ffdshow tryouts project: Discussion & Development

[onomatopellan]

Thanks there is no crash
Where can I download rev 3307?
You always can download the latest rev here. (thanks to xhmikosr)

http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1

[ikarad]

You always can download the latest rev here. (thanks to xhmikosr)

http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1

I tried rev3307 and It crash anew with YUY2.
with rev 3306 there is no crash.

[albain]

Also the build that I posted hold the new ffmpeg libraries that are unstable, so that (probably) explains the crashes you get

Please test again with regular builds.

About ffmpeg libs : I'll try to digg this around but this is difficult as long as I can't debug them.
There may be conversion methods that I didn't import, or a few things that I missed during the merge.

If there are candidates to this, here is the last patch (works on both 32 & 64 bits) :
http://damienbt.free.fr/ffdshow_ffmpeglibs_update2.patch

[ikarad]

With rev3306 it works well withtout any crash (edit: I had one crash with rev3306) but there are some problems with subtitles in the sample that i give you before.

1) some subs aren't displayed.
for example this sub is not displayed
148
00:10:09,692 --> 00:10:13,445
such speed, for such a large mass,
is quite unusual.

other example: between 10min34s and 10min 40 there is voice and no subtitle

2) Some subtitles appear with a time delay after the voice.

example, this subtitle appears with 1-2 seconds later
124
00:08:27,381 --> 00:08:28,590
Passing through the scorching sunshine
Storyvoards and Settings vy: Shinji Higuchi

3) some subtitles disappeared with a time delay
1
00:00:09,050 --> 00:00:12,720
'"2015 A.D. Perseus Arm, Milky Way Galaxy
(12,000 arsecs from Earth)'"

this sub must disappear at 12 seconds and in movie it disappears at 14 seconds.

I give you subtitle file extracted from bluray with suprip (there isn't all subs because ocr with suprip is not perfect)
http://www.zshare.net/info.html?73050006-7d75d90afaea102f4d92066e73874fde

sup file
http://www.zshare.net/info.html?73352376-976374543601cacbaa498463343cacfe

file containing image of all subs (extracted with suprip)
http://www.zshare.net/info.html?73352325-2a933698e98fb8caad63561bdd298763

example
part 1
http://www.zshare.net/info.html?72194780-3f3de5a523000d960412ca4e70f4959e
part2
http://www.zshare.net/info.html?72228704-870148d21c7371e7a2cb0c45e37239c1
part3
http://www.zshare.net/info.html?72228942-da539adec6856595c9fa215e89fcfdf2

part 4
http://www.zshare.net/info.html?72879609-e7005b47f97bd593cc1fac7d6620cda5

part 5
http://www.zshare.net/info.html?72894035-f4d916b0cbd3fbf999835142beff854c
part 6
http://www.zshare.net/info.html?72901975-036b32dfe47b4378533fe6974fb1237c

partie 7
http://www.zshare.net/info.html?72152693-cff6f6453a85c6f497a94c0fc61987f5

[tal.aloni]

albain,
v2 of the patch does not include the new files, only the modifications.

the build you posted based on v2 works well for me.

[JarrettH]

What should I do to confirm dxva is working in ffdshow? I've ticked off h264 in the dxva dialogue so far and am using ffmpeg-mt for the h264 decoder. Does a new tray icon for dxva video appear? Does it only work with EVR?

Running Windows 7, 7600 GT, ffdshow rev 3305, 196.21 drivers

??? :thanks:

[albain]

albain,
v2 of the patch does not include the new files, only the modifications.

the build you posted based on v2 works well for me.

Sorry, here is the new link
http://damienbt.free.fr/ffdshow_ffmpeglibs_update3.patch

and thanks for looking at it

And me too, I can't reproduce the crashes. Maybe I wronly compiled it...

[onomatopellan]

??? :thanks:
You should see this when dxva ffdshow is working. And yes, it only works with EVR in Vista/7
http://i49.tinypic.com/dg5v6x.png

[onomatopellan]

I have discovered with reliability monitor the reason of my crash with rev 3307. The fail module is libmplayer.dll

With output NV12 or YUY2 If Smoothing method for Bluray subs is "swscaler gaussian" then mpc-hc crash. If I select 'None','aproximate' or 'full' then it works. Output color (except rgb32) is still not good, color in rgb32 is good but it has some pixels on the top of the sub. example:
RGB32 output
http://i45.tinypic.com/2iizvkk.png

YUY2 output
http://i45.tinypic.com/2ikzll5.png

NV12 output (DXVA)
http://i45.tinypic.com/c3mup.png

[tal.aloni]

I have discovered with reliability monitor the reason of my crash with rev 3307. The fail module is libmplayer.dll
this is solved with albain v3 patch, IMO, it's stable enough to commit, this is the right direction.
(STaRGaZeR, could not recreate any of your crashes with my v3 build: http://iknowu.net/files/public/ffdshow/ffdshow_rev3307_20100305_dbt_.exe )

Output color (except rgb32) is still not good, color in rgb32 is good but it has some pixels on the top of the sub.

it's actually a complete single line, probably a tiny bug, it doesn't affect stability.
some colorspace conversion makes it break into several lines.

[onomatopellan]

Thanks tal.aloni. Indeed my crash disappeared with your v3 build. :D

[STaRGaZeR]

this is solved with albain v3 patch, IMO, it's stable enough to commit, this is the right direction.
(STaRGaZeR, could not recreate any of your crashes with my v3 build: http://iknowu.net/files/public/ffdshow/ffdshow_rev3307_20100305_dbt_.exe )

Weird, there must be something in my config that it's making it crash because a fresh installation of that build works fine. I'll try to find what's happening.

EDIT: Found it. When using default settings everything is OK. However, try to select Lanczos as the resizer and then change the number of taps. It'll crash on you with the libav error. Luma and chroma sharpen sliders don't work either. Actually "luma gaussian blur" is doing everything except blurring, I think that's actually the effect of luma sharpening. Are you sure these sliders work as they should after the patch?

[albain]

Weird, there must be something in my config that it's making it crash because a fresh installation of that build works fine. I'll try to find what's happening.

EDIT: Found it. When using default settings everything is OK. However, try to select Lanczos as the resizer and then change the number of taps. It'll crash on you with the libav error. Luma and chroma sharpen sliders don't work either. Actually "luma gaussian blur" is doing everything except blurring, I think that's actually the effect of luma sharpening. Are you sure these sliders work as they should after the patch?

You're right, I reproduce the crash.

The sliders should work and bring visible results

I'll have a look....after some sleeping

[STaRGaZeR]

Cool!

Have you guys seen that ffmpeg now decodes Bink video? :D

EDIT: clsid, with the lastest mingw64 (20100305) libavcodec vorbis works in x64.

[DigitalDeviant]

OK I tested the subtitle changes with 3307.

Blu-ray:
Still an issue with multiple subtitles
http://img534.imageshack.us/img534/2066/wb1.th.png (http://img534.imageshack.us/i/wb1.png/)
Sample (http://www.mediafire.com/?vthz5wtkihz)

DVD:
Subtitles are discolored.

Good
http://img13.imageshack.us/img13/316/gs2good.th.png (http://img13.imageshack.us/i/gs2good.png/)

Bad
http://img38.imageshack.us/img38/9990/gs2bad.th.png (http://img38.imageshack.us/i/gs2bad.png/)

[dann23]

Cool!

Have you guys seen that ffmpeg now decodes Bink video? :D

EDIT: clsid, with the lastest mingw64 (20100305) libavcodec vorbis works in x64.

how to use ffdshow to decode bink video?

[albain]

It crashes because of this :
swScaler: Exactly one scaler algorithm must be chosen

As a result the context is NULL. I am working on it

[ikarad]

With RGB 32, There are some problems with subtitles.

some subtitles are cut
example in this screen: the last word is cut
http://nsa13.casimages.com/img/2010/02/26/mini_100226113716480529.jpg (http://www.casimages.com/img.php?i=100226113716480529.jpg)

example of video that I give the page before

part 1
http://www.zshare.net/info.html?72194780-3f3de5a523000d960412ca4e70f4959e
part2
http://www.zshare.net/info.html?72228704-870148d21c7371e7a2cb0c45e37239c1
part3
http://www.zshare.net/info.html?72228942-da539adec6856595c9fa215e89fcfdf2

part 4
http://www.zshare.net/info.html?72879609-e7005b47f97bd593cc1fac7d6620cda5

part 5
http://www.zshare.net/info.html?72894035-f4d916b0cbd3fbf999835142beff854c
part 6
http://www.zshare.net/info.html?72901975-036b32dfe47b4378533fe6974fb1237c

partie 7
http://www.zshare.net/info.html?72152693-cff6f6453a85c6f497a94c0fc61987f5

https://sourceforge.net/tracker/?func=detail&aid=2959562&group_id=173941&atid=867360

There is the same problem with rev 3307 with patch v3.

With rev3306 it works well withtout any crash (edit: I had one crash with rev3306) but there are some problems with subtitles in the sample that i give you before.

1) some subs aren't displayed.
for example this sub is not displayed
148
00:10:09,692 --> 00:10:13,445
such speed, for such a large mass,
is quite unusual.

other example: between 10min34s and 10min 40 there is voice and no subtitle

2) Some subtitles appear with a time delay after the voice.

example, this subtitle appears with 1-2 seconds later
124
00:08:27,381 --> 00:08:28,590
Passing through the scorching sunshine
Storyvoards and Settings vy: Shinji Higuchi

3) some subtitles disappeared with a time delay
1
00:00:09,050 --> 00:00:12,720
'"2015 A.D. Perseus Arm, Milky Way Galaxy
(12,000 arsecs from Earth)'"

this sub must disappear at 12 seconds and in movie it disappears at 14 seconds.

I give you subtitle file extracted from bluray with suprip (there isn't all subs because ocr with suprip is not perfect)
http://www.zshare.net/info.html?73050006-7d75d90afaea102f4d92066e73874fde

sup file
http://www.zshare.net/info.html?73352376-976374543601cacbaa498463343cacfe

file containing image of all subs (extracted with suprip)
http://www.zshare.net/info.html?73352325-2a933698e98fb8caad63561bdd298763

example
part 1
http://www.zshare.net/info.html?72194780-3f3de5a523000d960412ca4e70f4959e
part2
http://www.zshare.net/info.html?72228704-870148d21c7371e7a2cb0c45e37239c1
part3
http://www.zshare.net/info.html?72228942-da539adec6856595c9fa215e89fcfdf2

part 4
http://www.zshare.net/info.html?72879609-e7005b47f97bd593cc1fac7d6620cda5

part 5
http://www.zshare.net/info.html?72894035-f4d916b0cbd3fbf999835142beff854c
part 6
http://www.zshare.net/info.html?72901975-036b32dfe47b4378533fe6974fb1237c

partie 7
http://www.zshare.net/info.html?72152693-cff6f6453a85c6f497a94c0fc61987f5

There is the same problem with rev 3307 with patch v3.

[albain]

I think this is fixed. Please tell me if the filters are working

New build (http://www.mediafire.com/file/nzjymyjtnjd/ffdshow_rev3307_20100306_dbt_newlibs_beta2.exe)


Patch v4 (http://www.mediafire.com/file/jdmfzqgnkke/ffdshow_ffmpeglibs_update4.patch)

[dann23]

I see now that ffmpeg supports bink and indeo 5 formats. Will ffdshow support this?

[roozhou]

I see now that ffmpeg supports bink and indeo 5 formats. Will ffdshow support this?
There's no working DShow bink splitter.
And there are more. e.g. cook, wmapro, monkey audio...

[albain]

The current bluray parser is buggy.

This subtitles format is a real pain !

I know what should be done but I would like to speed up my work by getting source code from existing and working bluray parsers.

I'll let you know

[STaRGaZeR]

I think this is fixed. Please tell me if the filters are working

New build (http://www.mediafire.com/file/nzjymyjtnjd/ffdshow_rev3307_20100306_dbt_newlibs_beta2.exe)


Patch v4 (http://www.mediafire.com/file/jdmfzqgnkke/ffdshow_ffmpeglibs_update4.patch)

These issues are still present, I've used a fresh install of ffdshow with everything at default values, just enabling each filter:

- The sliders issue in Resize is still present: luma gaussian blur controls luma sharpen instead, luma sharpen does nothing and can crash the player if you constantly move it, etc.
- ffmpeg deinterlacers like 5-tap lowpass, libav crash.
- Postprocessing, same as above. Funny thing is, it crashes with Xvid videos, not with H.264 material.
- Sharpen, everything works OK except swscaler, the black screen is gone but it doesn't do anything.
- In Blur, libavcodec error again when using mplayer temporal noise reducer.
- Luminance level fix in Picture properties, libavcodec crash.

These have been fixed:

- swscaler gaussian blur in Blur.
- Levels.

[albain]

These issues are still present, I've used a fresh install of ffdshow with everything at default values, just enabling each filter:

- The sliders issue in Resize is still present: luma gaussian blur controls luma sharpen instead, luma sharpen does nothing and can crash the player if you constantly move it, etc.
- ffmpeg deinterlacers like 5-tap lowpass, libav crash.
- Postprocessing, same as above. Funny thing is, it crashes with Xvid videos, not with H.264 material.
- Sharpen, everything works OK except swscaler, the black screen is gone but it doesn't do anything.
- In Blur, libavcodec error again when using mplayer temporal noise reducer.
- Luminance level fix in Picture properties, libavcodec crash.

These have been fixed:

- swscaler gaussian blur in Blur.
- Levels.

Oh yes, I messed up. I am fixing it

[ikarad]

The current bluray parser is buggy.

This subtitles format is a real pain !

I know what should be done but I would like to speed up my work by getting source code from existing and working bluray parsers.

I'll let you know

Good luck and I hope a perfect support of blu ray subs as soon as possible.

[albain]

I think that everything is fixed : libswscale and bluray subtitles thanks to Supread creator

New patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update5.patch)

New build (http://damienbt.free.fr/ffdshow_rev3307_20100306_dbt_newlibs_beta3.exe)

EDIT : there is still this problem with bluray subtitles using YUV colorspaces. I can't find why they are half-transparent

[onomatopellan]

Thanks albain! Now loading of external .sup subtitles with ffdshow would be very convenient for test and compare it with SUPread. Is this possible?

Also there still is the problem with line of pixels on the top of subtitles. They look perfect on SUPread.
http://forum.doom9.org/showthread.php?p=1380370#post1380370

[Gleb Egorych]

ForceWare 196.75 alert (http://www.nvidia.com/Download/index.aspx?lang=en-us)
We are aware that some customers have reported fan speed issues with the latest 196.75 WHQL drivers on NVIDIA.com. Until we can verify and root cause this issue, we recommend that customers do not download this driver. Instead, please stay with, or return to 196.21 WHQL drivers. Release 196.75 drivers have been temporarily removed from our website and we also are asking our partners and others to remove temporarily this 196.75 WHQL driver as well.
So it's safer to revert back to 196.21 WHQL or 196.34 beta.

[DigitalDeviant]

I think that everything is fixed : libswscale and bluray subtitles thanks to Supread creator

New patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update5.patch)

New build (http://damienbt.free.fr/ffdshow_rev3307_20100306_dbt_newlibs_beta3.exe)

EDIT : there is still this problem with bluray subtitles using YUV colorspaces. I can't find why they are half-transparent

It's still not working correctly with the multiple subtitle sample I posted. Both subtitles show completely now but not correctly positioned. They both appear at the top of the screen where the first is supposed to display and the first only displays for a split second until the second shows. The second subtitle should show at the bottom of the screen.

[albain]

I have fixed the crashing issue with DXVA + Fast forward/rewind.

The problem were reported to me while using media control and Haali splitter

@DigitalDeviant : okay, I think I know where it comes from. I think I take the placements for the next or previous subtitles. I already made this mistake at the beginning.

I am also interested in feedbacks using the FFDShow filters with the new libraries from my test build

Thanks

[anto]

Hi all,

I actually want to add PTS DTS on OSD Display.

Ca anybody could help me on how to deal with.

I've seen that PTS on Output are implemented in last versions of FFMPEG.

Thanks.

[STaRGaZeR]

I think that everything is fixed : libswscale and bluray subtitles thanks to Supread creator

New patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update5.patch)

New build (http://damienbt.free.fr/ffdshow_rev3307_20100306_dbt_newlibs_beta3.exe)

EDIT : there is still this problem with bluray subtitles using YUV colorspaces. I can't find why they are half-transparent

These issues are still present, like before I've used a fresh install of ffdshow with everything at default values, just enabling each filter:

- The sliders issue in Resize is still present: luma and chroma gaussian blur now work, but parameter, luma and chroma sharpen don't.
- ffmpeg deinterlacers like 5-tap lowpass, libav crash.
- Postprocessing, same as above.
- Sharpen, everything works OK except swscaler, the sliders and the filter itself do nothing.
- In Blur, libavcodec error again when using mplayer temporal noise reducer.
- Luminance level fix in Picture properties, libavcodec crash.

[albain]

Yes there is indeed some work left

ffmpeg scaler is not designed to handle different resize methods and values for luma and chroma so the ffdshow custom code breaks many things from the original code.

I am trying to put the pieces together : I now understand better how it works, I can see the end of the tunnel

Man, next time I'll do some translations, that's all lol
I don't know why I put myself into this situation lol

[STaRGaZeR]

You know we all love your work :D

[Mr VacBob]

Interesting changes should be submitted upstream so you don't have to maintain them ;)

[anto]

Hello,

I'm expecting just for little information, not a complete solution

I'm just expecting some help from anyone to find the Presentation Time Stamp in ffdshow because I've not found yet a valid PTS.

[hoborg]

Hi.
Just small question - is it normal that FFDshow reporting NV21 output instead of NV12 durning MPEG-2i decoding?

[tal.aloni]

Hello,

I'm expecting just for little information, not a complete solution

I'm just expecting some help from anyone to find the Presentation Time Stamp in ffdshow because I've not found yet a valid PTS.

we have "frame timestamp" under OSD.

[hoborg]

Hi.
Is there a way to use keyboard shortcut to swap between more audio/subtitles track (stream switcher enabled)? I didnt found it in keys and remote...

[anto]

we have "frame timestamp" under OSD.

I've ever seen it but frame timestamp aren't corresponding with real PTS, and there's no way to deduce PTS from it.

[THX-UltraII]

what output (for sound) renderer is recommended when you bitstream with the 5xxx series? directsound or the MPC audio renderer? (or another maybe?)

[albain]

New build with new libs/bluray subtitles support (http://www.mediafire.com/file/mutwkhv4wdd/ffdshow_rev3309_20100308_dbt_newlibs_beta4.exe)

Thank you for your feedbacks

[albain]

Hi.
Is there a way to use keyboard shortcut to swap between more audio/subtitles track (stream switcher enabled)? I didnt found it in keys and remote...

Which player are you using ?

[hoborg]

Which player are you using ?

Well, it is not about the player. This feature will alow switch betwen multiple subtitles/audio tracks undepended on player.

For example i use FFDShow on my HTPC based on GB-PVR. There is no support for any track selection in this software, it simply ask DirectShow to play video (like GraphStudio).

But i can map any keyboard shortcut on my remote-so such feture will be perfect solution i think.

[_xxl]

New build with new libs/bluray subtitles support (http://www.mediafire.com/file/mutwkhv4wdd/ffdshow_rev3309_20100308_dbt_newlibs_beta4.exe)

Thank you for your feedbacks

Cpu usage is very high, no filters enabled by default over 80-90%. Same file uses no more than 30% with normal ffdshow 3309.
Audio part wouldn't even load. Please create a new branch for all big changes in ffdshow. Thanks for your time!:)

[albain]

Branching doesn't work well, there are two many files

I gave up last time I tried (for HD audio)

I'll post the patch instead : here it is (http://damienbt.free.fr/ffdshow_ffmpeglibs_update6.patch)

I'll move on something else, because here I have no idea why the performances are so low

If someone wants to have a look at it, I think this is better and worth it if we are several to digg into this

libswscale now brings some interesting features (such as alpha scaling)

[STaRGaZeR]

New build with new libs/bluray subtitles support (http://www.mediafire.com/file/mutwkhv4wdd/ffdshow_rev3309_20100308_dbt_newlibs_beta4.exe)

Thank you for your feedbacks

I don't have any problems with perfomance, it's the same as before. But it's true that ffdshow audio doesn't load with the betas. I didn't mention it because I was testing the filters, but it's there.

- Resize is almost fixed! Bicubic produces green video, but everything else is working!
- ffmpeg deinterlacers like 5-tap lowpass, crash.
- Postprocessing, crash.
- Sharpen, everything works now!
- In Blur, libav error when using mplayer temporal noise reducer.
- Luminance level fix in Picture properties, libav crash.

[albain]

Well, it is not about the player. This feature will alow switch betwen multiple subtitles/audio tracks undepended on player.

For example i use FFDShow on my HTPC based on GB-PVR. There is no support for any track selection in this software, it simply ask DirectShow to play video (like GraphStudio).

But i can map any keyboard shortcut on my remote-so such feture will be perfect solution i think.

I asked this because some software implement this feature : MPC, media center+media control, media portal...

This should not be very complicated to do inside FFDSHow, the remote control API already holds the methods, this is just the GUI part which is missing


@Stargazer : thanks again for testing
I tested only the resize filter deeply, otherwise the other filters use the scaler but in a basic mode. This is rather libpostproc (which has also been updated) is used. However I didn't experience any crash with those. I'll make some deeper tests tomorrow if I have some time

[hoborg]

I asked this because some software implement this feature : MPC, media center+media control, media portal...

This should not be very complicated to do inside FFDSHow, the remote control API already holds the methods, this is just the GUI part which is missing

Sounds promising :)
:thanks:

[tal.aloni]

Cpu usage is very high

I don't have this issue with 'newlibs_beta4.exe', please try to identify the component responsible.

Thanks,
Tal

[STaRGaZeR]

@Stargazer : thanks again for testing
I tested only the resize filter deeply, otherwise the other filters use the scaler but in a basic mode. This is rather libpostproc (which has also been updated) is used. However I didn't experience any crash with those. I'll make some deeper tests tomorrow if I have some time

Like I said I only have the postprocessing crash with some videos, even with all the options forced, which is interesting to say the least. I'll test the new betas.

BTW, if someone is interested I've been compiling my own builds with this patch, which basically reverts r2976 since I've never faced such a situation and seeking is a lot faster with it reverted: http://www.mediafire.com/?ik2diohummm

[tetsuo55]

Like I said I only have the postprocessing crash with some videos, even with all the options forced, which is interesting to say the least. I'll test the new betas.

maybe you could provide debug treces or minidumps? and a sample that albain could use?

[DigitalDeviant]

Sorry to report I have no change in Blu-ray subtitle positioning. The sample still fails to position the second line at the bottom. DVD subtitles still look ugly as well.

[FreeFall]

First I would like to thank albain and all the other people who put their time and energy into working on these projects, your efforts are much appreciated.

Blu-Ray & IDX/SUB Subtitles

with DXVA disabled turning on subtitles crashes with swscaler gaussien (fault libavcodec.dll), other methods work without crashing.

Blu-Ray Disc Sword Of The Stranger (Anime)

wrong subtitle colours and cleanup needed (broken pixel lines above text), coloured subtitles should be blue.

YV12 & YUY2 output subs are orange, RGB output subs are yellow.

My System Setup:

Zoom Player Home Max 7.10 Alpha 3 EVR renderer Vista Business x32
Haali Matroska splitter
Gabest Mpeg Splitter
FFDshow 3309 20100308 dbt newlibs beta4


FreeFall

[albain]

I have fixed the crashes.

Now I am looking at issues with bluray subs

[albain]

Here is a new build with the fixes (http://damienbt.free.fr/ffdshow_rev3309_20100308_dbt_newlibs_beta5.exe)

Nothing done on subtitles side : still issues with semi-transparent subtitles in YUV mode and sometimes some outline of the subtitle box that appears.

Patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update7.patch)

[XhmikosR]

r3310 x64 ffmpeg compilation is broken:

libavcodec/fft.o:fft.c:(.text.unlikely+0x1de): undefined reference to `ff_fft_init_mmx'
collect2: ld returned 1 exit status
h:\progs\MSYS\mingw\bin\x86_64-w64-mingw32-dllwrap.exe: h:\progs\MSYS\mingw\bin\x86_64-w64-mingw32-gcc exited with status 1
make: *** [../../bin/libavcodec.dll] Error 1

H:\progs\Compiling\ffdshow\src\ffmpeg>

[STaRGaZeR]

I'll test the lastest beta tomorrow, I don't have time right now.

[TinTime]

Hi,

I've been experimenting with ffdshow to resize and convert to RGB during playback with very positive results so far.

However I've run into a problem with resizing. I'm using:
Resize to screen resolution
Process pixel aspect ratio internally
Keep original aspect ratio
Bilinear resize

I've also set the borders to a medium grey to prevent screen burn.

The problem is that before adding the borders it seems to pad the image to (at a guess) mod16 with black pixels. This leads to a black line down the right hand side of the image. Strangely though this black line disappears if I turn on the OSD. See the images below. They're not the same frame but it should show what I mean.

OSD off (http://i39.tinypic.com/2e51v2b.jpg)
OSD on (http://i43.tinypic.com/258v12f.jpg)

This seems to happen with all versions I've tried over the past day, including revision 3309.

Is this a bug or is there a setting somewhere I'm unaware of?

Thanks very much.

[pankov]

TinTime, guys,
how does one change the color of the blavk bars?

[TinTime]

The slider under:

Resize & aspect -> Borders -> Brightness of borders

[pankov]

10x
I like my borders black but a friend of mine wants them lighter ... he even wants to change the color to something not gray.
I don't think it's possible now, right?
btw
what about adding a feature to draw them dynamically change color depending on the current frame - something like the ambilight function of Philips TVs.
I know you guys, developers, have more important tasks right now, and I'm very grateful that you keep improving this great filter, so think of this like a very low priority feature request.

[blackEyEz]

Hello, maybe this has been adressed already but i have a problem with MKV's with multiple soundstracks (DTS/AC3) and the MPC splitter. I use MediaPortal and these movies dont seem to run. I got a tip of using streamswitcher but that didnt work. Could this be a FFDshow issue?

[jeremy33]

This file don't work with mpc with internal codecs or ffdshow but play perfectly with vlc.

http://www.megaupload.com/?d=GZEXML84

And second problem if I play a flac file in mpc with ffdshow audio prefered decoder and mpc internal flac codec unchecked ffdhsow don't want to decode a flac file.

[Atak_Snajpera]

This file don't work with mpc with internal codecs or ffdshow but play perfectly with vlc.

http://www.megaupload.com/?d=GZEXML84

remux this file in tsmuxer

[jeremy33]

It works but it's playing too fast 60fps instead of 25fps.

[Atak_Snajpera]

force frame rate in tsmuxer

[jeremy33]

No change too fast

[Atak_Snajpera]

it works ok on my ffdshow r3005 + MPC-hc (all internal decoders disabled)

[jeremy33]

if I remux with tsmuxer (forced 25 fps) and if I unchecked MPEG PS/TS/PVA source filter it work so it's a mpc-hc problem.

Another problem, if I play a flac file in mpc with ffdshow audio prefered decoder and mpc internal flac codec unchecked ffdhsow don't want to decode a flac file.

[Qaq]

Another problem, if I play a flac file in mpc with ffdshow audio prefered decoder and mpc internal flac codec unchecked ffdhsow don't want to decode a flac file.
use madFlac for flac (disable internal filters and codecs in MPC and ffdshow)

[sacharja]

I always have crashes with the new FFDShow DXVA codec when hardware acceleration is enabled. It always occurs on h.264 files when I seek very fast. ZPlayer just closes than. However, it's working with the old h.264 FFDShow codec.

rev3316 (and all further). Any ideas on that?

[rsd78]

I always have crashes with the new FFDShow DXVA codec when hardware acceleration is enabled. It always occurs on h.264 files when I seek very fast. ZPlayer just closes than. However, it's working with the old h.264 FFDShow codec.

rev3316 (and all further). Any ideas on that?

I know Albain is aware of this (this is with haali right?), as I reported it to him and he tried to fix it a few revisions ago but it still seems to happen.

I know he is very busy but I think he will make an attempt at fixing it if possible.

In the meantime, I know using the mpc/gabest splitter it does not crash. However this presents other problems related to using Albain's mediacontrol plugin for subtitles if you use windows media center. But seems like you are using another player, so try mpc/gabest splitter.

[86sixsix]

mpc standalone filters 1735, ffdshow 3305, win7_64_ult_rtm.

playing dts hd ma 7.1 24bit 96khz out over analogue - gets parsed as 48khz (unknown bit-depth) and everyone sounds like they're on helium.

also, if using 'mpc audio renderer' and the 'exclusive' checkbox is ticked in windoze mixer deely... i get odd pauses. unchecked - plays fine.

sample: http://www.trailerz.demo-world.eu/hd/hd_dts_orchestra_short_lossless.rar

using mpegsplitter.ax only dts-core gets pulled so be aware - but remuxing to mkv and using matroskasplitter.ax gets the 8 channels but only at 48khz.

same occurs after demuxing and just playing the .dts audio track.

if bitstreaming is selected - file gets correctly parsed at 96000 but i haven't got hdmi out and it rightly just sits there laughing at me.

i haven't got any 96khz truehd tracks to test that with.

how can i tell if it's a problem with the splitter or with ffdshow?

if it's ffdshow - can i request a 'i don't have hdmi, but let's have a go at 96k and see what happens' checkbox'?

EDIT: Folk might have to go to: http://www.demo-world.eu/trailers/high-definition-trailers.php and find the dts 7.1 samples...

[sacharja]

I know Albain is aware of this (this is with haali right?), as I reported it to him and he tried to fix it a few revisions ago but it still seems to happen.
No, I don't think that it's related to Haali. It's an h.264 AVI file and I don't use Haali for avi files.

However, FFDShow DXVA with hardware acceleration enabled is completely useless to me because I always encode my files as h.264 avi files and I can't seek fast in any of them :eek:

[albain]

No, I don't think that it's related to Haali. It's an h.264 AVI file and I don't use Haali for avi files.

However, FFDShow DXVA with hardware acceleration enabled is completely useless to me because I always encode my files as h.264 avi files and I can't seek fast in any of them :eek:

No, the problem comes from ffdshow and I thought I fixed it but I didn't. This is complicated to figure out because it occurs in ffmpeg DXVA code. I have not forgot but I just have too many things in parallel

[DoomzDayz]

I'm having problems with the mplayer noise average algorithm and high resolution source material

My ffdshow chain is resize to 1080P -> Add noise (to simulate film grain, was having problems with AVISynth). With 720P and higher source material, the noise has problems and is mostly on the right half. With less than 720P, the noise is uniform (well, I am using gaussian Smile across the video.

With average unchecked, the noise looks fine. Also, when I set it to generate noise before resizing, it works fine. The other noise algorithms work fine (no averaging allowed), changing the resize method doesn't change anything. Appears in mpc-hc and wmp with overlay, VMR7/9, and Haali. RAM and CPU usage less than 50%. beta7_rev3154_20091209.

Q6600 @ 3.2 GHz (doesn't change when going to stock)
nVIDIA 9600GSO
WinXP


Example (not always quite this bad):
http://img687.imageshack.us/img687/7915/noise.th.png (http://img687.imageshack.us/i/noise.png/)

[TFM_TheMask]

Can someone tell me what has changed between version rev 3154 and rev 3316 with AC3 and DTS Passthrough?

With version rev 3154 I could extract AC3 and DTS audio from a movie by setting the passthroughDTS and passthroughAC3 to 1 (in registry) with my application TFM Audio Tool but with rev 3316 this doesn't work anymore. DTS size and duration is about 2 minutes longer than with the rev 3154 version. This also applies to AC3 extraction.


And what does the other Passthrough options do? And does this also work with extracting audio using the FileWriter filter connected to the ffdshow audio decoder?

[sacharja]

No, the problem comes from ffdshow and I thought I fixed it but I didn't. This is complicated to figure out because it occurs in ffmpeg DXVA code. I have not forgot but I just have too many things in parallel
OK, thanks. If we can help let us know ;)

[Mixer73]

However, FFDShow DXVA with hardware acceleration enabled is completely useless to me because I always encode my files as h.264 avi files and I can't seek fast in any of them :eek:

Why would you create h.264 AVI files rather than MP4?

[tetsuo55]

Can someone tell me what has changed between version rev 3154 and rev 3316 with AC3 and DTS Passthrough?

With version rev 3154 I could extract AC3 and DTS audio from a movie by setting the passthroughDTS and passthroughAC3 to 1 (in registry) with my application TFM Audio Tool but with rev 3316 this doesn't work anymore. DTS size and duration is about 2 minutes longer than with the rev 3154 version. This also applies to AC3 extraction.


And what does the other Passthrough options do? And does this also work with extracting audio using the FileWriter filter connected to the ffdshow audio decoder?Someone else will have to provide more detail, but in a nutshell: Bitstreaming was refactored to support HD audio formats, and along the way a lot of bugs where found and fixed.

[TFM_TheMask]

Someone else will have to provide more detail, but in a nutshell: Bitstreaming was refactored to support HD audio formats, and along the way a lot of bugs where found and fixed.

Thanks tetsuo55,

Seems that also some bugs were created in the process of implementing this new HD audio support :p.

Does this mean that when a movie has DTS-HD Master Audio that I can write it to file when connecting ffdshow audio encoder to the Filewriter filter (with PassthroughDTSHD set to 1 in registry) or do I need special hardware for it to work?

Can someone else comment on the wrong duration with PassthroughAC3 and PassthroughDTS.

[zummi]

Hi there, i ran into a little WAF-issue today... On releases after rev3292 the MediaControl plugin doesn't detect that ffdshow is running. All builds before works fine... I'm running Win7x64 w/FFDSHOW x64. Don't know if this is a ffdshow or mediacontrol-issue though...but i'll try here first. Thanks :-)

[albain]

Hi there, i ran into a little WAF-issue today... On releases after rev3292 the MediaControl plugin doesn't detect that ffdshow is running. All builds before works fine... I'm running Win7x64 w/FFDSHOW x64. Don't know if this is a ffdshow or mediacontrol-issue though...but i'll try here first. Thanks :-)

You could try to disable the "enable media experience" checkbox in media control settings.
You can post your problem on media control forum with the logs in debug level
I am very busy at work right now, so don't expect an answer very soon

Thanks tetsuo55,

Seems that also some bugs were created in the process of implementing this new HD audio support :p.

Does this mean that when a movie has DTS-HD Master Audio that I can write it to file when connecting ffdshow audio encoder to the Filewriter filter (with PassthroughDTSHD set to 1 in registry) or do I need special hardware for it to work?

Can someone else comment on the wrong duration with PassthroughAC3 and PassthroughDTS.

Few people use this feature (write to file) and it has not been retested since the HD audio bitstream implementation. I'll take a look at it, but you should post a bug on the tracker so that we keep a trace on it

This is very difficult right now, we are only 2/3 active developers at the present moment

@Tetsuo : are you aware of this issue affecting the MPC m2ts splitter ? the subtitle streams are not detected until the first subtitles block is parsed (a few seconds before it is displayed). Then it appears within the context menu.
I saw this issue with bluray m2ts files (so holding PGS subtitles)

[rsd78]

@Tetsuo : are you aware of this issue affecting the MPC m2ts splitter ? the subtitle streams are not detected until the first subtitles block is parsed (a few seconds before it is displayed). Then it appears within the context menu.
I saw this issue with bluray m2ts files (so holding PGS subtitles)

Albain, this is also applicable to MKVs using the mpc-splitter too right? I believe this is the same issue I showed you when using the mpc-splitter and media control and ffdshow dxva.

[tetsuo55]

we are aware of A problem but have not found the cause.

Is it this bug? http://sourceforge.net/apps/trac/mpc-hc/ticket/16

We definately need help fixing the bugs in the mpeg splitter

[Sebastiii]

we are aware of A problem but have not found the cause.

Is it this bug? http://sourceforge.net/apps/trac/mpc-hc/ticket/16

We definately need help fixing the bugs in the mpeg splitter

Hi Tetsuo55,

It's also : http://sourceforge.net/apps/trac/mpc-hc/ticket/162

And my duplicate bug : http://sourceforge.net/apps/trac/mpc-hc/ticket/307

Thanks.

[tetsuo55]

ok i will triage that into the older ticket

[Betsy25]

Is the MPEG2 decoder (libavcodec) broken in rev 3158 ?

I get broken images like below, or video/audio out of sync.

http://i44.tinypic.com/i23lg1.jpg

[djesteban]

Ok, I think I found a pretty nasty bug...

I noticed that in clsid build 3267 (x64) and higher, there's a lot of image corruption/artifacts while watching h264 footage (in a mkv container in my case). I was able to detect the problem in multiple files (if not all); it seems to happen at sporadic times in each video, but when the problem occurs, it can always be replicated if you rewind to that particular part of the given file (if it happens at 6 min, and you rewind to that 6th min, you will see the problem occurring over and over again with what seemed to me as the same artifacts).
Please note that this problem seems to have appeared in build 3267 (ffdshow_rev3267_20100213_clsid_x64); the build before that, build 3222 (ffdshow_rev3222_20100123_clsid_x64) was unaffected. All build after 3267 are plagued with the issue (at least up until build 3316)
This should be tag important since it really affect the viewing experience; makes footage almost unwatchable. See the attached jpg for an example. I had a hard time pausing on one of those artifact, so I was able to get an example by pressing printscreen until I got one "captured".

Tested on 2 computers, and also I tested the same footage in VLC and it works perfectly.

I already logged a bug for this, but thought that it may get a little more visibility here. Anyone wants to try to repro this?

*EDIT* Here's the image, it doesn't seem to appear when I attach it
http://img532.imageshack.us/img532/9320/corruption01.jpg

[TFM_TheMask]

Few people use this feature (write to file) and it has not been retested since the HD audio bitstream implementation. I'll take a look at it, but you should post a bug on the tracker so that we keep a trace on it

This is very difficult right now, we are only 2/3 active developers at the present moment


Thanks for the answer Albain. I will post a bug report about this on the tracker. Can you also answer my other question:

Does this mean that when a movie has DTS-HD Master Audio that I can write it to file when connecting ffdshow audio encoder to the Filewriter filter (with PassthroughDTSHD set to 1 in registry) or do I need special hardware for it to work?

[Andrey /MAG/]

Is it possible to get output dimensions of video from ffdShow filter by remote control API?
I try to use constant IDFF_OSDtype_outputSize (1521), but it returns a zero.
I can't use deci.getOutputDimensions because of deci is need to be initalized.
How can I do it with Interface of filter loaded by another process (player)?
I have only ffdShow window handle (and player's PID).
Language: Delphi.

[clsid]

@djesteban
Please upload some sample files.

[rpm7200]

ffdshow's mixer doesn't work correctly. when i cut lfe channel it doesn't work. please fix it.

[rsd78]

Hi Tetsuo55,

It's also : http://sourceforge.net/apps/trac/mpc-hc/ticket/162

And my duplicate bug : http://sourceforge.net/apps/trac/mpc-hc/ticket/307

Thanks.

Tetsuo, I think I'm seeing the same issue with MKVs (specifically using ffdshow dxva + media control).

[djesteban]

@djesteban
Please upload some sample files.

@clsid
OK, I'm going to try sampling a part of a mkv where I am sure I can see the issue... and upload it somewhere. I'll try to get 1-2 example at least.

*EDIT*
@clsid
Ok, for some reason, I demuxed a whole 7Gb mkv, made a sample of the h264 stream using DGIndexNV and remuxed that in a mkv container and using this sample I was not able to reproduce the problem while playing it back. So after that, I decided to remux the whole 7Gb stream alone in a new mkv (to make sure it wasn't the soundtrack or something else in the original mkv that is causing the issue) and after doing that I am able to reproduce the problem... So I really don't know how long my sample needs to be for you in order to be able to repro it.
Now my question is: Do you have a connection that permits you to download large files? (i.e. 7Gb). If yes, let me know and I'll point you to where you can download the same file I am using. Now, please note that this happens with multiple (if not all animation/anime) mkv that I have... coming from multiple source... some that I have encoded, some that have been encoded by other people.
I really hope this problem can be fixed... I checked the changelog and I noticed that there's quite some changes between 3222 and 3267, but I cannot get all those builds... I really wish I could pinpoint where exactly in the changelog this has happened, but I can tell you for sure it's between 3222 and 3267

[clsid]

Maybe DGSplit can help to create a smaller sample:
http://forum.doom9.org/archive/index.php/t-143963.html

[HeadlessCow]

You can use mkvmerge to split mkvs as well. You might have better luck with that. In mkvmergeGUI just look on the second tab.

[djesteban]

Maybe DGSplit can help to create a smaller sample:
http://forum.doom9.org/archive/index.php/t-143963.html

:)
I will try making my sample using that, but like I said, my problem is not to make a sample, I was able to do that with DGIndexNV, the problem is that it seems it takes more than a small sample to repro the problem. If my sample is too small, I cannot reproduce the issue.

*EDIT*
Tried making the sample with DGSplit, and same thing happens; if the sample is too small, the problem doesn't happen...
Would 490 Mb be a ok size for you clsid? That is the smallest file where I can see it happening...
If you want that file, are you on IRC? Or let me know if I can PM you and send you the link or something.

[clsid]

Use http://www.mediafire.com/

[djesteban]

Use http://www.mediafire.com/

Can I upload a 490Mb file there?

[Atak_Snajpera]

I found small bug in installer

options untouched
http://img41.imageshack.us/img41/5523/new2s.png

but

isCompMgr=0 is not created in registry [SOFTWARE\GNU\ffdshow] . Result ffdshow shows compatibility manager for any unknown file (for example avs2yuv.exe or x264_x86.exe)

Can I upload a 490Mb file there?
zip your file in 100 mb parts

[djesteban]

zip your file in 100 mb parts

Yeah, that's what I figured.

@clsid
would it be ok if I upload it on netload instead of mediafire? I have an account there... let me know if that is fine with you, else I'll up it on mediafire...

[Atak_Snajpera]

It does not matter send your sample wherever you want :)

[djesteban]

It does not matter send your sample wherever you want :)

ok :)

Uploading now, should be done in an hour or something... my upload speed is... :(

[djesteban]

@clsid and Atak
Click here to get the files (http://netfolder.in/folder.php?folder_id=RW0b0fZ)
Let me know if there's any problem.

[JohnTitor]

Hey.
I just tried the same video as djesteban. The latest build (rev3316) really screws the image up - anime movies are basically unwatchable (I also tried build rev 3267, it glitches exactly the same way).

Live action seems ok, tough.

If I revert to build revision 3222, everything is ok.

Let me know if you want screenshots, but it looks exactly the same as what esteban posted.

[djesteban]

Hey.
I just tried the same video as djesteban. The latest build (rev3316) really screws the image up - anime movies are basically unwatchable (I also tried build rev 3267, it glitches exactly the same way).

Live action seems ok, tough.

If I revert to build revision 3222, everything is ok.

Let me know if you want screenshots, but it looks exactly the same as what esteban posted.

...I have to say, I've looked at 30 min of live action footage and indeed, I was not able to see any image corruption... so for some odd, crazy reason, that seems to affect primarily anime/animation/cartoon footage (?!?) if that makes any sense... it has too, I mean it was working fine before...

@clsid
Have you been able to check my sample I uploaded? Let me know if you have any question.

Cheers

[STaRGaZeR]

Here is a new build with the fixes (http://damienbt.free.fr/ffdshow_rev3309_20100308_dbt_newlibs_beta5.exe)

Nothing done on subtitles side : still issues with semi-transparent subtitles in YUV mode and sometimes some outline of the subtitle box that appears.

Patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update7.patch)

Sorry for being so late albain, here are the results of the test with newlibs_beta5:

- Resize with bicubic still produces green video, everything else is fixed.
- Deinterlacing, ffmpeg deinterlacers produce a extremely stuttering image, it doesn't look like a perfomance problem, CPU usage is at ~10% while without deinterlacing is at 20% (1080p H.264 high bitrate). That's with the default YUY2 output, with RGB32 I get a black screen. Yadif and the other deints are good.
- Postprocessing, with mplayer checked and YUY2 output, same behavior as deinterlacing: stuttering. RGB32, black screen. Crashes fixed!
- In Blur, stuttering with YUY2 and black screen with RGB32 again when using mplayer temporal noise reducer, it crashed once with error in libavcodec.
- Luminance level fix, stuttering and black screen with YUY2 and RGB32.

I didn't notice any other problems, all the crashes except that ocasional one are gone. Good work, you're almost there!

[clsid]

I have pinpointed the h264 problem to revision 3256. I will attempt to fix it later.

[albain]

Hi,

here is a new development build (http://damienbt.free.fr/ffdshow_rev3319_20100318_dbt_newlibs_beta6.exe)with the new ffmpeg scaling/postprocessing libraries.

Fixes : now the DVD subtitles are okay. The fix (committed in previous revision) as well as the new fixes should resolve the last crashing issues that could occur
Bluray subs may be cleaner too

I have also added new functions to the keyboard control : there is now fast forward/rewind commands, and next subtitle stream & next audio stream commands.


Please report back, thanks !

@Stargazer : I don't reproduce the issues you listed in the last build

[hoborg]

Hi,

I have also added new functions to the keyboard control : there is now fast forward/rewind commands, and next subtitle stream & next audio stream commands.

Please report back, thanks !

Great work!

BTW, it will be good to be able to setup "prefered" audio/subtitle track language :)

[hoborg]

@albain:

Some quick test with your test build:
- open any MKV video in grapstudio (using MPC-HC MKV splitter), play and click on FFDShow icons = crash
- it looks like FFDshow now refuse to accept AAC audio streams
- subtitles color are white (in graphstudio) even if i setup to use yellow
- hitting CTRL+ALT+F4 cause graphstudio crash instead of next subtitle stream
- FFMPEG-MT causing h.264 video stuttering, libavcodec play fine here on IntelAtom CPU.

...do not have time for more test, hope my report will help you.
Thanks again for your work.

EDIT:
here is test file (http://hobring.esero.net/saf/ffdshow_test.zip) i am working with:

[djesteban]

I have pinpointed the h264 problem to revision 3256. I will attempt to fix it later.

Are you talking about the problem I reported? Jut want to make sure cause you didn't quote :P

If yes, that's awesome, hope it won't be too hard to fix

[sacharja]

#3319 is still crashing while seeking when hardware acceleration is enabled.

[clsid]

Are you talking about the problem I reported? Jut want to make sure cause you didn't quote :P

If yes, that's awesome, hope it won't be too hard to fix
Yes, it is the problem you reported.
Fixing will require some trial and error, but should be doable.

[TFM_TheMask]

Thanks for the answer Albain. I will post a bug report about this on the tracker. Can you also answer my other question:

Does this mean that when a movie has DTS-HD Master Audio that I can write it to file when connecting ffdshow audio encoder to the Filewriter filter (with PassthroughDTSHD set to 1 in registry) or do I need special hardware for it to work?

Nobody to answer my question above?

[STaRGaZeR]

Hi,

here is a new development build (http://damienbt.free.fr/ffdshow_rev3319_20100318_dbt_newlibs_beta6.exe)with the new ffmpeg scaling/postprocessing libraries.

Fixes : now the DVD subtitles are okay. The fix (committed in previous revision) as well as the new fixes should resolve the last crashing issues that could occur
Bluray subs may be cleaner too

I have also added new functions to the keyboard control : there is now fast forward/rewind commands, and next subtitle stream & next audio stream commands.


Please report back, thanks !

@Stargazer : I don't reproduce the issues you listed in the last build

Testing is done as always, clean install of ffdshow with everything default, then activating each filter at a time. The results of beta 6 are the same as beta 5. I can reproduce them with all renderers, codecs, etc. so I don't know why you can't.

BTW, the issue with that white shadow in the DVD subs that I uploaded earlier is still there, but I don't know if that's supposed to be fixed in this beta.

Oh and thanks for the new keyboard control features, they're very useful.

[albain]

Nobody to answer my question above?

This should work as always : in file mode there is a different behaviour for writing the content of the input stream instead of encapsulating it in bitstream format.

However this does not apply to TrueHD : there is special code for this format. So in the TODO list

[TFM_TheMask]

Ok, looking forward to it.

[albain]

Testing is done as always, clean install of ffdshow with everything default, then activating each filter at a time. The results of beta 6 are the same as beta 5. I can reproduce them with all renderers, codecs, etc. so I don't know why you can't.

BTW, the issue with that white shadow in the DVD subs that I uploaded earlier is still there, but I don't know if that's supposed to be fixed in this beta.

Oh and thanks for the new keyboard control features, they're very useful.

I noticed this white shadow but I thought that this was normal. Until now the alpha channel was buggy on DVD subs so I don't know if this is normal or not.

EDIT : ok, I nailed it, this was a bug. This will be fixed in the next release
However I don't understand why I can't reproduce your stutterring and black screen issues...

[STaRGaZeR]

For example:

Everything default (output is YUY2 when using EVR), deinterlacing checked and using 5-tap lowpass (I forced it because the source is progressive, but it doesn't matter):

http://thumbnails25.imagebam.com/7254/87a21372534035.gif (http://www.imagebam.com/image/87a21372534035)

The same, with RGB32 as output. No stuttering but black screen:

http://thumbnails27.imagebam.com/7254/56c45772534039.gif (http://www.imagebam.com/image/56c45772534039)

The bicubic green screen:

http://thumbnails14.imagebam.com/7254/903af472534414.gif (http://www.imagebam.com/image/903af472534414)

[clsid]

@Atak_Snajpera
Fixed in 3322

@djesteban
Fixed in 3320

[albain]

Ok, thanks I reproduce it. Don't know why I didn't before

Except the stutterring in YUY2 mode, don't understand why. I thought that it could come from the cpu optimizations but you have an Intel CPU like me.

Color problems are probably due to missing colorspace conversions that I did not import

RGB32 : I don't know why yet but it should be easy to fix (I can see a few pictures before it becomes black)

[Betsy25]

Is the MPEG2 decoder (libavcodec) broken in rev 3158 ?

I get broken images like below, or video/audio out of sync.

http://i44.tinypic.com/i23lg1.jpg

50 MB sample of problematic behaviour of MPEG2 decoder (libavcodec) on .VOB files.

http://www.mediafire.com/?yynu455wtkz

[NiFa]

How can I set wanted audio profile to autoload only when playing files from specified folder? I tried volume name match but it didn't work.

[albain]

@Betsy25: this problem is indeed annoying
I don't have time to work on it. I hope that clsid or Tal will

@Stargazer : this time I think this is good
The bicubic scaler is fixed
The deinterlacers are fixed.

I don't reproduce the stutterring so I hope that this is fixed too

Here is the link (http://damienbt.free.fr/ffdshow_rev3322_20100318_dbt_newlibs_beta7.exe)

New patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update8.patch)

@developers
It would be great to release a new beta very soon.
Many things have been fixed since beta7 : DXVA2, HD audio, bluray subtitles support

I have nearly finished my work on libswscale and libpostproc libraries, but the commit should done only after this new beta.
What do you think ?

[fastplayer]

@developers
It would be great to release a new beta very soon.
Many things have been fixed since beta7 : DXVA2, HD audio, bluray subtitles support

I have nearly finished my work on libswscale and libpostproc libraries, but the commit should done only after this new beta.
What do you think ?
I'm for it if we drop the beta-tag.

[clsid]

I am currently working on automatically using the SVN revision as part of the ffdshow version, like is done in MPC-HC.

Does anyone know what rebase.exe does? It doesn't seem to actually be used right now. So I am thinking of just removing it.

[Blight]

Hi,
Did something get changed/broken in the FFDShow API?, trying to set a subtitle file using the "IffDecoder2:putParamStrW" function causes ffdshow to throw an Access Violation error.

Also, how do you get BluRay subtitles to load? I tried feeding a BluRay image '.m2ts' file to the gabest mpeg source filter (from MPC-HC) and connecting the video to video pin and the subtitle to text pin (no errors), connected the audio pin to the ffdshow audio decoder and default directsound audio renderer.

I could get the connections working in graph edit, but no subs were showing in the video renderer (I checked the ffdshow subtitle and bluray subtitle configuration pages and everything was enabled).

I the main movie file from both "SpaceBalls" and "The Fifth Element".

[STaRGaZeR]

@Betsy25: this problem is indeed annoying
I don't have time to work on it. I hope that clsid or Tal will

@Stargazer : this time I think this is good
The bicubic scaler is fixed
The deinterlacers are fixed.

I don't reproduce the stutterring so I hope that this is fixed too

Here is the link (http://damienbt.free.fr/ffdshow_rev3322_20100318_dbt_newlibs_beta7.exe)

New patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update8.patch)

@developers
It would be great to release a new beta very soon.
Many things have been fixed since beta7 : DXVA2, HD audio, bluray subtitles support

I have nearly finished my work on libswscale and libpostproc libraries, but the commit should done only after this new beta.
What do you think ?

- Bicubic is fixed, however changing the parameter slider produces green garbage on the screen. Also I didn't notice this earlier, Gauss also produces green vertical lines when the parameter slider is over 5.70, so does 2xSaI.
- Deinterlacing, RGB32 still black screen. With YUY2 they work, but only if the content is interlaced. If you force it with a progressive source ffdshow outputs the correct frame rate, but dropping one of each two frames. The result is double speed but same number of frames. I think I know the reason of the stuttering: resolution. When tested with interlaced SD video they work fine, with interlaced HD video they also work fine but I get that stuttering. CPU usage is normal, actually lower than with deinterlacing disabled.
- Postprocessing looks like it doesn't work anymore, I see no changes whatsoever in the image.
- In Blur, libavcodec crash with YUY2 and black screen with RGB32 when using mplayer temporal noise reducer.
- Luminance level fix, it works now with YUY2, but with the double speed thingy. RGB32 is still black.
- I don't know if this beta includes the fix to the white shadow in DVD subs, but it's still there. I've found that it only happens with RGB32 output, YUV is fine.

- NEW ISSUE: NV12 output, black screen with no filters enabled.

BTW, I posted this in the HD audio thread but got ignored, I sometimes get silence when seeking in m2ts files with TrueHD audio, decoded with libavcodec, no bitstreaming. It didn't happen before, but I don't know the exact revision. Has anyone reproduced this?

[albain]

Hi,
Did something get changed/broken in the FFDShow API?, trying to set a subtitle file using the "IffDecoder2:putParamStrW" function causes ffdshow to throw an Access Violation error.

Also, how do you get BluRay subtitles to load? I tried feeding a BluRay image '.m2ts' file to the gabest mpeg source filter (from MPC-HC) and connecting the video to video pin and the subtitle to text pin (no errors), connected the audio pin to the ffdshow audio decoder and default directsound audio renderer.

I could get the connections working in graph edit, but no subs were showing in the video renderer (I checked the ffdshow subtitle and bluray subtitle configuration pages and everything was enabled).

I the main movie file from both "SpaceBalls" and "The Fifth Element".

Check out the last version I posted on HD audio thread

FFDShowAPI has not changed since a while. I'll take a look at it sometimes.

@Stargazer : thank you for your big todo list ;-)

[_xxl]

I have nearly finished my work on libswscale and libpostproc libraries, but the commit should done only after this new beta.
What do you think ?
Yes, we should release a new beta soon.

[dann23]

will someone add support for mpeg2 in dxva?

[tal.aloni]

regarding releasing a new beta, I personally think that some of the new features are still a work in progress, and I still want to implement a default white list for DXVA like clsid suggested.
(we have to do this in order to avoid problems with many programs)

I think that the updated swscale library should be included and tested before releasing the next beta, but if you all think differently than it's ok by me.

[ikarad]

Check out the last version I posted on HD audio thread

FFDShowAPI has not changed since a while. I'll take a look at it sometimes.

@Stargazer : thank you for your big todo list ;-)

I try 3326 (version on xvidvideo.ru) rev and It crash when I select bluray subs with YUY 2.

With 3306 it worked
I tried rev3307 and It crash anew with YUY2.
with rev 3306 there is no crash.

Before 3306 it didn't worked. I don't try ffdshow between 3308 and 3325 included.

EDIT: I try also 3222-dbt-newlibs-beta8 and there is the same problem.
EDIT 2: I try with RGB32 and there is no crash but with YUY2 there is crash.

[Jeremy Duncan]

ffdshow 3326: http://www.mediafire.com/?lftymn35zy2

patch (http://www.mediafire.com/?hezjwyrniym)

The patch offers frame accurate seeking in ffdshow avisynth when normally it wouldn't be available with some filters and settings.

[STaRGaZeR]

Albain, after applying your beta 8 patch to a clean svn checkout I get the following two errors when compiling:

Error 40 fatal error C1083: Cannot open include file: 'postproc/postprocFilters.h': No such file or directory e:\ffdshow\src\settings\filters\TpostprocSettings.cpp 21
Error 42 fatal error C1083: Cannot open include file: 'postproc/swscale.h': No such file or directory e:\ffdshow\src\mplayer\Tlibmplayer.cpp 24

These two files were removed by the patch itself, the entire src/mplayer/postproc folder is empty. Maybe that's why I didn't see any effect when enabling postprocessing? Did you do a full rebuild of ffdshow?

Also, after putting these two files in there and compile, if you open ffdshow properties when playing a DVD for example (weird because it happens with some files, not all of them), the whole list of filters is missing and the player hangs:

http://thumbnails6.imagebam.com/7289/fb29f772880265.gif (http://www.imagebam.com/image/fb29f772880265)

[Sebastiii]

ffdshow 3326: http://www.mediafire.com/?lftymn35zy2

patch (http://www.mediafire.com/?hezjwyrniym)

The patch offers frame accurate seeking in ffdshow avisynth when normally it wouldn't be available with some filters and settings.

Thanks :)

[ikarad]

With 3326 and rgb32 output I find several problems with blu rays subtitles (I can't test with YUY2 because It crash).
1)first problem: some subs are cut (Same problem with 3306 and other ffdshow version untill 3326)
For example there is this sub at 8min33
"Your back is heavy as I catch up to you"

http://nsa14.casimages.com/img/2010/03/21/mini_100321091909495937.png (http://www.casimages.com/img.php?i=100321091909495937.png)
In the video the "you" is cut
http://nsa15.casimages.com/img/2010/03/21/mini_100321092046662446.jpg (http://www.casimages.com/img.php?i=100321092046662446.jpg)

Same problem with this sub at 8min46
in the video "It feels so great when I run throu"
The original sub is
It feels so great when I run through the wind
http://nsa14.casimages.com/img/2010/03/21/mini_100321092808826318.png (http://www.casimages.com/img.php?i=100321092808826318.png)

2) second problemThere is some problems with the time of displaying of subtitles.
For example in this video (that I also give you) The first sub :
1
00:00:09,050 --> 00:00:12,720
'"2015 A.D. Perseus Arm, Milky Way Galaxy
(12,000 arsecs from Earth)'"

is well displayed.
Although the second sub:
http://nsa14.casimages.com/img/2010/03/21/mini_100321090446191729.png (http://www.casimages.com/img.php?i=100321090446191729.png)
the fourth wave of ennemy photons torpedoes is approching
is displayed and is removed immediately. There is no time to read it and it's not normal.

Same problem with this sub
128
00:08:39,143 --> 00:08:44,272
The heat of your body reaches mine
Directors of Photography: Masahide Okino
In the movie it disappeared at 00:08:41 and not at 00:08:44,272

There many other examples in the sample that I give you.

3) Third problem: If I use resize function (lanczos) of ffdshow: the display of subs is very very bad. All subs are displayed in the high left corner instead of the middle of screen and many subs are cut. Sound is cut at some moments.



I give you subtitle file extracted from bluray with suprip (there isn't all subs because ocr with suprip is not perfect)
http://www.zshare.net/info.html?73050006-7d75d90afaea102f4d92066e73874fde

sup file
http://www.zshare.net/info.html?73352376-976374543601cacbaa498463343cacfe

file containing image of all subs (extracted with suprip)
http://www.zshare.net/info.html?73352325-2a933698e98fb8caad63561bdd298763

example
part 1
http://www.zshare.net/info.html?72194780-3f3de5a523000d960412ca4e70f4959e
part2
http://www.zshare.net/info.html?72228704-870148d21c7371e7a2cb0c45e37239c1
part3
http://www.zshare.net/info.html?72228942-da539adec6856595c9fa215e89fcfdf2

part 4
http://www.zshare.net/info.html?72879609-e7005b47f97bd593cc1fac7d6620cda5

part 5
http://www.zshare.net/info.html?72894035-f4d916b0cbd3fbf999835142beff854c
part 6
http://www.zshare.net/info.html?72901975-036b32dfe47b4378533fe6974fb1237c

partie 7
http://www.zshare.net/info.html?72152693-cff6f6453a85c6f497a94c0fc61987f5

[albain]

@Stargazer : the patch is buggy indeed, I have to make a new one

I am suffering with this new libs, I have fixed the bicubic and gauss resize issues, but I still don't understand why postprocessing does not work

@Ikarad : ok I'll take a look at it this week, thanks for your reports

[STaRGaZeR]

clsid, after r3323 the icl10 project file has an error:

http://thumbnails25.imagebam.com/7298/5babdd72979526.gif (http://www.imagebam.com/image/5babdd72979526)

As you can see, ffdshow is configured to x64 when buiding for Win32, plus it's actually skipped because it's not selected.

[djesteban]

@Atak_Snajpera
Fixed in 3322

@djesteban
Fixed in 3320

Thanks! I tested this thoroughly with build 3326 and it now work as expected... no weird image corruption! Happy you could find the culprit!

[Eliminateur]

in the RGB wiki help page: http://ffdshow-tryout.sourceforge.net/wiki/video:rgb_conversion there's a link about dithering for img comparison: http://forum.doom9.org/showthread.php?p=1287242#post1287242 which states it tries to make the same result as with deband.
Unfortunately, dithering does not work as expected and has little to no effect(nowhere near deband which produces a "smooth" transition of colors) and increases CPU load enough for my machine not to be able to smoothly play 1080p H264 (Xeon 3075 @3GHz).

It would be interesting to see if the algorithm needs a revision as deband is quite cpu hungry, ¿is it multithreaded?, ¿is there any other filter with the same effect and better performance?.

Here are the screenshots showing the difference very clearly(at least in my monitor which has HC range), aoutput is set to RGB32 HQ conversion:
No deband or dither:
http://img707.imageshack.us/img707/7039/nodebandordither.th.png (http://img707.imageshack.us/i/nodebandordither.png/)

Dither only:
http://img402.imageshack.us/img402/6498/ditheronly.th.png (http://img402.imageshack.us/i/ditheronly.png/)

Deband 1.22 and Dither:
http://img201.imageshack.us/img201/6779/debandanddither.th.png (http://img201.imageshack.us/i/debandanddither.png/)

i haven't included a deband-only image as the difference with deband+dither is negligible(the bands are clearly seen and the number of bands is the same, the difference is that some bands are lighter/darker)

it would be good to update the wiki link or a better explanation to prevent banding

[madshi]

in the RGB wiki help page: http://ffdshow-tryout.sourceforge.net/wiki/video:rgb_conversion there's a link about dithering for img comparison: http://forum.doom9.org/showthread.php?p=1287242#post1287242 which states it tries to make the same result as with deband.
Unfortunately, dithering does not work as expected and has little to no effect(nowhere near deband which produces a "smooth" transition of colors)
The purpose of dithering in the YCbCr -> RGB conversion is to not introduce any additional banding, respectively to not make source banding more evident/ugly. Basically using dithering produces a more correct image compared to not using dithering. However, dithering (as implemented in the YCbCr -> RGB conversion) does not in any way reduce banding which is hard coded into the source material. In your images you can see that dithering does not reduce the source's banding. However, in the right top of the images the banding steps are more equally spread when using dithering compared to when not using dithering. So it seems to me that dithering does work as intended.

The "deband" post processing aims at detecting and removing banding which is hard coded into the source material. So "deband" has a different purpose compared to the RGB conversion dithering. "deband" is a more complicated algorithm and is expected to consume more CPU power compared to dithering.

I do think the wiki should be changed to make things less confusing. Currently it reads:

"Dithering is an option to enhance visual quality. It reduces banding that RGB conversion tends to add."

I'd replace that with:

"Dithering is an option to maintain the source's full visual quality when doing YCbCr -> RGB conversion. Without dithering RGB conversion can result in added banding."

[fastplayer]

I do think the wiki should be changed to make things less confusing. Currently it reads:

"Dithering is an option to enhance visual quality. It reduces banding that RGB conversion tends to add."

I'd replace that with:

"Dithering is an option to maintain the source's full visual quality when doing YCbCr -> RGB conversion. Without dithering RGB conversion can result in added banding."
Done.

[albain]

Is it normal that the postprocessing filter is applicable only to the following codecs :
Raw codecs, then :
case CODEC_ID_MPEG1VIDEO:
case CODEC_ID_MPEG2VIDEO:
case CODEC_ID_LIBMPEG2:
case CODEC_ID_MPEG4:
case CODEC_ID_MSMPEG4V1:
case CODEC_ID_MSMPEG4V2:
case CODEC_ID_MSMPEG4V3:
case CODEC_ID_H263:
case CODEC_ID_SVQ1:
case CODEC_ID_FLV1:
case CODEC_ID_INDEO2:
case CODEC_ID_INDEO3:
case CODEC_ID_XVID4:
case CODEC_ID_MJPEG:
case CODEC_ID_MJPEGB:
case CODEC_ID_MSVIDEO1:
case CODEC_ID_CINEPAK:
case CODEC_ID_VP5:
case CODEC_ID_VP6:
case CODEC_ID_VP6F:


For example it is disabled for H264 formats

[STaRGaZeR]

I don't think that's normal, you should be able to apply postprocessing to all sources if you want to.

[fastplayer]

For example it is disabled for H264 formats
Yep, doesn't make much sense for it to be enabled for H.264 content.

[Snowknight26]

You mean disabled?

[fastplayer]

No, I mean enabled. Post-processing an already post-processed (deblocked) video doesn't make much sense to me.

[Eragon4ever]

The deblocking is part of any valid decoding. Therefore it is no post-processing IMHO. Filters should not be concerned by what the decoding of a certain codec does.

[Px]

Is it normal that the postprocessing filter is applicable only to the following codecs :
In general case it is normal, strange is that DIVX missing in this list

[STaRGaZeR]

I think you guys have a confusion between two deblocking filters. Albain is not talking about H.264's inloop deblocking, but the deblocking filter inside ffdshow's postprocessing filter. Nothing to do with each other. ffdshow's postprocessing filter is like any other filter like resize or subtitles, you apply it if you want to. Any restrictions would be illogical.

[albain]

That's right, there are presets to apply those restrictions

Hello guys, I need some help : I spent a lot of time on this, I still don't understand why the postprocessing filter does not work correctly (I get a black picture when enabled)

Here is the source patch (http://damienbt.free.fr/ffdshow_ffmpeglibs_update10.patch), if one of you could have a look at it

libswscale seems to be okay now
libpostproc seems to be working on the deinterlacer part, but not on the deblocking (postprocessing) part.

The parameterts that are sent to the filter seem to be fine, so this has to be something inside the postprocess methods

Thank you for your help

[_xxl]

In general case it is normal, strange is that DIVX missing in this list
Isn't missing:
case CODEC_ID_MPEG4
case CODEC_ID_XVID4

[_xxl]

I think you guys have a confusion between two deblocking filters. Albain is not talking about H.264's inloop deblocking, but the deblocking filter inside ffdshow's postprocessing filter. Nothing to do with each other. ffdshow's postprocessing filter is like any other filter like resize or subtitles, you apply it if you want to. Any restrictions would be illogical.

Was removed in 2006 because H.264 has internal inloop deblocking and additional postprocessing wouldn't help.

[STaRGaZeR]

Was removed in 2006 because H.264 has internal inloop deblocking and additional postprocessing wouldn't help.

Internal inloop deblocking is not a sustitute of a deblocking filter. Plus an H.264 stream can be encoded without it for any obscure reason, and it will probably look like garbage without any postprocessing.

EDIT: albain, your patch is password protected.

[albain]

Internal inloop deblocking is not a sustitute of a deblocking filter. Plus an H.264 stream can be encoded without it for any obscure reason, and it will probably look like garbage without any postprocessing.

EDIT: albain, your patch is password protected.

Sorry : here is the right link (http://damienbt.free.fr/ffdshow_ffmpeglibs_update10.patch)

[_xxl]

I think you guys have a confusion between two deblocking filters. Albain is not talking about H.264's inloop deblocking, but the deblocking filter inside ffdshow's postprocessing filter. Nothing to do with each other. ffdshow's postprocessing filter is like any other filter like resize or subtitles, you apply it if you want to. Any restrictions would be illogical.

We should ask H.264 experts on this forum.

[Px]

albain
As you checking the postprocessing filter, please take a look on SPP Deblocking (and fast SPP deblocking in particular case), with full postprocessing enabled it very slow even on modern cpus, while on builds from near 2006-2008 it was a lot faster.
Of course, I taking into account grown of resolution in video files :)

[clsid]

regarding releasing a new beta, I personally think that some of the new features are still a work in progress, and I still want to implement a default white list for DXVA like clsid suggested.
(we have to do this in order to avoid problems with many programs)Do you have any estimate when you will have time to work on this?

I have made an initial list of the most commonly used players:
"bsplayer.exe;ehshell.exe;gom.exe;graphedit.exe;graphedt.exe;graphstudio.exe;graphstudio64.exe;kmplayer.exe;mpc-hc.exe;mpc-hc64.exe;mplayerc.exe;wmplayer.exe;zplayer.exe;"

[STaRGaZeR]

Sorry : here is the right link (http://damienbt.free.fr/ffdshow_ffmpeglibs_update10.patch)

Good, some issues fixed as you said: bicubic, gauss, the hangs and the missing list of filters fixed too :)

We should ask H.264 experts on this forum.

http://mewiki.project357.com/wiki/X264_Settings#no-deblock

Check "--no-deblock". Also note that x264's ultrafast preset has --no-deblock in it.

[tal.aloni]

@clsid

since I develop features in parallel, I have a main trunk directory and several non-svn directories.
with your latest changes, I can't build any of them.... :(

what should we do?


1>Performing Pre-Build Event...
1>SubWCRev : 'd:\Development\ffdshow\3330' is not a working copy
1>Compiling resources...



Edit:
maybe something like this?

@if ERRORLEVEL 6 goto :NoSubWCRev


p.s.
why update_version is pre-build event?
isn't it only needed for innosetup?
Tal

[clsid]

albain
As you checking the postprocessing filter, please take a look on SPP Deblocking (and fast SPP deblocking in particular case), with full postprocessing enabled it very slow even on modern cpus, while on builds from near 2006-2008 it was a lot faster.
Of course, I taking into account grown of resolution in video files :)The 'preset' slider controls how many passes the SPP deblocker performs. Set it to for example 2 or 3, and it will be much lighter on the CPU.

[clsid]

@clsid

since I develop features in parallel, I have a main trunk directory and several non-svn directories.
with your latest changes, I can't build any of them.... :(

what should we do?


1>Performing Pre-Build Event...
1>SubWCRev : 'd:\Development\ffdshow\3330' is not a working copy
1>Compiling resources...

Edit:
maybe something like this?

@if ERRORLEVEL 6 goto :NoSubWCRev


p.s.
why update_version is pre-build event?
isn't it only needed for innosetup?
Tal
The SVN revision is used in the file resource too (and in the future perhaps also on the About page in settings). So that is why it is done pre-build.

I am fixing it now.

[tal.aloni]

I am fixing it now.

Thanks,
for TortoiseSVN 1.6.7 (latest), the error code returned is 6 (SVN error), 10 doesn't work (seems like a bug on their side).
adding both 6 and 10 will be the best solution here.

Edit:
IF ERRORLEVEL construction has one strange feature, it returns TRUE if the return code was equal to or higher than the specified errorlevel.

here is the suggested fix:

@if ERRORLEVEL 6 if not ERRORLEVEL 7 goto :NoSubWCRev
@if ERRORLEVEL 10 if not ERRORLEVEL 11 goto :NoSubWCRev

[clsid]

In that case I will change it to:
if %ERRORLEVEL%==6
if %ERRORLEVEL%==10

[rsd78]

@developers
It would be great to release a new beta very soon.
Many things have been fixed since beta7 : DXVA2, HD audio, bluray subtitles support

I have nearly finished my work on libswscale and libpostproc libraries, but the commit should done only after this new beta.
What do you think ?

Albain, I apologize if I'm coming off as nagging (which is not my intent), but I'd love it if the issues with DXVA + ff/rew could be fixed before putting out an official beta (isn't that an oxymoron? :) )

To summarize for others (@Tetsuo for the mpc-hc splitter issue ?), using a player other than mpc-hc (i.e. Windows Media Center) using dxva + haali + ff/rew crashes. Using dxva + gabest + ff/rew works however gabest splitter (at least for mkvs) fails to consistently "see" the subtitle/audio streams (using Albain's Media Control).

[tal.aloni]

In that case I will change it to:
if %ERRORLEVEL%==6
if %ERRORLEVEL%==10

seem to work as well.
Thanks!

[stax76]

@clsid

Please add StaxRip to the whitelist.

[tal.aloni]

this patch makes the compatibility manager ("show dialog when an unknown application tries to load ffdshow") independent and not shared between ffdshow / DXVA / audio decoder,

we also enabled here a short DXVA whitelist.

http://iknowu.net/files/public/ffdshow/DXVA/DXVA-Compatibility.patch

[albain]

Albain, I apologize if I'm coming off as nagging (which is not my intent), but I'd love it if the issues with DXVA + ff/rew could be fixed before putting out an official beta (isn't that an oxymoron? :) )

To summarize for others (@Tetsuo for the mpc-hc splitter issue ?), using a player other than mpc-hc (i.e. Windows Media Center) using dxva + haali + ff/rew crashes. Using dxva + gabest + ff/rew works however gabest splitter (at least for mkvs) fails to consistently "see" the subtitle/audio streams (using Albain's Media Control).

I don't reproduce the crash anymore in the last revisions. Tested it in Windows 7 x64 + media center + Haali + DXVA mode + FF/RW on a H264 720p mkv

[rsd78]

I don't reproduce the crash anymore in the last revisions. Tested it in Windows 7 x64 + media center + Haali + DXVA mode + FF/RW on a H264 720p mkv

Thanks Albain, I will test again. I was using not the latest version of Haali before so I will try to duplicate.

Any chance of finding a work around for the subs/audio stream issue with the mpc-hc mkv splitter? I'd prefer to use this splitter long term since it supports the HD Audio better.

Thanks again, for all your hard work.

[clsid]

@stax76
Doing so right now.

@tal
isCompMgrChanged is now obsolete right? If so, that could also be removed from the code.

[Px]

The 'preset' slider controls how many passes the SPP deblocker performs. Set it to for example 2 or 3, and it will be much lighter on the CPU.
I knew about that, but "before" it works in 6 or 5 passes on most videos, and now, at twice powerful cpu - with only 3 or 4 :)

[albain]

Thanks Albain, I will test again. I was using not the latest version of Haali before so I will try to duplicate.

Any chance of finding a work around for the subs/audio stream issue with the mpc-hc mkv splitter? I'd prefer to use this splitter long term since it supports the HD Audio better.

Thanks again, for all your hard work.

I have applied a fix (rather a workaround) to media control, I have to post the new builds

[rsd78]

I have applied a fix (rather a workaround) to media control, I have to post the new builds

Awesome, thank you again Albain.

[albain]

FYI, I posted new media control builds on sourceforge and my website (6.0.4)

[rsd78]

FYI, I posted new media control builds on sourceforge and my website (6.0.4)

@Albain,

I upgraded to MediaControl 6.04, Ffdshow 3334, Gabest splitter 1764, latest Haali (I'm using Shark's pack basically), and unfortunately I"m having the same issues. This is on Win 7 x32. NOTE: All the issues I'm seeing are only when using ffdshow dxva and not ffdshow software decoding (everything works as expected in this case).

I tested a couple of 720p MKVs and I get crashes still using haali if I ff/rew for any length of time.

With the mpc-splitter, I still have the issues with it detecting the subs. Context menu doesn't pop up (because it didn't detect the individual embedded subs), but even when I manually go into the Sub/Audio stream selection menu of Media Control it simply has the Checkbox for "Show Subtitles". I can turn off/on subs correctly from there, however it fails to enumerate the embedded subs (i.e. it's just blank, it doesn't list any of the embedded subs. I haven't tried with external subs).

Hopefully someone else can confirm what I'm seeing, but if you need logs/info please let me know. Sorry this has been a persistent problem, but I do appreciate the effort you have put into it.

EDIT: Also I noticed a small Media Control bug I think. It kept failing to detect ffdshow dxva running even after applying the Recommended Configuration. I looked into the settings and noticed that in the ffdshow dxva control/inputs settings were not set to use "Remote Control API" and the "User Defined Settings". I think your "Apply Min/Recommended Changes" buttons seemed to only do this for regular ffdshow. Once I made these changes MediaControl detected ffdshow dxva.

EDIT2: Another note/bug with MC. From your front page the Download button goes to an error page (not found). I had to download it manually through Sourceforge. Also on Sourgeforge the current version listed is 6.02, I had to go to show all files to get to 6.04.

[Mtz]

I'm using Subtitle Workshop from many years.
Starting with ffdshow rev. 3314 I always have this error:
http://thumbnails4.imagebam.com/7352/1af96d73515640.gif (http://www.imagebam.com/image/1af96d73515640)
so I cannot use Subtitle Workshop to sync the subtitles.
Please fix this error because I cant use any newer version of ffdshow after rev. 3309.

enjoy,
Mtz

[albain]

Ok, I'll fix those 2

About DXVA + Haali + FF/RW : I'll make some deeper tests with 720p mkv to reproduce it

The only difference with you is that I own windows 7 x64

[STaRGaZeR]

albain, if you still don't have a good solution for it can you tell me what to change in the code to disable the only 2 change of codecs when dealing with TrueHD and AC3 for the time being? A lot of European Blu-ray titles have a TrueHD English track and then all the others in standard DD, and changing between them is driving me insane, having to reopen the movie. Or at least can you tell me how to increase the number of allowed changes? Thanks!

[rsd78]

Ok, I'll fix those 2

About DXVA + Haali + FF/RW : I'll make some deeper tests with 720p mkv to reproduce it

The only difference with you is that I own windows 7 x64

Thanks Albain.

Out of curiosity, any idea why the latest MediaControl did not fix that issue with the mpc-splitter? What was it supposed to do?

[TFM_TheMask]

@Albain: Did you by any chance had a look at the bug in the dts and ac3 passtrough when connecting it to the filewriter filter to extract the original AC3 or DTS stream?

[albain]

@Stargazer : a good solution must be thought according to the following.
- NewSegment event is raised when a there is a seek or when there is a stream switching
- The same stream can hold AC3 blocks & TrueHD blocks or DTS blocks with DTSHD blocks. So the audio parser must be put in the situation where the blocks it scans may not correspond to the resulting stream format (may be TrueHD if meets AC3 block or DTSHD if meets DTS core block)

So the right solution would be to scan for the stream format only a few seconds after a newsegment is raised. The solution is here but I don't have the time for now (as for new ffmpeg libraries)

Currently the audio parser stops scanning for the stream format after 2 changes which is not the best solution I agree

@rsd78 : I have changed the way media control scans the streams.
Formerly the streams were scanned for only once, now they are scanned anytime you raise the a menu that needs access to it.
The MPC bug is not fixed yet, this is a workaround : so if the streams don't appear at first, this is because the MPC splitter did not detect them. You have to wait a few seconds that the splitter detects them and then raise the Media Control menu


@TFM_TheMask : no, sorry

[STaRGaZeR]

@Stargazer : a good solution must be thought according to the following.
- NewSegment event is raised when a there is a seek or when there is a stream switching
- The same stream can hold AC3 blocks & TrueHD blocks or DTS blocks with DTSHD blocks. So the audio parser must be put in the situation where the blocks it scans may not correspond to the resulting stream format (may be TrueHD if meets AC3 block or DTSHD if meets DTS core block)

So the right solution would be to scan for the stream format only a few seconds after a newsegment is raised. The solution is here but I don't have the time for now (as for new ffmpeg libraries)

Currently the audio parser stops scanning for the stream format after 2 changes which is not the best solution I agree

Thanks for the detailed explanation. Can you tell me how to change ffdshow's code to stop scanning after, let's say, 50 changes instead of 2? If it's complicated forget about it, but if it's changing a 2 for a 50 somewhere in the code please tell me so I can live with my own builds until you have time to do the right thing :D

[rsd78]

@rsd78 : I have changed the way media control scans the streams.
Formerly the streams were scanned for only once, now they are scanned anytime you raise the a menu that needs access to it.
The MPC bug is not fixed yet, this is a workaround : so if the streams don't appear at first, this is because the MPC splitter did not detect them. You have to wait a few seconds that the splitter detects them and then raise the Media Control menu


Thanks Albain. Unfortunately it still doesn't seem to work for me. I played an mkv that had 5+ sub streams in it, waited 60 seconds into it, got into the menu and still did not show the individual sub streams.

[albain]

Thanks Albain. Unfortunately it still doesn't seem to work for me. I played an mkv that had 5+ sub streams in it, waited 60 seconds into it, got into the menu and still did not show the individual sub streams.

Do the streams appear within the MPC splitter menu ?

[albain]

Thanks for the detailed explanation. Can you tell me how to change ffdshow's code to stop scanning after, let's say, 50 changes instead of 2? If it's complicated forget about it, but if it's changing a 2 for a 50 somewhere in the code please tell me so I can live with my own builds until you have time to do the right thing :D

It is located inside TaudioParser.cpp. I don't have the code here but there are comments to describe the behaviour (I think this is located at the end of the file)

[albain]

I'm using Subtitle Workshop from many years.
Starting with ffdshow rev. 3314 I always have this error:
http://thumbnails4.imagebam.com/7352/1af96d73515640.gif (http://www.imagebam.com/image/1af96d73515640)
so I cannot use Subtitle Workshop to sync the subtitles.
Please fix this error because I cant use any newer version of ffdshow after rev. 3309.

enjoy,
Mtz

Do you have the exact faulty revision ? It would help a lot if you could narrow it down a little more

Only ffmpeg updates were made between rev 3309 and 3314

[rpm7200]

when i enable lfe croosover libavcodec is using 32 bit floating point but when i don't use lfe croosover, libavcodec is using 16 bit integer. is this a bug?

[albain]

Thanks for the detailed explanation. Can you tell me how to change ffdshow's code to stop scanning after, let's say, 50 changes instead of 2? If it's complicated forget about it, but if it's changing a 2 for a 50 somewhere in the code please tell me so I can live with my own builds until you have time to do the right thing :D

Hi again, can you try this build (http://damienbt.free.fr/ffdshow_rev3335_201011_dbt_.exe) and let me know if this fixes this issue ?

[rsd78]

Do the streams appear within the MPC splitter menu ?

If I right click on the ffdshow dxva tray icon and then go to Subtitle Sources, yes I do see the list of individual subs while playing a movie.

After doing that I checked with Media Control and it still didn't enumerate those same subs from the Media Control Menu (doesn't list any individual streams).

[STaRGaZeR]

Hi again, can you try this build (http://damienbt.free.fr/ffdshow_rev3335_201011_dbt_.exe) and let me know if this fixes this issue ?

No, it doesn't :(. In an MKV with AC3 and THD, changing from the AC3 track to the THD track results in silence, ffdshow reports 621 kbps ac3 (liba52). If I change back to the AC3 track I get silence too, with the same message. Info & CPU is empty in both cases.

[albain]

No, it doesn't :(. In an MKV with AC3 and THD, changing from the AC3 track to the THD track results in silence, ffdshow reports 621 kbps ac3 (liba52). If I change back to the AC3 track I get silence too, with the same message. Info & CPU is empty in both cases.

Okay, I'll make a build with debug traces

I can't test it by myself because I don't have any HD receiver for now

EDIT : I also forgot something in my code, it should work with this build (I hope)

Here is the new link (http://damienbt.free.fr/ffdshow_rev3335_20100327_dbt_.exe) : build that should fix audio stream switching

Also I am going to post a build with the new ffmpeg swscale and postprocessing libraries : I don't reproduce the black picture with postprocessing enabled anymore (!). I just made a clean/rebuild all. Maybe I had some crap that was not up to date

Here is the link (http://damienbt.free.fr/ffdshow_rev3335_20100327_dbt_newlibs_beta9.exe) for the ffdshow build with the new ffmpeg libraries

Thank you for your feedbacks

[STaRGaZeR]

With the audio build: AC3-->THD results in silence, ffdshow reports 6xx kbps ac3 (liba52). But now if I change back to the AC3 track I get sound. It's like ffdshow doesn't detect the format change. Oh I forgot to say, I use analog audio, no bitstreaming so I can't test that.

I will report my findings with the new libraries beta in a moment.

EDIT, newlibs report:

RGB32:

- Luminance fix in picture properties, black screen. But wait, if you enable deband and you place it after picture properties everything is fine! The filter works and the image is not black. Maybe this can help you with debugging.
- Deinterlacing, same as luminance fix, it works perfectly fine but only if you put it before deband and you activate it, otherwise black screen with ffmpeg deints.
- Postprocessing with mplayer checked, I get black screen. Again if I activate deband after PP it works perfectly fine.
- mplayer temporal noise reducer in Blur, same as above, with deband after it it works fine.
- White shadow in DVD subs, still present. Not a problem with the libs, but still there.

YUY2:

With SD content, exact same issues as RGB32, but changing the black screens with the double speed but same number of frames issue I reported earlier. With HD content, exact same issues as RGB32 but changing the black screens with stuttering. Activating deband after the filters fixes the double speed and the stuttering.

NV12: black screen always. This time deband does nothing.

It looks like everything works OK with deband, maybe that filter does something that should be added in the other filters too. Find this thing and everything will work like a charm (at least the bugs I've found) :)

EDIT 2: Log (http://www.mediafire.com/?onzkyyhdmmr). MKV file with AC3 as primary audio. Then at about second 45 I change to the THD track, sound is gone. Then again at second 55 I change back to the AC3 track, sound is heard again. This log is done with "ffdshow_rev3335_20100327_dbt_.exe"

[albain]

Ok thanks, please post the debugview logs of the tests with AC3 <=> TrueHD switches

[video_magic]

@ rsd78:

Did you know there is a new Haali Splitter? 27-03-10

http://haali.su/mkv/

[albain]

Thanks stargazer

Here is a new build (http://damienbt.free.fr/ffdshow_rev3335_20100327_dbt_.exe) on the audio part

I disabled the jitter correction for TrueHD & MLP codecs

About the build with the new libraries, I don't understand, I really don't reproduce the black screen issues anymore. Although I used the same sample as before (which caused me black screen). The only thing that changed is I had to reformat my drive (so new fresh installation)

Have you tried a clean install ?

[albain]

@Ikarad : I had to reformat my disk and I lost your sample. Can you reupload part2 (the others are still downloadable) ?

[STaRGaZeR]

Thanks stargazer

Here is a new build (http://damienbt.free.fr/ffdshow_rev3335_20100327_dbt_.exe) on the audio part

I disabled the jitter correction for TrueHD & MLP codecs

About the build with the new libraries, I don't understand, I really don't reproduce the black screen issues anymore. Although I used the same sample as before (which caused me black screen). The only thing that changed is I had to reformat my drive (so new fresh installation)

Have you tried a clean install ?

Are you sure this is the new build? The filename is the same, the download was instantaneous as if the build was cached in my browser and the log still shows jitter correction when I change to the THD track. The results are the same as before. It looks like ffdshow didn't change to the THD decoder.

About the newlibs beta, yes, all the tests are with clean installs as always. Deband after every filter with problems fixes them.

[ikarad]

@Ikarad : I had to reformat my disk and I lost your sample. Can you reupload part2 (the others are still downloadable) ?

here the part 2
http://www.zshare.net/info.html?74314450-ba2c28d26a51156a2b08c17414f5f336

[hoborg]

@albain:
Hi.
Tested 3336 build from http://www.xvidvideo.ru/.
How "...subtitle/audio streams switching" supose to work?
Audio switcher and keys & remotes are enabled, MPC-HC mkv source filter is in use, but pressing CTRL+ALT+F4 or nunlock 1 simply does nothing for me, no OSD info.

Tested on this sample (http://hobring.esero.net/saf/ffdshow_test.zip) in graphstudio.
BTW, FF/RW working fine.

[Sebastiii]

No, it doesn't :(. In an MKV with AC3 and THD, changing from the AC3 track to the THD track results in silence, ffdshow reports 621 kbps ac3 (liba52). If I change back to the AC3 track I get silence too, with the same message. Info & CPU is empty in both cases.

Hi, Can you provide a sample ?
Thanks,
Seb.

[kolak]

Is there a way to force makeavis to create YUY2 file (with YUY2 header) for a software that reads files natively (can't access directshow or VFW filters)?

Thx,
Andrew

[Casshern]

v3336 breaks sp/dif passthrough on some dts cores of dts master audio tracks. Before if you just enabled dts passthrough and disabled dts hd passthrough ffdshow would pass the dts core just fine to the sp/dif output. This sometimes does not work anymore (e.g. GI Joe Bluray main m2ts), ffdshow decodes the dts hd stream and does not extract the core:
Filter : ffdshow Audio Decoder - CLSID : {0F40E1E5-4F79-4988-B1A9-CC98794E6B55}

- Connected to:

CLSID: {1365BE7A-C86A-473C-9A41-C0A6E82C9FA3}
Filter: G:\GIJOE\BDMV\STREAM\01002.m2ts
Pin: Audio

- Connection media type:

Audio: DTS 48000Hz 6ch 1509kbps

AM_MEDIA_TYPE:
majortype: MEDIATYPE_Audio {73647561-0000-0010-8000-00AA00389B71}
subtype: MEDIASUBTYPE_DTS {E06D8033-DB46-11CF-B4D1-00805F6CBBEA}
formattype: FORMAT_WaveFormatEx {05589F81-C356-11CE-BF01-00AA0055595A}
bFixedSizeSamples: 1
bTemporalCompression: 0
lSampleSize: 1
cbFormat: 18

WAVEFORMATEX:
wFormatTag: 0x2001
nChannels: 6
nSamplesPerSec: 48000
nAvgBytesPerSec: 188719
nBlockAlign: 2012
wBitsPerSample: 0
cbSize: 0 (extra bytes)

pbFormat:
0000: 01 20 06 00 80 bb 00 00 2f e1 02 00 dc 07 00 00 . ..€»../á..Ü...
0010: 00 00

Interestingly other titles with the same info work! Also GI Joe works with older versions of ffdshow.... ..

[leeperry]

I was wondering, sometimes 5.1 AC3 seems to be tagged as stereo...I've got a DVD here, where KMPlayer says it's stereo, but both MediaInfo and ffdshow show it as 6 channels.

too bad my "stereo <> 6 channels" rule in ffdshow audio doesn't work and reverts to stereo...yet it shows as 6 channels input in ffdshow audio, I guess I'm SOL?

I've put a sample here: http://www.mediafire.com/file/mn23gioeoug/Sample.ac3

[albain]

@Casshern : please copy/paste the debugview logs (check "send strings to debugger" in ffdshow)

@hoborg : are you able to see the streams inside the ffdshow video context menu ?

[hoborg]

@hoborg : are you able to see the streams inside the ffdshow video context menu ?

Yes, and i can swap them by mouse without any problem
http://hobring.esero.net/saf/ffdshow/audio.jpg
http://hobring.esero.net/saf/ffdshow/sub.jpg

[ikarad]

@Casshern : please copy/paste the debugview logs (check "send strings to debugger" in ffdshow)

@hoborg : are you able to see the streams inside the ffdshow video context menu ?

Have you succeed to download part2?

[albain]

Have you succeed to download part2?

I am downloading it right now, works fine, thanks

[albain]

@hoborg : weird, could you post the content of Media Control logfile (first you have to set loglevel to debug from the configuration program)

Thx

[hoborg]

@hoborg : weird, could you post the content of Media Control logfile (first you have to set loglevel to debug from the configuration program)

Thx

sorry, i am not using WMC or Media Control...

[rsd78]

@hoborg : weird, could you post the content of Media Control logfile (first you have to set loglevel to debug from the configuration program)

Thx

Albain, could Hoborg be seeing the same issues I had using the mpc splitter and no enumeration of individual audio/sub streams?

I also could see the streams listed under the ffdshow properties, but not via 6.04 Media Control using mpc splitter and ffdshow dxva. Again, everything works fine if ffdshow regular is being used and not dxva.

EDIT: Let me know if you want logs, since I'm using Media Control and Hoborg is not. But ffdshow dxva + mpc-splitter seem to be in common between our two cases.

[albain]

Yes I am interested in such logs.

Actually I was referring to this problem but I mixed up the issues.

About Hoborg issue, I should be able to reproduce it on my own

[rsd78]

Yes I am interested in such logs.

Actually I was referring to this problem but I mixed up the issues.

About Hoborg issue, I should be able to reproduce it on my own

Ok I will get those to you tonight

EDIT: Attached is a log file. I tested a few things.
1. Used ffdshow dxva + mpc-splitter. Pressed blue button to get to menu but it failed to find individual sub streams.
2. Used ffdshow + mpc-splitter. Pressed blue button to get to menu but it failed to find individual sub streams.
3. Still using ffdshow + mpc-splitter but resumed the same file and this time it found subtitle streams automatically (i.e. the context menu popped up while previously it did not) as well as going through the menu.

All tests done using same mkv with several embedded subtitles.

[baribal]

Guys, just wanted to know if someone has kmplayer + ffdshow dxva working? I have last version of kmplayer and 3336 x86 version of ffdshow from the xvidvideo.ru site. I have also tried earlier ffdshow versions. The problem is kmplayer always using ffdshow video decoder instead of ffdshow dxva video decoder. I decrease ffdshow video decoder merit and increase ffdshow dxva video decoder merit as high as possible. I also turned on in ffdshow video decoder ffmpeg-mt for H264/AVC. Nothing helped. :( And of course i turned on H264 and VC1 in hardware acceleration DXVA options in ffdshow DXVA video decoder, EVR renderer in player, win7 x64 OS. MPC HC working like a charm with ffdshow dxva video decoder. Could someone help me to resolve the issue? Thanks in advance.

[onomatopellan]

Guys, just wanted to know if someone has kmplayer + ffdshow dxva working? I have last version of kmplayer and 3336 x86 version of ffdshow from the xvidvideo.ru site. I have also tried earlier ffdshow versions. The problem is kmplayer always using ffdshow video decoder instead of ffdshow dxva video decoder. I put ffdshow video decoder merit low and ffdshow dxva video decoder merit as high as possible. I also turned on in ffdshow video decoder ffmpeg-mt for H264/AVC. Nothing helped. :( And of course i turned on H264 and VC1 in hardware acceleration DXVA options in ffdshow DXVA video decoder. MPC HC working like a charm with ffdshow dxva video decoder. Could someone help me to resolve my issue? Thanks in advance.
It doens't work with TheKMPlayer. You can try PotPlayer (from the same author), it works but the player crash when you close it.

[Casshern]

@Casshern : please copy/paste the debugview logs (check "send strings to debugger" in ffdshow)

@hoborg : are you able to see the streams inside the ffdshow video context menu ?
Here is the debug run. The v3336 just does not pass the dts core to sp/dif. It is as if is just does not recognize that there is a dts core. It works with older versions - and with the mpc-hc inbuilt dts decoder. The strange thing is that other titles work!


[1484] Loading modified build of detours library designed for MPC-HC player (http://sourceforge.net/projects/mpc-hc/)
[1484] TffdshowBase::Constructor
[1484] TffdshowDecAudio::Constructor
[1484] Join filter graph
[1484] Removed from filter graph
[1484] Join filter graph
[1484] TffdshowDecAudio::CheckConnect (input)
[1484] TffdshowDecAudio::CheckInputType
[1484] TffdshowDecAudio::getCodecId
[1484] TffdshowDecAudio::getCodecId: codecId=dts (107)
[1484] TinputPin::SetMediaType
[1484] TffdshowDecAudioInputPin::initAudio
[1484] TffdshowDecAudio::getCodecId
[1484] TffdshowDecAudio::getCodecId: codecId=dts (107)
[1484] initPreset
[1484] TffdshowDecAudio::getOutsf PCM 1
[1484] TffdshowBase::Constructor
[1484] TffdshowDecAudio::Constructor
[1484] Join filter graph
[1484] Removed from filter graph
[1484] Join filter graph
[1484] Removed from filter graph
[1484] TffdshowDecAudio::getOutsf PCM 1
[1484] TffdshowDecAudio::CheckConnect (output)
[1484] initPreset
[1484] TffdshowDecAudio::getOutsf PCM 1
[1484] TffdshowDecAudio::CheckTransform From :
[1484] Media Type Structure
[1484] Format type : FORMAT_WaveFormatEx
[1484] Sub type : MEDIASUBTYPE_DTS
[1484]
[1484] WAVEFORMATEX :
[1484] wFormatTag : DTS wave
[1484] Channels : 6
[1484] Bits per sample : 0
[1484] Samples per second : 48000
[1484] nBlockAlign : 2012
[1484] nAvgBytesPerSec : 188719
[1484]
[1484] TffdshowDecAudio::CheckTransform To :
[1484] Media Type Structure
[1484] Format type : FORMAT_WaveFormatEx
[1484] Sub type : MEDIASUBTYPE_PCM
[1484]
[1484] WAVEFORMATEXTENSIBLE :
[1484] subFormat : MEDIASUBTYPE_PCM
[1484]
[1484] wSamplesPerBlock : 16
[1484] Valid bits per sample : 16
[1484] dwChannelMask : 63
[1484]
[1484] WAVEFORMATEX :
[1484] wFormatTag : WAVE_FORMAT_EXTENSIBLE
[1484] Channels : 6
[1484] Bits per sample : 16
[1484] Samples per second : 48000
[1484] nBlockAlign : 12
[1484] nAvgBytesPerSec : 576000
[1484]
[1484] TffdshowDecAudio::CheckInputType
[1484] TffdshowDecAudio::getCodecId
[1484] TffdshowDecAudio::getCodecId: codecId=dts (107)
[1484] TffdshowDecAudio::getOutsf PCM 1
[1484] TffdshowDecAudio::CheckTransform To generated by FFDShow :
[1484] Media Type Structure
[1484] Format type : FORMAT_WaveFormatEx
[1484] Sub type : MEDIASUBTYPE_PCM
[1484]
[1484] WAVEFORMATEXTENSIBLE :
[1484] subFormat : MEDIASUBTYPE_PCM
[1484]
[1484] wSamplesPerBlock : 16
[1484] Valid bits per sample : 16
[1484] dwChannelMask : 63
[1484]
[1484] WAVEFORMATEX :
[1484] wFormatTag : WAVE_FORMAT_EXTENSIBLE
[1484] Channels : 6
[1484] Bits per sample : 16
[1484] Samples per second : 48000
[1484] nBlockAlign : 12
[1484] nAvgBytesPerSec : 576000
[1484]
[1484] TffdshowDecAudio::CheckTransform result 0
[1984] TffdshowBase::Constructor
[1984] TffdshowDecVideo::Constructor
[1984] TffdshowDecVideoOutputPin::Constructor
[1984] TffdshowDecVideoDXVA constructor
[1984] TffdshowDecVideo::Destructor
[1984] TffdshowDecVideoOutputPin::Destructor
[1984] TffdshowBase::Constructor
[1984] TffdshowDecVideo::Constructor
[1984] TffdshowDecVideoOutputPin::Constructor
[1984] Join filter graph
[1984] TffdshowVideoInputPin::ReceiveConnection

[zummi]

Hi there,

I too are having some trouble with MediaControl and ffdshow DXVA. In short, with DXVA enabled I can't switch audio streams or subtitles by using MediaControl. MediaControl detects and displays all streams and subs, but nothing happens when i click them. I can still right-click the ffdshow tray icon and switch.

This only happens when I use DXVA. With "regular" ffdshow MediaControl works as intended.

running x64 and MPC-HC splitter (haali gives me DTS-HD bitstreaming issues), MediaControl 6.0.4 x64 and the latest x64 ffdshow from xvidvideo.ru (build 3336). Any ideas...?

[LigH]

ffdshow_rev3337_20100329_sse_icl11 from xvidvideo.ru flickers while decoding AVC. It looks like the deblocking at edges does not work correctly and differently for most frames. Happens with both libavcodec and ffmpeg-mt decoder.

ffdshow_rev3304_20100304_sse_icl11 does not show this effect.

Look at the following images (they are carefully web-optimized JPEGs, compression doesn't harm the effect too much), especially compare the borders around the cyan color bar, the near-bottom left blue field and the bottom-center dark blue field.

http://www.ligh.de/pics/ffdshow/50p_Frame0000.jpg
http://www.ligh.de/pics/ffdshow/50p_Frame0001.jpg
http://www.ligh.de/pics/ffdshow/50p_Frame0012.jpg
http://www.ligh.de/pics/ffdshow/50p_Frame0025.jpg
http://www.ligh.de/pics/ffdshow/50p_Frame0026.jpg

The MP4 file contains all relevant details about the x264 encoding.

http://www.ligh.de/pics/ffdshow/50p.mp4 (160 KB).

[rsd78]

Hi there,

I too are having some trouble with MediaControl and ffdshow DXVA. In short, with DXVA enabled I can't switch audio streams or subtitles by using MediaControl. MediaControl detects and displays all streams and subs, but nothing happens when i click them. I can still right-click the ffdshow tray icon and switch.

This only happens when I use DXVA. With "regular" ffdshow MediaControl works as intended.

running x64 and MPC-HC splitter (haali gives me DTS-HD bitstreaming issues), MediaControl 6.0.4 x64 and the latest x64 ffdshow from xvidvideo.ru (build 3336). Any ideas...?

Check the Keys/Input section in ffdshow dxva and make sure it matches regular ffdshow. I don't think Media Control's configuration command affects dxva yet. Basically this was my issue, because the remote control settings weren't configured properly.

[zummi]

Oh crap...i'm so stupuid i shouldn't be allowed to live. Well, thanks a lot, mate :-)

Ah, just one question...is there a way to set prefered stream/sub languages either in MediaControl or the MPC-HC splitter - like you can in Haali?

[albain]

Yes, I have to fix this

@Ikarad : here is a test build with better management of bluray subtitltes
Thanks to supread author I was able to do a better implementation.
Issue #3 may not be fixed yet (I haven't tested it)

http://damienbt.free.fr/ffdshow_rev3336_20100327_dbt_newlibs_beta9.exe

[ikarad]

Yes, I have to fix this

@Ikarad : here is a test build with better management of bluray subtitltes
Thanks to supread author I was able to do a better implementation.
Issue #3 may not be fixed yet (I haven't tested it)

http://damienbt.free.fr/ffdshow_rev3336_20100327_dbt_newlibs_beta9.exe

I try and it crash many times with rgb32 output.
When it works subs are not displayed. In the first 35 secondes the only sub displayed is "abandon ship" at 29 second.

I can't test if the problem that I speak before are corrected
With 3326 and rgb32 output I find several problems with blu rays subtitles (I can't test with YUY2 because It crash).
1)first problem: some subs are cut (Same problem with 3306 and other ffdshow version untill 3326)
For example there is this sub at 8min33
"Your back is heavy as I catch up to you"

http://nsa14.casimages.com/img/2010/03/21/mini_100321091909495937.png (http://www.casimages.com/img.php?i=100321091909495937.png)
In the video the "you" is cut
http://nsa15.casimages.com/img/2010/03/21/mini_100321092046662446.jpg (http://www.casimages.com/img.php?i=100321092046662446.jpg)

Same problem with this sub at 8min46
in the video
The original sub is

http://nsa14.casimages.com/img/2010/03/21/mini_100321092808826318.png (http://www.casimages.com/img.php?i=100321092808826318.png)

2) second problemThere is some problems with the time of displaying of subtitles.
For example in this video (that I also give you) The first sub :


is well displayed.
Although the second sub:
http://nsa14.casimages.com/img/2010/03/21/mini_100321090446191729.png (http://www.casimages.com/img.php?i=100321090446191729.png)

is displayed and is removed immediately. There is no time to read it and it's not normal.

Same problem with this sub

In the movie it disappeared at 00:08:41 and not at 00:08:44,272

There many other examples in the sample that I give you.

3) Third problem: If I use resize function (lanczos) of ffdshow: the display of subs is very very bad. All subs are displayed in the high left corner instead of the middle of screen and many subs are cut. Sound is cut at some moments.

[DMD]

ffdshow is coupled with other applications except PowerDVD, then
launching PowerDVD because I see this message?

http://www.allfreeportal.com/imghost2/images/159821Sccreenshot 001.png (http://allfreeportal.com)

http://www.allfreeportal.com/imghost2/images/102569Sccreenshot 002.png (http://allfreeportal.com)


THANKS

[clsid]

I shall add powerdvd10.exe to the default blacklist.

Choose "Don't use ffdshow (always)".

[STaRGaZeR]

Hi, Can you provide a sample ?
Thanks,
Seb.

Sure: http://www.mediafire.com/?jmyyddtm0mi (153MB)

It'll start with the AC3 track if you use Haali. With r3334 and previous you can change to the THD track and hear it perfectly fine, but if you change back to AC3 you will get silence. After r3334 if you change to the THD track you'll get silence, but if you change back to the AC3 track you'll hear it. Tested with r3338.

[Sebastiii]

Thanks, i'll try this evening :)
Did you test only with haali ?
Cheers,
Seb.

[DMD]

I shall add powerdvd10.exe to the default blacklist.

Choose "Don't use ffdshow (always)".
Thanks!

I activated the list does not use

http://www.allfreeportal.com/imghost2/images/790886Sccreenshot 003.png (http://allfreeportal.com)

[albain]

@Ikarad
@Stargazer

I have finally fixed this blackscreen problem with new ffmpeg libraries. The CPU optimizations weren't correctly taken into account and could cause weird issues (such as this one)

This build (http://www.mediafire.com/file/juf01drmly4/ffdshow_rev3342_20100331_dbt_newlibs_beta10.exe) fixes it and fixes also the stability problems with bluray subs

[Casshern]

@Ikarad
@Stargazer

I have finally fixed this blackscreen problem with new ffmpeg libraries. The CPU optimizations weren't correctly taken into account and could cause weird issues (such as this one)

This build (http://www.mediafire.com/file/juf01drmly4/ffdshow_rev3342_20100331_dbt_newlibs_beta10.exe) fixes it and fixes also the stability problems with bluray subs

Any news on the recently introduced bug that ffdshow only picks up the DTS core when files are played from the begining? If you directly seek and start playback in the middle ffdshow only decodes the DTS HD stream.

[LigH]

@ albain:

Does this also affect the issues mentioned in #11222 (http://forum.doom9.org/showpost.php?p=1387394&postcount=11222)?

[clsid]

@LigH
I don't see those artifacts with the generic build. Could you test that?

[Sebastiii]

@Ikarad
@Stargazer

I have finally fixed this blackscreen problem with new ffmpeg libraries. The CPU optimizations weren't correctly taken into account and could cause weird issues (such as this one)

This build (http://www.mediafire.com/file/juf01drmly4/ffdshow_rev3342_20100331_dbt_newlibs_beta10.exe) fixes it and fixes also the stability problems with bluray subs

Hi,

The file is corrupt on my side ? i have down twice.
And you ?
Thx :)
Seb.

[fastplayer]

Hi,

The file is corrupt on my side ? i have down twice.
And you ?
Thx :)
Seb.
Alternative download link (http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1)

[ikarad]

@Ikarad
@Stargazer

I have finally fixed this blackscreen problem with new ffmpeg libraries. The CPU optimizations weren't correctly taken into account and could cause weird issues (such as this one)

This build (http://www.mediafire.com/file/juf01drmly4/ffdshow_rev3342_20100331_dbt_newlibs_beta10.exe) fixes it and fixes also the stability problems with bluray subs

I have no crash now with rgb32 output but with YUY2 there is crash like with other versions

But I have the same bug with 3336: many subs are not displayed. For example in the first 35 secondes the only sub displayed is "abandon ship" at 29 second. With 3326 there wasn't this problem.

[Atak_Snajpera]

I think it is time to get rid of some useless algos under Deinterlace section

http://img714.imageshack.us/img714/4338/new1f.png

Screenshots -> http://www.mediafire.com/file/zhm3itwym2n/deinterlaced_pictures.7z

You will see that most of them provide poor quality. The best in my opinion is TomsMoComp + Vertical . However it does not deinterlace to full frame rate. Second would be Yadif.

[albain]

I have no crash now with rgb32 output but with YUY2 there is crash like with other versions

But I have the same bug with 3336: many subs are not displayed. For example in the first 35 secondes the only sub displayed is "abandon ship" at 29 second. With 3326 there wasn't this problem.

Ok for yuy2 but I think that the behaviour is correct in rev 3336

I know that rev 3326 displayed the first subs but it did not respect the PGS format

In the new revision I have to wait for the next subtitle to know if the current one is really finished to parse which explains why you can't see it during the first seconds of the video. But after the first subs it should be fine

Also note that the MPC splitter is buggy and does not detect the PGS streams early enough. But once the stream is detected and you have selected it, you can go back to the beginning of the video and the first subs should be displayed

Maybe this new behaviour could be improved but I would like to confirm first that the implementation is correct (YUY2 colorspace apart)

Any news on the recently introduced bug that ffdshow only picks up the DTS core when files are played from the begining? If you directly seek and start playback in the middle ffdshow only decodes the DTS HD stream.

I'll take a look on it tomorrow. Don't hesitate to remind me it if I forget

[STaRGaZeR]

Thanks, i'll try this evening :)
Did you test only with haali ?
Cheers,
Seb.

Gabest MKV splitter has always worked fine, but not really an option for me because it doesn't have an internal stream switcher and I need the splitter to honor the default track flags present in my MKVs :p. Keep in mind that with Gabest's all the tracks are sent at the same time to ffdshow and then you choose which one you want with the stream switcher, so it doesn't have to check for changes in a single stream, that's why it works while Haali doesn't.

@Ikarad
@Stargazer

I have finally fixed this blackscreen problem with new ffmpeg libraries. The CPU optimizations weren't correctly taken into account and could cause weird issues (such as this one)

This build (http://www.mediafire.com/file/juf01drmly4/ffdshow_rev3342_20100331_dbt_newlibs_beta10.exe) fixes it and fixes also the stability problems with bluray subs

As Sebastii says the file seems to be corrupted. Can't wait to test it.

Alternative download link (http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1)

Those are SVN builds only.

[clsid]

Please also test the latest version of Haali splitter if you haven't already. It supposedly has less problems with FF and RW.

[ikarad]

In the new revision I have to wait for the next subtitle to know if the current one is really finished to parse which explains why you can't see it during the first seconds of the video. But after the first subs it should be fine

Also note that the MPC splitter is buggy and does not detect the PGS streams early enough. But once the stream is detected and you have selected it, you can go back to the beginning of the video and the first subs should be displayed




I tried and it doens't work. With 3326 it works but not with 3336 and 3342.

Even if I go back to the beginning of the video, subs are not displayed except the first sub 1
00:00:09,050 --> 00:00:12,720
'"2015 A.D. Perseus Arm, Milky Way Galaxy
(12,000 arsecs from Earth)'"

and
7
00:00:28,570 --> 00:00:30,237
Abandon ship!

for the first 35 seconds. All the others subs are not displayed.
I try untill 2 minutes and the only subs displayed are the first and abandon ship. The others are not displayed

I try at 8 minutes and there is the same problem. Some subs are displayed and some are not displayed.
With 3326 there wasn't this problem.

I try with mpc hc 1613 or 1771 and there is the same problem. I try with vmr9renderless or evrcustom and there is the same problem.

I can't make a video to show this problem because fraps doesn't work with mpc-hc since many months

[albain]

Ok I have not paid attention enough.

I'll work on it tomorrow

[albain]

Here is a new build (http://damienbt.free.fr/ffdshow_rev3343_20100401_dbt_newlibs_beta11.exe)

Fixes :

Bluray subtitles fixes : now all are displayed
Stability fixes on bluray subtitles
Crashes fixed when using YUY2 colorspace with subtitles (this is a workaround inside ffmpeg but Tspuimage should not rescale picture plane by plane but all the planes at the same time)
New ffmpeg libraries inside (libswscale and libpostproc)

[clsid]

Albain, while you're working on the subtitle part, I have some suggestions to make the options a bit more user-friendly:

1) Move positioning/alignment/letterbox/stereoscopic options to a new subpage called "Positioning".
2) Move all checkboxes for enabling formats to the top of the main subtitles page. Move all remaining options on that page down a bit to make room at the top.
So there will be three checkboxes for enabling the various subtitle formats: Text-based, Closed Captions, PGS & VobSub. Plus "SSA/ASS" which is a child of Text-based. Plus "accept embedded" which applies to all.
Afaik it is currently not yet possible to disable handling of Text-based subs, while keeping for example PGS enabled. But I think that could be useful.
Perhaps PGS and VobSub could also be splitted into two format checkboxes?

[ikarad]

Here is a new build (http://damienbt.free.fr/ffdshow_rev3343_20100401_dbt_newlibs_beta11.exe)

Fixes :

Bluray subtitles fixes : now all are displayed
Stability fixes on bluray subtitles
Crashes fixed when using YUY2 colorspace with subtitles (this is a workaround inside ffmpeg but Tspuimage should not rescale picture plane by plane but all the planes at the same time)
New ffmpeg libraries inside (libswscale and libpostproc)


I try With YUY2 there is no crash thanks.
There is an improvement for displaying of subs but the same problem still exists.
There is more subs displayed with 3343 but not all subs are displayed like with 3326.
For example during the first 35 seconds these subs are displayed
1
00:00:09,050 --> 00:00:12,720
'"2015 A.D. Perseus Arm, Milky Way Galaxy
(12,000 arsecs from Earth)'"
7
00:00:28,570 --> 00:00:30,237
Abandon ship!

9
00:00:31,406 --> 00:00:32,948
Repeat, abandon ship!


theses subs are not displayed

3
00:00:14,597 --> 00:00:15,723
ETA, 20 seconds!

4
00:00:15,890 --> 00:00:17,975
Full speed evasive maneuvers, on the douvle!

5
00:00:18,059 --> 00:00:19,476
Full speed evasive maneuvers!
8
00:00:30,280 --> 00:00:31,321
We are avandoning ship.
10
00:00:33,241 --> 00:00:34,575
One more!

There is an mprovement because with the other version between 3336 and 3343 only two subs were displayed

1
00:00:09,050 --> 00:00:12,720
'"2015 A.D. Perseus Arm, Milky Way Galaxy
(12,000 arsecs from Earth)'"
7
00:00:28,570 --> 00:00:30,237
Abandon ship!

[STaRGaZeR]

Here is a new build (http://damienbt.free.fr/ffdshow_rev3343_20100401_dbt_newlibs_beta11.exe)

Fixes :

Bluray subtitles fixes : now all are displayed
Stability fixes on bluray subtitles
Crashes fixed when using YUY2 colorspace with subtitles (this is a workaround inside ffmpeg but Tspuimage should not rescale picture plane by plane but all the planes at the same time)
New ffmpeg libraries inside (libswscale and libpostproc)


I'm happy to announce that all these issues have been fixed:

RGB32:

- Luminance fix in picture properties, black screen. But wait, if you enable deband and you place it after picture properties everything is fine! The filter works and the image is not black. Maybe this can help you with debugging.
- Deinterlacing, same as luminance fix, it works perfectly fine but only if you put it before deband and you activate it, otherwise black screen with ffmpeg deints.
- Postprocessing with mplayer checked, I get black screen. Again if I activate deband after PP it works perfectly fine.
- mplayer temporal noise reducer in Blur, same as above, with deband after it it works fine.

YUY2:

With SD content, exact same issues as RGB32, but changing the black screens with the double speed but same number of frames issue I reported earlier. With HD content, exact same issues as RGB32 but changing the black screens with stuttering. Activating deband after the filters fixes the double speed and the stuttering.


Bugs that are still in there:

- NV12 output: black screen.
- White shadow in DVD subs.

Great work!

[FredThompson]

EDIT2: Another note/bug with MC. From your front page the Download button goes to an error page (not found). I had to download it manually through Sourceforge. Also on Sourgeforge the current version listed is 6.02, I had to go to show all files to get to 6.04.That's common with SourceForge. It's automatically generated but periodically, not interrupt-driven.

[onomatopellan]

With latest build the line of pixels problem (http://i45.tinypic.com/2iizvkk.png) is almost fixed. Now sometimes appear a vertical line at the end of the subs.
http://i42.tinypic.com/5lbdow.png

It seems pretty stable with RGB32 output though, great work!

[zummi]

Hi, quick question. Is there a way to give DXVA higher priority than non-DXVA? I've tried moving the merit-bars around, but regular ffdshow always "wins" unless i completely disable it - which then again gives me trouble with non-dxva content. I'm using regular WMP through W7 MediaCenter... Thanks :-)

[SamuriHL]

See? I'm not the only one with merit problems with dxva vs non! :)

[rsd78]

See? I'm not the only one with merit problems with dxva vs non! :)

Using Shark's codec pack I never ran into this issue on W7 WMC. I know people tend to frown on packs (including myself in the past), but for W7 it works well at least for me.

[SamuriHL]

Using Shark's codec pack I never ran into this issue on W7 WMC. I know people tend to frown on packs (including myself in the past), but for W7 it works well at least for me.

I hate codec packs. I'm not even going to entertain that thought. It's as simple as that.

[rsd78]

I hate codec packs. I'm not even going to entertain that thought. It's as simple as that.

I never used them for Vista and earlier (I've had a dedicated "HTPC" for close to 7 years now, and my background is in Software Engineering so I'm not an idiot when it comes to computers/software), but I have had very good experiences with this one because it let's me make major changes in settings with a few clicks and I can always uninstall it w/o it leaving any permanent marks on the system. At the end of the day it worked, and saved me a little less effort in terms of getting my htpcs up and running. But I totally respect your stance as well :)

[SamuriHL]

I totally respect that. I never used them for Vista and earlier, but I have had very good experiences with this one because it let's me make major changes in settings with a few clicks and I can always uninstall it w/o it leaving any permanent marks on the system.

That's the sign of a good codec pack, but, I find that it's always better to install EXACTLY what you need rather than a blanket codec pack. (And yes, I do realize that the good codec packs allow you to control what gets installed, but, I'd rather do it myself as needed). Just a philosophical thing. For me, I forced the merit on ffdshow dxva in MPC-HC and called it good. But, I was also told this wasn't necessary because it should be higher than the non-dxva version. I suspected that was wrong but I was the only one reporting it so I left it at that. :)

[clsid]

The default merit for ffdshow is FF800001 and for DXVA FF800002.

Be aware that for WMP/MC on Windows 7 you need to use tweaks to change the preferred decoder in order to override the Microsoft decoders.

[SamuriHL]

I haven't tried to use it outside of MPC-HC. For whatever reason it decided to choose ffdshow instead of DXVA until I forced it by adding it to external filters and setting it to prefer. Then it was all good. As long as no one else is running into it I'm good with that. You're right about having to tweak it on W7. MS decoders just don't like going away on their own. :D

[zummi]

The default merit for ffdshow is FF800001 and for DXVA FF800002.

Be aware that for WMP/MC on Windows 7 you need to use tweaks to change the preferred decoder in order to override the Microsoft decoders.

I've been running the Windows 7 DS filter tweaker tool, but it only lets me choose ffdshow, not dxva specifically...

[DigitalDeviant]

Here is a new build (http://damienbt.free.fr/ffdshow_rev3343_20100401_dbt_newlibs_beta11.exe)

Fixes :

Bluray subtitles fixes : now all are displayed
Stability fixes on bluray subtitles
Crashes fixed when using YUY2 colorspace with subtitles (this is a workaround inside ffmpeg but Tspuimage should not rescale picture plane by plane but all the planes at the same time)
New ffmpeg libraries inside (libswscale and libpostproc)


Still no luck with this (http://www.mediafire.com/?anywvxwl3m4) sample. Now instead of displaying both subtitles at the top of the screen (the 2nd is supposed to be at the bottom) it just shows the 1st and does not display the 2nd at all. Previous versions would have ended the 1st subtitle and displayed the 2nd at the top in it's place.

[zummi]

Using Shark's codec pack I never ran into this issue on W7 WMC. I know people tend to frown on packs (including myself in the past), but for W7 it works well at least for me.

if a codec pack can do it, I'm guessing it's also possible to do it manually somehow. Just gotta figure out how. Glad I'm not the only one with this issue though :)

[rica]

I've been running the Windows 7 DS filter tweaker tool, but it only lets me choose ffdshow, not dxva specifically...

Try this tool:

http://alax.info/blog/1111

And MPC-HC,

so as to give an idea: (have a look at internal filters setup)

http://forum.doom9.org/showthread.php?t=152520

[SamuriHL]

Try this tool:

http://alax.info/blog/1111

And MPC-HC,

so as to give an idea: (have a look at internal filters setup)

http://forum.doom9.org/showthread.php?t=152520

I had disabled all the internal filters. Didn't matter.

[clsid]

I've been running the Windows 7 DS filter tweaker tool, but it only lets me choose ffdshow, not dxva specifically...
Added in version 3.1 ;)

[zummi]

Superb...just superb :) will non-dxva ffdshow be selected if the video file isn't dxva-compatible?

[clsid]

I haven't tested it, but I assume it will then simply pick whichever other filter with the highest merit, which should be normal ffdshow.

[albain]

I have brought new fixes to bluray subtitles management, but this is not perfect : simultaneous subtitles are not handled correctly (previous one stops when new one starts whereas previous one should continue)

I'll work on new ffmpeg libraries, I hope that I'll be able to commit them soon

[Sebastiii]

I have brought new fixes to bluray subtitles management, but this is not perfect : simultaneous subtitles are not handled correctly (previous one stops when new one starts whereas previous one should continue)

I'll work on new ffmpeg libraries, I hope that I'll be able to commit them soon

Thanks again for all work you made :)
Seb.

[zummi]

I haven't tested it, but I assume it will then simply pick whichever other filter with the highest merit, which should be normal ffdshow.

It works! Thanks a million :-)

[magic144]

just stumbled across this forum looking into BD subtitle support!

has anybody seen this problem report and is this a known issue at all??

http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?t=1448

cheers,
m

[EDIT]
raised this under Bugs forum:
http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?p=4994#4994
and on tracker
https://sourceforge.net/tracker/?func=detail&aid=2981790&group_id=173941&atid=867360

[Piyoko]

Hello,

I noticed that on all builds after r3274 I'm getting a 1 pixel white outline to the right on all DVD subtitles, like so:
http://img338.imageshack.us/img338/8094/subs.png (http://img338.imageshack.us/i/subs.png/)

Is this a known bug?

[STaRGaZeR]

Yup, albain already knows about it.

[zummi]

Hi guys, two questions:

1. When seeking in an mkv with TrueHD wmp crashes within 2 seconds. Haven't seen much about this problem, so it could just be me? I'm using win7x64, ffdshow x64, mpc-hc splitter and ATI 5000-series for bitstreaming. DTS HD works like a charm.

2. Where can I find a "donate"-button for this project?

[DigitalDeviant]

Still no luck with this (http://www.mediafire.com/?anywvxwl3m4) sample. Now instead of displaying both subtitles at the top of the screen (the 2nd is supposed to be at the bottom) it just shows the 1st and does not display the 2nd at all. Previous versions would have ended the 1st subtitle and displayed the 2nd at the top in it's place.

With 3345 both lines show and in the correct position. I think the first subtitle ends too early though.

[MatLz]

Hi!
Maybe found a bug
Found on 3326, updated to 3345 but still bug:

For testing I outputed a YV16 avi with Vdub.

If I use directshowsource on this avi in an avs script, it (the report points ffdshow.ax) crashes.
I noticed that happens ONLY if, in the image properties processing, contrast OR brightness are used (default values don't crash)

But strangely, all is fine on playback with MPC...:confused:

[Nyquist]

After upgrading ffdshow (now rev 3345), videos encoded with AVC/H.264 are looking very blurry when post processing is enabled. On earlier revisions, post processing didn't do anything on H.264 encoded videos.

Maybe a bug? Or a new feature?

[fastplayer]

Maybe a bug? Or a new feature?
A new "feature" since rev 3335.

[_xxl]

Deblocking for H.264?
After upgrading ffdshow (now rev 3345), videos encoded with AVC/H.264 are looking very blurry when post processing is enabled. On earlier revisions, post processing didn't do anything on H.264 encoded videos.
Maybe a bug? Or a new feature?
Yes it should be removed from pp list. Why do need h.264 video's pp?

[Nyquist]

Yes it should be removed from pp list. Why do need h.264 video's pp?

I don't know.

The deblocking routines are working quite aggressively for h.264 videos. Much more than they should in my opinion.

[_xxl]

H.264 pp from original ffdshow:
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout?view=rev&revision=99

[sacharja]

Hi guys, two questions:

1. When seeking in an mkv with TrueHD wmp crashes within 2 seconds. Haven't seen much about this problem, so it could just be me? I'm using win7x64, ffdshow x64, mpc-hc splitter and ATI 5000-series for bitstreaming. DTS HD works like a charm.

2. Where can I find a "donate"-button for this project?
I already posted this crash here. Happens for me for every h.264 file. Albein said he's going to fix it, but it seems he didn't have enough time.

[Nyquist]

H.264 pp from original ffdshow:
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout?view=rev&revision=99

Hm, looks like there's a separate function for deblocking h.264? Normally I'm using SPP for DIVX/MPEG2 with different strenghts. But on a h.264 video even weak deblocking is blurrying the whole thing.

[Astrophizz]

H.264 has its own in-loop deblocking. Any further deblocking would blur the image as you have seen.

[Mr VacBob]

There is no reason to use spp or pp as a postprocessor for FF_QSCALE_TYPE_H264. Something like gradfun is more appropriate.

[Nyquist]

H.264 has its own in-loop deblocking. Any further deblocking would blur the image as you have seen.

I know. I'm only wondering, why the builtin post processing do so much blur on h-264 videos. On good videos I cannot see an effect of applying SPP deblocking to MPEG2 or MPEG4-ASP videos, no blurrying. I'd expect this behaviour for h.264, too.

[Mr VacBob]

H.264 has more than 10x larger qp values, and it's ignoring that the values mean something different for it, so it applies more than 10x too much postprocessing.

[Nyquist]

H.264 has more than 10x larger qp values, and it's ignoring that the values mean something different for it, so it applies more than 10x too much postprocessing.

Sounds like a reasonable explaination why. Thanks!

[hoborg]

Hi.

I just tested ffdshow_rev3345_20100402.exe build, CTRL+ALT+F4 or CTRL+ALT+numpad1 still doesnot working for me.
Only me ?

[LigH]

@LigH
I don't see those artifacts with the generic build. Could you test that?

Sorry for the late reply ...


I do not see this issue with the following builds:

- ffdshow_rev3304_20100304_sse_icl11 (ICL11, x86; xvidvideo.ru)

- ffdshow_rev3305_20100304_clsid_icl10 (ICL10, x86; SourceForge)

- ffdshow_rev3330_20100323 (generic, x86; xvidvideo.ru)
- ffdshow_rev3330_20100323_sse_icl11 (ICL11, x86; xvidvideo.ru)

- ffdshow_rev3334_20100325 (generic, x86; xvidvideo.ru)
- ffdshow_rev3334_20100325_sse_icl11 (ICL11, x86; xvidvideo.ru)


I see this issue with the following builds:

- ffdshow_rev3336_20100327 (generic, x86; xvidvideo.ru)
- ffdshow_rev3336_20100327_sse_icl11 (ICL11, x86; xvidvideo.ru)

- ffdshow_rev3337_20100329 (generic, x86; xvidvideo.ru)
- ffdshow_rev3337_20100329_sse_icl11 (ICL11, x86; xvidvideo.ru)

- ffdshow_rev3342_20100331_clsid (generic, x86; SourceForge)

- ffdshow_rev3345_20100402 (generic, x86; xvidvideo.ru)
- ffdshow_rev3345_20100402_sse_icl11 (ICL11, x86; xvidvideo.ru)


My CPU:
QuadCore AMD Phenom II X4 945 (Deneb), 3000 MHz (15 x 200), RB-C2;
x86, x86-64, MMX, 3DNow!, SSE, SSE2, SSE3, SSE4A


Probably related to rev. 3335: "added H264 format to postprocessing filter" -- maybe I should have changed some option there?

[clsid]

I tested with PP disabled, so that would likely be the cause.
That change has been reverted in 3347, so your problem is probably fixed now.

[Killerattacks]

I have a DTS passthrough issue with ffdshow. I'm currently using ffdshow_rev3342_20100331_clsid and a Radeon HD4xxx's HDMI output (supports SPDIF, but no 7.1/HD-audio).

I previously used the ffdshow version of the CCCP which is totally outdated but there I had no problems. For the ffdshow settings see the attached screenshot.

File 1:
ID : 2
Format : DTS
Format/Info : Digital Theater Systems
Format-Profil : ES
Codec-ID : A_DTS
Dauer : 1h 40min
Bitraten-Modus : konstant
Bitrate : 755 Kbps
Kanäle : 6 Kanäle
Kanal-Positionen : Front: L C R, Surround: L R, LFE
Samplingrate : 48,0 KHz
Auflösung : 24 bits
Stream-Größe : 544 MiB (27%)
Titel : 5.1 DTS
This file works fine with the old and the most recent ffdshow.

Audio
ID : 3
Format : DTS
Format/Info : Digital Theater Systems
Format-Profil : MA
Codec-ID : A_DTS
Dauer : 45min
Bitraten-Modus : variabel
Kanäle : 6 Kanäle
Kanal-Positionen : Front: L C R, Surround: L R, LFE
Samplingrate : 48,0 KHz
Auflösung : 16 bits
This file only works with the old ffdshow version or when I use MPC-HC internal DTS decoder for SPDIF passthrough.
I know it's DTS-MA but the core should still be a DTS bitstream right?
PS: Enabling DTS HD bitstream doesn't make any difference. My AV Receiver is Onkyo TX-SR607.

[albain]

Hi,

about bluray subtitles and subtitles in general, I am still working on it.

Pel (supread creator) and I are working on making a better PGS format support

The current implementation is not correct but I think I am close to the solution

There is also the new ffmpeg libs on their way, the last issues to fix

[Shankster]

Hi,

about bluray subtitles and subtitles in general, I am still working on it.

Pel (supread creator) and I are working on making a better PGS format support

The current implementation is not correct but I think I am close to the solution

There is also the new ffmpeg libs on their way, the last issues to fix

Hi albain, does your work include work on the dvd format vobsub subs aswell?

I am mainly thinking of this bug stated on the front page:

2. In some cases VobSub subtitle files (.idx/.sub) are loaded, but no subs are displayed. VobSub support is still incomplete.

As it is now, some movies will start with subtitles working fine, but after a while they disappear.

If it helps, vsfilter does not have this issue.

[STaRGaZeR]

Albain, did you finally disable jitter correction for THD? With the lastest revisions I get huge jitter (30 segs or more) that is causing crashes, silence and audio desync with m2ts files here. No issues with matroska though, jitter is always at ~0. Can anybody confirm?

[DigitalDeviant]

Albain, did you finally disable jitter correction for THD? With the lastest revisions I get huge jitter (30 segs or more) that is causing crashes, silence and audio desync with m2ts files here. No issues with matroska though, jitter is always at ~0. Can anybody confirm?

Confirmed.

[STaRGaZeR]

http://img338.imageshack.us/img338/8094/subs.png (http://img338.imageshack.us/i/subs.png/)

It seems this was introduced in r3284. r3283 had the subs half transparent but correctly displayed, r3284 fixed the transparency issue but introduced the white stuff.

[swordsman]

Hi, I use ffdshow in delphi app, and every time you guys
change iffdecoder I had to update my prog.
But you didn't update iffdecoder.h (nor delphi interface) in time.
Please update interface units.
Thanks

[rsd78]

Ok, I'll fix those 2

About DXVA + Haali + FF/RW : I'll make some deeper tests with 720p mkv to reproduce it

The only difference with you is that I own windows 7 x64

Hi Albain,

Just wanted to give you a heads up, but I tested with the newest Haali and I still get crashing while ff/rew with any mkv while using ffdshow dxva. I know you are aware (and very busy) but I wanted to let you know. Would love this to be fixed personally, as its the only thing from preventing me from using ffdshow dxva within Windows Media Center.

Thanks

[albain]

Hi, I use ffdshow in delphi app, and every time you guys
change iffdecoder I had to update my prog.
But you didn't update iffdecoder.h (nor delphi interface) in time.
Please update interface units.
Thanks

Could you explain a little more ?

@All : I have fixed the last issues on new ffmpeg libraries (nv12 black screen). Now everything is working
I'm doing some code cleanings but the modifications are ready to be committed.

[Sebastiii]

:thanks:
Seb.

[Casshern]

DTS Spdif passthrough problems with m2ts files:

1) The DTS core does get picked up when playing the file from the beginning
2) I had MPC-HC set to remember last position, so it essentially seeked before playing -> this results in the core not being picked up
3) When seeking sound gets totally screwed, only multiple reseeking helps to get it back

Here is a small sample (which makes seeking hard, but try to play it NOT from the beginning to replicate)

http://www.mediafire.com/file/mztm2kmzu4u/testdtscore.m2ts

[ikarad]

Hi,

about bluray subtitles and subtitles in general, I am still working on it.

Pel (supread creator) and I are working on making a better PGS format support

The current implementation is not correct but I think I am close to the solution

There is also the new ffmpeg libs on their way, the last issues to fix
I try 3348 and there is no problem with blu ray subs. It is perfect with the sample that I give you. All subs are displayed with the correct timing. Thanks

But there is a new problem.
With 3343 you have corrected crash with YUY 2. With 3348 there are crashes with YUY2. It works only with rgb32

[albain]

I try 3348 and there is no problem with blu ray subs. It is perfect with the sample that I give you. All subs are displayed with the correct timing. Thanks

But there is a new problem.
With 3343 you have corrected crash with YUY 2. With 3348 there are crashes with YUY2. It works only with rgb32

The crash fix on YUY2 was not committed, but it was embedded in the test build I posted.
It is located in the new ffmpeg libraries that I am about to commit

[STaRGaZeR]

http://www.xtremesystems.org/forums/images/smilies/clap.gif

[swordsman]

Could you explain a little more ?

@All : I have fixed the last issues on new ffmpeg libraries (nv12 black screen). Now everything is working
I'm doing some code cleanings but the modifications are ready to be committed.

I mean you changed intrface unit or add some entries to it, but
version in svn repository is some months old, so I can't change my delphi units to work with last version of ffdshow.

[STaRGaZeR]

No problems found so far with the new libs, great work albain!

[Sebastiii]

Hi,

I've just made some test and it works good :) (great works)

I post some screenshot and i found on some sub this :

http://sebinternet.free.fr/ffdshow_3350-1.png

http://sebinternet.free.fr/ffdshow_3350-2.png

http://sebinternet.free.fr/ffdshow_3350-3.png

Also with Graphstudio and WMP, if i play .mkv, i can see sub (embedded or external) but if i open ffdshow windows to setup size of sub or other, the sub don't appear. (only when windows is open)

With .m2ts it's ok, i don't know why lol.
Seb and again great work :)

:thanks:

[ikarad]

The crash fix on YUY2 was not committed, but it was embedded in the test build I posted.
It is located in the new ffmpeg libraries that I am about to commit
I try rev3350 and it crash with rgb32 output and bluray subs. With 3348 no crash.

[albain]

I try rev3350 and it crash with rgb32 output and bluray subs. With 3348 no crash.

I don't reproduce the crash. Where did you get the rev3350 ?

I'll bet on a bad compilation, revision 3350 includes the new ffmpeg libraries and a clean compilation has to be made

Try this build (32 bits) (http://damienbt.free.fr/ffdshow_rev3350_20100407_dbt_.exe)

[onomatopellan]

rgb32 seems perfect, great work albain!
But yuy2 an dxva output still shows strange colors.
http://i41.tinypic.com/2cf9icl.png

[Thunderbolt8]

I have a problem with mpc-hc and ac3filter, but don't know whether its one or the other causing it (or maybe both). I used to have the problem for 1-2 years that the preferred ac3filter was just 'dropped' when I switched from one audio track inside a .mkv file to another. so basically when I had two 5.1 tracks or at least one 5.1 which was not the default one, then ac3filter was dropped after switching the track which was really bad for me, because I use ac3filter to downmix from 5.1 to 2.0 channels (because I only have 2 speakers).

since recently this problem seems to be solved though for 5.1 tracks, I can now switch between two 5.1 tracks inside a .mkv file and it's all fine, ac3filter still remains active. however, now this problem emerges when switching to 2.0 audio tracks, theres no sound at all anymore then. theoretically it's not that bad here, because I don't need ac3filter to downmix the channels. but I still use it change the gain & master volumes etc, so sometimes I still need it even for 2.0 channels. I use madlfac, ac3filter and reclock all together for video & audio decoding of movies (or at least ac3filter and reclock if no flac is involved).

the problem seems to be reclock (and/or madflac?) here, when I disable reclock, switching to the 2.0 is OK and the flac track plays with madflac only, but ac3filter becomes dropped if active. but as soon as reclock is active, it won't play any more, no matter if I use reclock + madflac or reclock + madflac + ac3filter. madflac (and ac3filter, if active as well) get dropped as active filters when switching to the 2.0 track in this situation, if reclock is active as well. they also remain dropped when switching back to the 5.1 track again so that you have to restart the player in order to get them back (this problem was the same for the 5.1 tracks back then, but only seems to apply now when switching back from the 2.0 to the 5.1 again; switching between two 5.1 tracks is fine, nothing gets dropped).

so when madflac and ac3filter are active:
from 5.1 --> 2.0 = ac3filter gets dropped. when switching back to 5.1 it remains dropped and 5.1 are not put out as 2.0 any more

when madlfac and ac3filter and reclock are active:
from 5.1 --> 2.0 = madflac & ac3filter get dropped, no sound output at all. same when switching back to 5.1 again


so which prog is to blame here, mpc-hc, ac3filter, madflac or reclock?

EDIT: actually switching to 5.1 is only fine when switching from another 5.1 track to 5.1 again. when having a 2.0 track as default, then switching to 5.1 results in dropping of ac3filter and madflac again, only reclock remains active. same when switching back to 2.0 again. seems to be some 5.1 and 2.0 track interference of filters or something like that. so I was wrong, nothing has been solved recently, unfortunately.

[Casshern]

I don't reproduce the crash. Where did you get the rev3350 ?

I'll bet on a bad compilation, revision 3350 includes the new ffmpeg libraries and a clean compilation has to be made

Try this build (32 bits) (http://damienbt.free.fr/ffdshow_rev3350_20100407_dbt_.exe)

With this build the dts core of dts MA tracks still does only get played over sp/dif if the file is played from the beginning (like it should if only DTS and not DTS HD is checked under output). If played from the last position (essentially seeking before starting playback) ffdshow incorrectly decodes the dts MA track and outputs according to the mixer options, instead of extracting the dts core.

Interestingly if played from beginning, further seeking is no problem ffdshow correctly picks up the dts core and outputs it over sp-dif. If playback starts in the middle of the m2ts, the dts core is not picked up (DTSMA gets decoded and output - in my case - as stereo) and further seeking does not restore the correct dts core sp/dif playback - even if seeking to the beginning of the file. Somehow the seeking code seems different from initial playback code. A couple of versions back ffdshow worked fine.

This is the behaviour with the mpc m2ts splitter. Unfortunately haali is even worse in that it plays the affected files only in stutter motion.

[albain]

rgb32 seems perfect, great work albain!
But yuy2 an dxva output still shows strange colors.
http://i41.tinypic.com/2cf9icl.png

I spent a lot of time on this, I don't understand why the subtitles are half transparent

The blending formula that I use with SSE instructions works perfectly for DVD subs
Maybe the problem comes from the ffmpeg scale routines
This problem was already there before.
I'll need some help from other developers on this

About DTS core decoding : I didn't work on this yet

@Sebastii : these borders are not normal. Do you have this problem in both RGB/YUV colorspace ? Could you post a sample ? Thanks

@Stargazer : do you still have border problems with DVD subtitles ? Could you send me a sample with visible results because this is very difficult to see these shadows on my DVDs. Thx

[albain]

Hi, I use ffdshow in delphi app, and every time you guys
change iffdecoder I had to update my prog.
But you didn't update iffdecoder.h (nor delphi interface) in time.
Please update interface units.
Thanks

This file has not been updated ok but what should be added ?

I added 3 methods to manipulate the audio/subtitle streams inside. Are you interested in those ?

[STaRGaZeR]

@Stargazer : do you still have border problems with DVD subtitles ? Could you send me a sample with visible results because this is very difficult to see these shadows on my DVDs. Thx

Yup, with r3350 the probem is still there. Sample here (http://forum.doom9.org/showthread.php?p=1377467#post1377467). Also see this (http://forum.doom9.org/showthread.php?p=1389462#post1389462).

[swordsman]

Would you update IffDecoder.h in svn repository?
I need new interface alot.
Thanks

[Sebastiii]

@Sebastii : these borders are not normal. Do you have this problem in both RGB/YUV colorspace ? Could you post a sample ? Thanks


Hi Albain :)

My test and setting are :
FFDshow (no DXVA) 3350, RGB32 Output.

The sample : FDSAMPLE.rar (http://sebinternet.free.fr/FDSAMPLE.rar)

I've to try in YUY to see what happen :)
I get my sample @ work to see :)

Thx,
Seb.


Edit : In YUY it's not exactly the same issue

This is the screenshot :

http://sebinternet.free.fr/ffdshow_3350-3-YUY.png


Edit2 : it's like en echo (or a shadow on sub)

[clsid]

@Thunderbolt8
Stream switching in Haali works by dynamically reconstructing part of the filter graph. You can avoid your problems by not using reclock. Or use the Gabest matroska splitter, which results in a graph with multiple decoders, of which just one will be actually doing active decoding if you enable the audio switcher in either MPC or ffdshow.

[onomatopellan]

@albain: ok, I'll try only with rgb32 output for now, thanks.

@Sebastiii: I have downloaded your sample and with rgb32 output (disabling all Yxx and nv12 colorspaces) and it works ok. Maybe it's a drivers problem? I'm using ATI Catalyst 9.2.
http://i41.tinypic.com/sw71jl.png

[STaRGaZeR]

http://sebinternet.free.fr/ffdshow_3350-3.png

I have the same problem with Sebastii's sample, RGB32 output.

[ikarad]

I don't reproduce the crash. Where did you get the rev3350 ?

I'll bet on a bad compilation, revision 3350 includes the new ffmpeg libraries and a clean compilation has to be made

Try this build (32 bits) (http://damienbt.free.fr/ffdshow_rev3350_20100407_dbt_.exe)

xvidvideo.ru

[Sebastiii]

Thanks guy's to test it,

I have the same issue on 3 PC(s).
ATI 5770, Nvidia 9600GT and GMA Q965.

Thanks,
Seb.

[Thunderbolt8]

@Thunderbolt8
Stream switching in Haali works by dynamically reconstructing part of the filter graph. You can avoid your problems by not using reclock. Or use the Gabest matroska splitter, which results in a graph with multiple decoders, of which just one will be actually doing active decoding if you enable the audio switcher in either MPC or ffdshow.disabling reclock doesnt help, as ac3filter gets dropped anyway, no matter if reclock is used or not. and I need ac3filter to downmix to 2.0 channels.

I'd like to try with the internal matroska splitter, but it seems then I cannot choose the different (audio) tracks in the .mkv file any more. so I don't know how to change the audio tracks then, it doesnt work the way as with the haali splitter anymore.

[albain]

This is not a driver issue but it depends on the scaler you choose in DVD & bluray settings

Some scalers will clean those shadows, but there is indeed a problem. If you use no scaler you will see them clearly

Do you have the problem in YUV mode (apart the transparency problem) ?

It may be due to the SSE algorithm

[obieobieobie]

I get a blurry picture with AVC in 3351. I have postprocessing enabled and disabling it makes the picture go back to normal.

[clsid]

disabling reclock doesnt help, as ac3filter gets dropped anyway, no matter if reclock is used or not. and I need ac3filter to downmix to 2.0 channels.

I'd like to try with the internal matroska splitter, but it seems then I cannot choose the different (audio) tracks in the .mkv file any more. so I don't know how to change the audio tracks then, it doesnt work the way as with the haali splitter anymore.
ffdshow can downmix as well.

MPC menu -> Play -> Filters -> bottom filter -> there you can switch streams

[DigitalDeviant]

Albain, did you finally disable jitter correction for THD? With the lastest revisions I get huge jitter (30 segs or more) that is causing crashes, silence and audio desync with m2ts files here. No issues with matroska though, jitter is always at ~0. Can anybody confirm?

I've done some testing and I believe there was an issue created between 3337 and 3342 causing silence, desynched or garbled audio when seeking. I've only tested this with AVC/TruHD Blu-ray though

[DigitalDeviant]

My original multiple subtitle sample plays fine now, a million thanks Albain. However, here is another problem sample I just ran into.

In addition to some fugly lines through the yellow subs at the bottom and fugly shadowing (see pic), the subtitle positions get switched around.

http://img442.imageshack.us/img442/6645/badsub.th.png (http://img442.imageshack.us/i/badsub.png/)

Sample (http://www.mediafire.com/download.php?mgztymdmdjn)

[Thunderbolt8]

ffdshow can downmix as well.

MPC menu -> Play -> Filters -> bottom filter -> there you can switch streamsnot sure what you mean. theres at least no field or option 'bottom filter'. all the filters for each stream are listed seperately there so like 3x madflac, ac3filter and reclock. I can click on them, but it only gives the PIN info and such, but does change the audio stream. the last one is the .mkv file, but same here, it contains all the PIN infos about all streams and I can click them seperately here, but it doesn't change which stream is being played. so how can I find the option to change the streams? ;S

[EpsilonX]

I get a blurry picture with AVC in 3351. I have postprocessing enabled and disabling it makes the picture go back to normal.

I can confirm this as well...
3350 exhibit the same problem...
3347 is working fine...
Maybe the changes in 3347 got reverted back again... :D

[albain]

I can confirm this as well...
3350 exhibit the same problem...
3347 is working fine...
Maybe the changes in 3347 got reverted back again... :D

That's right, I didn't pay attention to the automatic merge. I'll revert this in the next revision

[fps]

I've got a problem with the DXVA part of ffdshow/MPC-HC.
Every time I seek (H.264, 720p) the picture becomes blocky until the next keyframe. This only happens in fullscreen mode, otherwise everything is still fine.
The problem only happens with ffdshow's DXVA enabled. Exactly the same applies to the internal filter of MPC-HC (DXVA).
On my Radeon HD 4830 this bug was introduced since Catalyst 10.1, previous versions work fine.

Win 7 32 bit
Radeon HD 4830 (Catalyst 10.1 and upwards)
Intel Q9400
ffdshow_rev3350_20100407_dbt_
MPC-HC svn 1782

[Sebastiii]

This is not a driver issue but it depends on the scaler you choose in DVD & bluray settings

Some scalers will clean those shadows, but there is indeed a problem. If you use no scaler you will see them clearly

Do you have the problem in YUV mode (apart the transparency problem) ?

It may be due to the SSE algorithm

Hi, i have change and effectively it's better.
I have tested in RGB32 and YUY -> OK (i have put Full)
Big Thanks,
Seb.

[err]

hallo, I would like to ask something, I use mpc (1775) with ffdshow tryouts 20100331_clsid_x64 and I encounter two problems:
1)I cannot get the haali media splitter work with hd audio no matter what I do I get unsynchronised audio with video very slow.Is there any workaround?
2)when i play videos other than mkv, with dolby digital sound (ac3, mp3 whatever) I cannot get the sound without interruptions.Every 10-20secs the sound stops for 2 secs and starts again.the codecs I use is ffdshow or the built in mpc codecs and I always get the same result.The only way to make this work is to set it to speakers instead of spdif in the built in player just as the screenshot and I get multichannel pcm in my receiver even though it is really 2.1.What can I do?I use radeon 5850 with 10.3 drivers.Thanx in advance

[mark0077]

Guys, I am using ffdshow mainly to decode, convert output to nv12, and change the interlace flag manually to Force Weave or Force Bob for my badly encoded DVDs.

My question is regarding conversion to nv12. When I am choosing to output nv12, along with force weave or force bob for the interlace output flag, does the "RGB Conversion" -> "Interlace or progressive" field have any impact.

It says its also applies to yuy2 <-> yv12 conversion but I am unsure should I be manually setting this to interlace or progressive when i manually change the interlace flag field?

[Thunderbolt8]

@Thunderbolt8
Stream switching in Haali works by dynamically reconstructing part of the filter graph. You can avoid your problems by not using reclock. Or use the Gabest matroska splitter, which results in a graph with multiple decoders, of which just one will be actually doing active decoding if you enable the audio switcher in either MPC or ffdshow.
is the internal mpc-hc splitter the same as the gabest splitter? if not, wheres the option to reactivate the gabest splitter? (haali installation gives the me choice to deactivate it when installing, but how can I activate it again without deinstalling haali?)

I tried my .mkv file with the mpc-hc internal matroska splitter instead of haali, but essentially it was the same. ac3filter remained listed as filter for all 3 tracks, but there was only sound input and output for the default 5.1 track - no sound (modification) through ac3filter for the other track inside that file.

[clsid]

Options -> Internal Filters

[Thunderbolt8]

but is the internal filter = gabest splitter? because when installing haali, Im being asked for both, whether I want to deactivate the internal and the gabest splitter which would not be necessary if both were the same thing.

[Keiyakusha]

but is the internal filter = gabest splitter? because when installing haali, Im being asked for both, whether I want to deactivate the internal and the gabest splitter which would not be necessary if both were the same thing.

They are the same. It just can be internal and external.
EDIT: internal means built-in into MPC-HC, external - separate filter.

[Thunderbolt8]

hm, but why has the haali installation them as 2 seperate entries you can checkmark, disable internal mpc-hc spitter and also disable gabest splitter?

[Keiyakusha]

because even if they are technically the same, they are not connected with each other.

[rica]

hallo, I would like to ask something, I use mpc (1775) with ffdshow tryouts 20100331_clsid_x64 and I encounter two problems:
1)I cannot get the haali media splitter work with hd audio no matter what I do I get unsynchronised audio with video very slow.Is there any workaround?
2)when i play videos other than mkv, with dolby digital sound (ac3, mp3 whatever) I cannot get the sound without interruptions.Every 10-20secs the sound stops for 2 secs and starts again.the codecs I use is ffdshow or the built in mpc codecs and I always get the same result.The only way to make this work is to set it to speakers instead of spdif in the built in player just as the screenshot and I get multichannel pcm in my receiver even though it is really 2.1.What can I do?I use radeon 5850 with 10.3 drivers.Thanx in advance


is the internal mpc-hc splitter the same as the gabest splitter? if not, wheres the option to reactivate the gabest splitter? (haali installation gives the me choice to deactivate it when installing, but how can I activate it again without deinstalling haali?)

I tried my .mkv file with the mpc-hc internal matroska splitter instead of haali, but essentially it was the same. ac3filter remained listed as filter for all 3 tracks, but there was only sound input and output for the default 5.1 track - no sound (modification) through ac3filter for the other track inside that file.

Guys, you may find out some common answers to your questions here:

(even you'll find the answer of how to make Haali as easily selectable filter in MPC-HC.)

http://forum.doom9.org/showthread.php?p=1370134#post1370134

Gabest splitter and Haali are completely different filters.
Gabest comes with MPC-HC (or as an external filter).
Haali is different.

Here is the comparison of them:

http://forum.doom9.org/showthread.php?p=1382497#post1382497

btw err, pls read my post carefully and you have two options:

1) Untick internal LPCM filter in MPC-HC and leave ffdshow configuration as LPCM enabled.
2) Untick LPCM filter in MPC-HC and unable LPCM in ffdshow and try WASAPI exclusive(for PCM) via ReClock (haali needed this time) and/or untick the rest of all audio formats and bitstream with DirectSound.

[Thunderbolt8]

why do I have to install mpc-hc after all other filters? or even reinstall it again after having changed a version of ffdshow or reclock?

and why to untick those splitter options for gabest etc. at haali installation? as soon as I tick 'matroska' for internal filters, it uses the internal filter and not haali any more.

[rica]

If you install haali with its default settings, even if you select (tick) internal filters in MPC, Haali will be selected as preferred renderer.

why do I have to install mpc-hc after all other filters? or even reinstall it again after having changed a version of ffdshow or reclock?


Sorry, it's not my choice, this is a "defacto" situation.

[swordsman]

Yes, I do

[Thunderbolt8]

If you install haali with its default settings, even if you select (tick) internal filters in MPC, Haali will be selected as preferred renderer.well, what do you mean by that? when I tick the internal renderer and I play a remuxed HD movie with flac audio I cannot see haali being loaded. no taskbar icon, nothing of haali mentioned in filters. the menu & track list in mpc-hc doesnt look like haali at all then.
Sorry, it's not my choice, this is a "defacto" situation."what" is the defacto situation? I definately can install mpc-hc after installing all the filters and it still plays all my remuxed HD movies fine. Ive always done it this way so far.

[swordsman]

Ok, nobody wants to update lffdecoder.h, so I can't recompile my prog to use new versions of ffdshow ;-(
Another question - is where any interface description to ffdshow DXVA?

[Keiyakusha]

If you install haali with its default settings, even if you select (tick) internal filters in MPC, Haali will be selected as preferred renderer.

Renderer? Do you mean splitter? If so, this is not true. Enabling internal splitter will override haali splitter. Haali won't be loaded.

[Thunderbolt8]

^^thats what I mean as well

[albain]

Ok, nobody wants to update lffdecoder.h, so I can't recompile my prog to use new versions of ffdshow ;-(
Another question - is where any interface description to ffdshow DXVA?

You can do it yourself, you know delphi's coding, the effort is not huge

I don't have time for now

[rica]

Renderer? Do you mean splitter? If so, this is not true. Enabling internal splitter will override haali splitter. Haali won't be loaded.

Yes it was a typo; isn't it very clear?


^^thats what I mean as well

I guess you were asking something, were you?

It seemed to me you were confusing splitters, so i gave the second link in my post.
The remaining is might be related with the only current configuration method of MPC-HC or Haali. (it's just the formal side of the content.)
But thanks anyway, i'll give it a go some time.

[Keiyakusha]

Yes it was a typo; isn't it very clear?
Yes, its very clear. So clear that normally no one will point that out, but some less experienced users may be confused. Anyway the main part of my post was about splitter loading.

[rica]

Anyway the main part of my post was about splitter loading.

Yes i know.
Let me switch off myself since it's not the right thread in which discusssing on unrelated stuff.

[Thunderbolt8]

ffdshow can downmix as well.

MPC menu -> Play -> Filters -> bottom filter -> there you can switch streams
well, ive tried it now with the internal splitter as well and also without reclock, but my problem still remains: ac3filter gets dropped or doesnt have any sound for the other tracks. so is this now a kind of bug? if so, what prog/filter is responsible?

[swordsman]

You can do it yourself, you know delphi's coding, the effort is not huge

I don't have time for now

Ok, I can do it myself, but what had I do if I don't
know what you changed in lffdecoder.h (which functions you add,
remove or rearranged)

[fastplayer]

Ok, I can do it myself, but what had I do if I don't
know what you changed in lffdecoder.h (which functions you add,
remove or rearranged)
...but you do know:
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/src/IffDecoder.h?view=log

[swordsman]

Yes, I have this one.
I thought, that as Albain said he "added 3 methods to manipulate the audio/subtitle streams inside", the interface was changed.
Sorry, my mistake.:stupid:

[clsid]

well, ive tried it now with the internal splitter as well and also without reclock, but my problem still remains: ac3filter gets dropped or doesnt have any sound for the other tracks. so is this now a kind of bug? if so, what prog/filter is responsible?
Are you using S/PDIF output? If so, it is probably related to the fact that only one filter instance can have exclusive access to it.

[Thunderbolt8]

Are you using S/PDIF output? If so, it is probably related to the fact that only one filter instance can have exclusive access to it.no, just normal output.

[hoborg]

Finally, in ATI Catalyst 10.5 preview, ATI introduced full h.264 DXVA decoding support including profile 5.1 and 16ref frames (DXVA 2.0)
Look on video on this page:
http://www.diit.cz/clanek/akcelerace-prehravani-videa-na-ati-radeonech-hd-konecne-funguje/36026/

[tetsuo55]

Finally, in ATI Catalyst 10.5 preview, ATI introduced full h.264 DXVA decoding support including profile 5.1 and 16ref frames (DXVA 2.0)
Look on video on this page:
http://www.diit.cz/clanek/akcelerace-prehravani-videa-na-ati-radeonech-hd-konecne-funguje/36026/I googled some more on this because that site is a bit vague.

I couldnt find much english chatter about this but based on google translate i read the following:

the 10.5 driver is a seperate branch from the 10.3/10.4 trunk.
Besides having a new avivo core that moves both VC1 and H264 to the "L5.1" limits and improving the transcoder many other things are changed.

OpenGL/CL drivers are up by 15 revisions!
Longstanding bugs in DirectX games appear to have been fixed, and performance is up in underperforming ones.

Please try it out! (also with the ffdshow dxva decoder)

[_xxl]

MSVC 2010 can't compile ffdshow.ax, a get errors with svn version of tuple.h. Has someone compiled successfully with MSVC2010?

[XhmikosR]

Nope, last time I tried MSVC 2010 I got some errors too.

http://pastebin.com/SMR7E2nz

[SamuriHL]

I've yet to download and install it. I hear there's some good improvements.

[hoborg]

Hi.
Devs with ATI, can you look on this sample (http://rapidshare.com/files/376893014/_aF_Gurren.Lagann.-.16.-.Zusammenfassung_E27E7284___1_-009.mkv)?
There is pixelation if ATI catalyst 10.3 is in use on my 5770 (WinXP).
Strange is, this doesnot happend if PDVD9 decoder (DXVA on) is in use.
Can be problem in FFDshow DXVA decoder?

[Snowknight26]

Any word on whether the lossless H.264 with DXVA (http://forum.doom9.org/showthread.php?p=1372277#post1372277) issue will be fixed?

[DMD]

ffdshow video is a problem or mpchc? :confused:

happens with 2 different releases of ffdshow:

pin SubPicure DVD navigator connected to pin MPC

ffdshow_rev2099_20080903_clsid_sse_icl10
http://www.allfreeportal.com/imghost2/thumbs/190149ffdshow_rev2099.png ('http://www.allfreeportal.com/imghost2/images/190149ffdshow_rev2099.png')


because the pin SubPicure MPC is not automatically linked?

one of the latest releases ffdshow_rev3356_20100411_sse_icl11
http://www.allfreeportal.com/imghost2/thumbs/28207ffdshow_rev3356.png ('http://www.allfreeportal.com/imghost2/images/28207ffdshow_rev3356.png')

[ikarad]

http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1
What is the difference between ffdshow_rev3361_20100415_xhmikosr.exe and ffdshow_rev3361_20100415_xhmikosr_icl11_core2.exe?
The second file is just for double core?

[tetsuo55]

Since albain is getting closer to a stable build i would like to ask everyone to help out getting a clean libavcodec for both ffdshow and mpc-hc.

the following looks like a clean solution:

src/libavcodec/ffmpeg/ (this contains the clean and complete export from ffmpeg

src/libavcodec/patches (this contains:)
* ffmpeg_directshow_patch (this turns it into a dshow filter)
* ffmpeg_mediafoundation_patch (this turns it into a MF filter)
* ffmpeg_h264_dxva_patch (adds DXVA1, 2 and HD)
* ffmpeg_mpeg2_dxva_patch (adds DXVA1, 2 and HD)
* ffmpeg_VC1_dxva_patch (adds DXVA1, 2 and HD)
* ffmpeg_windows_configpanel_patch (adds the settings panel)
* ffmpeg_bitstreaming_patch (adds support for bitstreaming (HD-)audio)

Both combined become the clean and correct codec.ax that can be used standalone, in ffdshow and in mpc-hc.

This way we can report bugs upstream again and updating becomes less painfull, also this part of the code can be completely shared between both projects lowering the time taken to work on that section of code.

[clsid]

Sorry, but this is a useless suggestion. First of all, except for dxva, all of those things are done outside of the ffmpeg library. Secondly, it is totally impractical to maintain code like you suggest.

If you want to simplify things for MPC, then consider using ffdshow's libavcodec.dll instead of linking it directly into MPC/MPCVideodec. Or better yet, stop trying to replicate ffdshow functionality and just tell MPC users to use ffdshow.

[tetsuo55]

Sorry, but this is a useless suggestion. First of all, except for dxva, all of those things are done outside of the ffmpeg library. Secondly, it is totally impractical to maintain code like you suggest.

If you want to simplify things for MPC, then consider using ffdshow's libavcodec.dll instead of linking it directly into MPC/MPCVideodec. Or better yet, stop trying to replicate ffdshow functionality and just tell MPC users to use ffdshow.What you suggest is the reason for my post. We still intend to include libavcodec but as a direct clean mirror from ffdshow instead.

The current implementation is messy and difficult to maintain, your post does not suggest an alternative that makes it easier.

[Keiyakusha]

is it possible to strip all decoders in mpc-hc (including DXVA ones), integate ffdshow in mpc-hc so there will be ffdshow settings page where internal filters are located and bundle ffdshow with mpc-hc?

[clsid]

The current implementation is messy and difficult to maintain, your post does not suggest an alternative that makes it easier. Exactly what do you think is messy? Updating the FFmpeg code in MPC takes maybe half an hour. A few minutes per update if done regularly. There is little to none MPC-specific code left, so most files can just be copied from ffdshow.
is it possible to strip all decoders in mpc-hc (including DXVA ones), integate ffdshow in mpc-hc so there will be ffdshow settings page where internal filters are located and bundle ffdshow with mpc-hc? It is possible to compile MPC without internal decoders. There is no need to 'integrate' ffdshow. Just install it and it will automagically get used when MPC has no internal filters enabled/included.

[STaRGaZeR]

Another bug bites the dust: wrong shadows in DVD subtitles fixed in r3363!

[Keiyakusha]

It is possible to compile MPC without internal decoders. There is no need to 'integrate' ffdshow. Just install it and it will automagically get used when MPC has no internal filters enabled/included.

What I mean is some people was against MPC-HC without internal filters but I don't understand why do MPC-HC still needs them. Why not just drop and forget about them? In what way internal filters better than ffdshow? So I thought maybe some more integration will be fine for these people...

[STaRGaZeR]

@devs

I've fixed a problem with SSA subs. If a color is written as an integer instead of an hexadecimal and the value is negative, ffdshow outputs a greenish color while MPC outputs black, since a negative value doesn't make any sense here. With this patch ffdshow is consistent with VSFilter and MPC's internal subtitle renderer.

Patch: http://www.mediafire.com/?4wwhlztljlm
Build: http://www.mediafire.com/?lizgymxzyzh
SSA script for testing: http://www.mediafire.com/?gntnyw2tkwv

[tetsuo55]

Exactly what do you think is messy? Updating the FFmpeg code in MPC takes maybe half an hour. A few minutes per update if done regularly. There is little to none MPC-specific code left, so most files can just be copied from ffdshow.Several developers have tried to update ffmpeg but have been unable to do so.

You are most experienced in this area, could you write a detailed guide to updating? We could take over from there.

What I mean is some people was against MPC-HC without internal filters but I don't understand why do MPC-HC still needs them. Why not just drop and forget about them? In what way internal filters better than ffdshow? So I thought maybe some more integration will be fine for these people...MPC-HC is a standalone player that does not need any codecs or admin rights to work, this way it can be used on locked down environments.

[madshi]

MPC-HC is a standalone player that does not need any codecs or admin rights to work, this way it can be used on locked down environments.
I understand your goal, but I think you're realizing it the "wrong" way. Of course you can try compiling all needed filters into MPC, but I dare say that if you do that, you will always lose out somewhere because there are always going to be external (open or closed source) filters that will be better than MPC internal filters. Your current approach to MPC means that you're giving up on external filters in locked down environments.

IMHO the "proper" way to handle locked down environments would be to:

(1) Add a folder named "external filters" to the MPC root folder. Users could just drop in any external codec files there.
(2) MPC would enumerate through these files, load all of them via LoadLibrary, and call their "DllRegisterServer" routines during MPC initialization.
(3) To make (2) succeed even in locked down environments, MPC would hook all registry related APIs (by using Detours) and build up a temporary virtual registry storage (in RAM only). If you do this correctly, the DllRegisterServer method of most external filters should succeed, and the registration would be in your virtual registry storage. Also you should be able to fully use the external filters this way, in the same way, as if they were installed for real.

I think something like the approach above is the usual way to handle locked down environments. Instead of avoiding admin stuff, you emulate/virtualize it by hooking the APIs. Of course doing it this way is somewhat complicated. And it might fail with a few external filters (e.g. it might fail with filters that have some kind of copy protection). But I'm quite sure that it would work with by far most free external filters.

This logic would also have the added benefit that even in non-locked down environments MPC HC users could use many external filters without having to install them.

[tetsuo55]

Your description is very close to how MPC(-HC) already works.
The only difference is that we include the codecs internally as backups for when no codec is installed on the OS, or you have not selected a specific (registered or unregistered) codec.


To make MPC-HC behave like you have described all you have to do is uncheck the internal codecs, in that case they will only be used when there is no alternative. To get an unregistered codec working you select it from the external codec page.

For some features to work correctly we need the codecs to be internal (mostly DXVA-1)

I believe we should take a more modular approach and bring these two projects closer together.
It would be nice to have a single file for the player, codecs and postprocessing but on the other hand be able to install only 1 of the 3.

[madshi]

Your description is very close to how MPC(-HC) already works.
Keiyakusha asked why MPC HC still has internal filters. Your reply was that they're needed for locked-down environments. Now you're contradacting yourself. Either the internal filters are needed for locked-down environments or they're not. If they're not needed, your reply to Keiyakusha was incorrect. If they are needed, your reply to me is incorrect. I'm confused... :confused:

[tetsuo55]

Good point.

I oversimplified my reply to Keiyakusha.

the correct reply is:

MPC-HC includes codecs to provide a backup for when the OS or user do not provide the required codec for the file. This also provides us with the opertunity to add some advanced features like DXVA1 support with subtitles.

By default the included codecs are prefered over external ones to prevent codec hell, improve the quality of feature requests and bug reports and to make the dxva codecs available by default.

[It's not only codecs though, the same is true for source filters, splitters(parser filters), subtitle renderer and audio-/video-renderers]

[madshi]

Ok, thanks, that makes more sense to me... :)

But then the question is: Instead of trying to compile ffdshow into MPC HC, wouldn't it make more sense to remove those parts from MPC HC which are duplicate to ffdshow and instead rely on ffdshow to take over that work? Maybe you could distribute the full standalone ffdshow package with MPC HC to make things easier for MPC HC users (don't know if that's possible legally, though). You could also ship other external filters with MPC HC, if you like them and their license allows that.

IMHO the focus of the MPC HC team should *not* be to replace all external filters with internal stuff. The focus instead should be to give MPC HC users the best possible experience out of the box. If there are external filters available that are better than the internal filters, MPC HC should consider using those external filters by default and (as said above) maybe even distribute them with MPC HC, if legally possible. That would IMHO be the best approach for MPC HC users, although maybe it could hurt the pride of MPC HC developers...

But this kind of discussion probably belongs into the MPC HC thread...

[tetsuo55]

what you describe is basically what codecpacks like k-lite and cccp do.

We do not intend to do any work on the codecs, the idea to make ffdshows ffmpeg implementation cleaner is to remove the worry about codecs from the mpc-hc team completley so they can focus on stuff that actually matters.

[madshi]

what you describe is basically what codecpacks like k-lite and cccp do.
In case you haven't noticed, MPC HC currently is a codecpack, too! ;) It's just a more limited codecpack, because it limits itself to only self-written codecs, some of which are good, some are less good, and many are duplicates to ffdshow.

We do not intend to do any work on the codecs, the idea to make ffdshows ffmpeg implementation cleaner is to remove the worry about codecs from the mpc-hc team completley so they can focus on stuff that actually matters.
So in other words: MPC HC will stay a codecpack, and since you don't plan to work on the codecs, it will only get more outdated over time. And to make things even worse, you plan to continue prefering most internal codecs of your codecpack over external codecs by default.

IMHO you should invest some thinking over whether you want MPC HC to be a codecpack or not. If yes, you should go the full mile and include all codecs (external or not) which you find best for any given format. If no, you should move all good MPC HC filters to an external package and make MPC HC a pure media player. Right now it's neither fish nor flesh. Maybe ffdshow could be extended to also home source filters? This way all good MPC HC filters could be moved to ffdshow. But of course that's only my 2 cents... :)

[tetsuo55]

In case you haven't noticed, MPC HC currently is a codecpack, too! ;) It's just a more limited codecpack, because it limits itself to only self-written codecs, some of which are good, some are less good, and many are duplicates to ffdshow.


So in other words: MPC HC will stay a codecpack, and since you don't plan to work on the codecs, it will only get more outdated over time. And to make things even worse, you plan to continue prefering most internal codecs of your codecpack over external codecs by default.

IMHO you should invest some thinking over whether you want MPC HC to be a codecpack or not. If yes, you should go the full mile and include all codecs (external or not) which you find best for any given format. If no, you should move all good MPC HC filters to an external package and make MPC HC a pure media player. Right now it's neither fish nor flesh. Maybe ffdshow could be extended to also home source filters? This way all good MPC HC filters could be moved to ffdshow. But of course that's only my 2 cents... :)The same is true for all players, we have no intention to change the current behaviour. We are also the filter upstream for many open and closed source projects so changes there could wreak havok downstream.

The only thing we are trying to do (and albain has already done most of the work here) is to use ffdshows libavcodec as the (maintainable) upstream for the ffmpeg codecs included in mpc-hc as the code is 99% the same but updating it twice takes double the effort.

(Also filter A is better than filter B will have to be proven with a scientific test, we simply use the best open source upstream which is ffmpeg)

[clsid]

MPC already has the functionality to load external filters without requiring those filters to be registered on the system (/filter command line parameter). That could be used to replace the internal decoders if the functionality is extended with:
1) An optional bundle of recommended external filters that can be installed (or extracted) in specific subdir of MPC location. This could contain stuff like ffdshow, haali, madflac, etc.
2) Ability to select in MPC options which of those filters should be used. Filters that are detected as registered on the system should be hidden (so that registered filters take precedence over the bundled filters). There should also be buttons for opening filter properties dialogs to configure their settings.
3) MPC should create a virtual registry for these filters to store their settings, like madshi suggested. This must be stored as a file, so settings are remembered. A file with initial recommended settings must be included in the filter bundle. The virtual registry file should be in some user readable format, so that it can be easily edited.

The above would allow MPC to make use of many excellent filters, while still being portable. The only difficult part is the virtual registry.

[tetsuo55]

I do think this change in filter loading behaviour is a good idea regardless of where and how we store the default codecs.

A patch that enables this (or at least feature request on the tracker) would be great.
(We have few devs, and those we have are not interested in changes like this, which is one of the reasons for upstreaming ffmpeg)

Maybe this creates some more clarity.

A dshow graph can be very complex and mpc-hc provides every required filter internally.
example:

player > source filter > splitter >
> video codec > video software postprocessor > video renderer (> hardware postprocess)
> Subtitle renderer > video renderer (either directly or through a 2nd surface)
> audio codec > audio software postprocessor > audio renderer (> hardware postprocess)

Thats about 10 filters (but currently mpc-hc does not offer postprocessing as both gabest and casimir seem to not like that, if they had both projects would have merged a long time ago)

[clsid]

Don't expect use to help out with the internal decoders of MPC. We don't have the manpower for it either, nor do we care much about it, since we obviously prefer to use ffdshow instead. I don't mind updating the ffmpeg stuff in MPC once in a while (when I have time), but that is all we can do for you.

[tetsuo55]

Yes you are right, we don't expect any help for our side of the story, mpc-hc devs will have to fix stuff in the ffdshow trunk.

[ikarad]

http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1
What is the difference between ffdshow_rev3361_20100415_xhmikosr.exe and ffdshow_rev3361_20100415_xhmikosr_icl11_core2.exe?
The second file is just for double core?

up! thanks

[fastplayer]

up! thanks
Read the first post of this thread, the FAQ, Google etc.

[ikarad]

Read the first post of this thread, the FAQ, Google etc.

It's not explained.
I don't speak about icl11 but about core2 in the name of file

[hoborg]

Hi devs.
I didnt think you will have some free time to look, but if you have - this sample (http://hobring.ic.cz/samples/mp4.AAC+AVC1_(1280x720@50fps).rar) (recorded by SAMSUNG camcodrder) doesnot work using FFDshow DXVA decoder - you will see only black screen if you will use MPC-HC MP4 splitter.
Becouse it is working with standart FFDshow decoder, it seems to be some kind of DXVA decoder issue.

It will start to work by using haali as the splitter or Cyberlink h.264 DXVA decoder, or by conversion in to MKV (MKV merge) - but that is not solution for me :(
Thanks.

[kieranrk]

In case you haven't noticed, MPC HC currently is a codecpack, too! ;)

Except without the conflicts, crashes and ancient ACM/VCM libraries that codec packs have. (also the hundreds of tray icons)

In my opinion the setup is fine. It's the right mix between standalone player and mplayer/vlc type app with support for hundreds of codecs and dozens of various options. Though there is some cruft like RealMedia and Quicktime support that's a bit dated.

Also internal DXVA is better than anything else for 1080p50 material.

[clsid]

That is just nonsense. A good codec pack will give you LESS problems than MPC-HC with all of its internal filters enabled.

[tetsuo55]

do you have any proof of that?

[Keiyakusha]

Also internal DXVA is better than anything else for 1080p50 material.
Not as part of the current discussion about codec packs and stuff I want to ask what is the difference compared to ffdshow?

[clsid]

do you have any proof of that?Were you responding to me? If so, there are a few examples on MPC's bugtracker, like the mod16 bug, amr issues. Those bugs do not exist in ffdshow.

[tetsuo55]

Were you responding to me? If so, there are a few examples on MPC's bugtracker, like the mod16 bug, amr issues. Those bugs do not exist in ffdshow.afaik those exist because our version of ffmpeg is broken, hence why i rather do a straight copy from ffdshow's one.

[kieranrk]

That is just nonsense. A good codec pack will give you LESS problems than MPC-HC with all of its internal filters enabled.

oh dear...

Not as part of the current discussion about codec packs and stuff I want to ask what is the difference compared to ffdshow?

The various copying between GPU and PC memory isn't suitable for high-bandwidth 1080p50.

[clsid]

afaik those exist because our version of ffmpeg is broken, hence why i rather do a straight copy from ffdshow's one.The ffmpeg code isn't the cause, its the code that uses it that needs work.
oh dear...
Indeed.

[kieranrk]

Indeed.

I'm not going to go there.

[Mixer73]

With respect, I disagree strongly to clsid, MPC-HC has a place as an all in one that works REALLY well when lots of other things don't, especially for users who don't know enough to mess with filter priority and stuff, that can be pretty evil.

Especially once machines have had codec packs loaded on them and stuff no longer works, MPC-HC is like old faithful.

[Keiyakusha]

With respect, I disagree strongly to clsid, MPC-HC has a place as an all in one that works REALLY well when lots of other things don't, especially for users who don't know enough to mess with filter priority and stuff, that can be pretty evil.

Especially once machines have had codec packs loaded on them and stuff no longer works, MPC-HC is like old faithful.

I'm sorry but I think this is not something where you can disagree with clsid with only these arguments. This discussion started from proposition to make something, so maintaining internal filters and bug fixing will be easier and not because someone just wants to change something for no reason. And it was said that mpc-hc team have no plans to work on codecs.
Also even with replacing internal filters by ffdshow it is possible to have no admin rights requirement and playability on old systems "where all codecs messed up". However yes, it will require some amount of work.

[madshi]

Except without the conflicts, crashes and ancient ACM/VCM libraries that codec packs have. (also the hundreds of tray icons)
I'm not saying that the well known big codec packs out there would be any good. But that doesn't mean that it wouldn't be possible to create a good codec pack.

In my opinion the setup is fine. It's the right mix between standalone player and mplayer/vlc type app with support for hundreds of codecs and dozens of various options. Though there is some cruft like RealMedia and Quicktime support that's a bit dated.
MPC HC is currently my main media player, and I really like it. I don't find the internal filter logic (preferring most internal filters over external filters, without considering the filter quality) right, though.

Also internal DXVA is better than anything else for 1080p50 material.
Why is ffdshow slower? Probably because it wants to enable post processing and subtitles, right? I think it would be possible to solve that problem. E.g. post processing could be disabled to avoid that GPU <-> RAM copy stuff. Subtitles could be sent to the video renderer directly, I'd guess. If all else fails, the MPC HC video renderers could expose some new interfaces to make ffdshow DXVA subtitle rendering possible in a fast way. Not sure how much work that would be. But I'm quite sure it would be possible to implement DXVA in ffdshow without speed penalty, as long as the user doesn't require post processing.

With respect, I disagree strongly to clsid, MPC-HC has a place as an all in one that works REALLY well when lots of other things don't, especially for users who don't know enough to mess with filter priority and stuff, that can be pretty evil.

Especially once machines have had codec packs loaded on them and stuff no longer works, MPC-HC is like old faithful.
There's no doubt that bad codec packs will mess up anyone's PC. But that's not the point. clsid was talking about a *good* codec pack. E.g. imagine a codec pack only consisting of MPC HC and ffdshow, where all those MPC HC's internal filters which are duplicates to ffdshow are removed, and where MPC HC automatically "preferred" ffdshow instead. That would be a pretty good start for a "good" codec pack. It would still have all the positive properties you mentioned, plus it would fix a couple of bugs. There may be a few more freeware filters which would be worth being part of such a codec pack, but that's a whole different topic.

Please understand that we're not trying to bully the MPC HC folks into distributing ffdshow with MPC HC. It's just that according to tetsuo55, MPC HC's aim is to provide an all-in-one solution. And if I, personally, were trying to create a really good all-in-one solution, I would find no way around including some selected external DirectShow filters.

[tetsuo55]

What kierank means with DXVA is that the implementation in ffdshow needs to do a very slow system memory > gpu memory copy operation.

about 3rd party codecs, almost every codec that we can legally include is equal to or worse than ffmpeg.
I would gladly hear about cases where another open source codec is better than ffmpeg.

[madshi]

What kierank means with DXVA is that the implementation in ffdshow needs to do a very slow system memory > gpu memory copy operation.
Yes, and what I mean is that there would probably be ways to avoid that under specific circumstances (e.g. with disabled post processing and subtitles, see previous post).

about 3rd party codecs, almost every codec that we can legally include is equal to or worse than ffmpeg.
I would gladly hear about cases where another open source codec is better than ffmpeg.
Why limiting things to open source? I was talking about *freeware* 3rd party codecs, which you could include in the MPC HC distribution to create an even better all-in-one solution. Examples would be ffdshow, Haali Media Splitter, madFlac, ... It's quite possible that Haali's license wouldn't allow you to redistribute it, though. Again: I'm not trying to bullying you into include the mentioned filters. I'm just saying that if a great all-in-one solution is your aim, it might make sense to include a few selected external filters. Or at least, if you detect that some specific external filters are installed, you could disable some internal filters by default and prefer known good external filters instead. Just a suggestion, though. Any way in which MPC HC could be improved to automatically work better with good external filters, without having to tweak MPC HC settings, would be nice.

[clsid]

Postprocessing is disabled by default in ffdshow. The small speed difference might be due to colorspace conversion and some sanity checks that MPC does not perform. Btw, MPC does not have any multi-threaded decoding, so on a multicore CPU ffdshow will be much faster with decoding H.264 in software mode.

It always makes me laugh when people say that codec packs are evil and break your system. While that may be true for ancient stuff such as Nimo and ACE, it is just blatantly untrue for any modern codec pack such as CCCP or K-Lite. In fact, for example K-Lite is equipped with special functionality to fix things. This includes detection (and removal) of several known bad codecs that cause crashes, memory leaks and other stability issues. It includes detection (and removal) of incorrect codec references in the registry (left-over stuff). It includes detection of some known audio and DirectShow issues. It includes a proper uninstaller. It adjusts what it installs based on what is already installed. Everything can be customized. Etc. Etc.

Btw, the update that I did yesterday has made software H264 decoding in MPC significantly faster. Your welcome ;)

[tetsuo55]

It is possible to make a more complicated dshow fallback system than we currently have in MPC-HC and i can see how this improves the experience for some.

We believe that software should be open source, that is the main reason for not including closed source filters.
I think k-lite/cccp solve the problem quite nicely as their codec choices favor open source and free codecs replacing them only with others when they are truly better.

The problem with ffdshow's DXVA implementation has nothing to do with the postprocessing its a design flaw in DXVA2 (mpc-hc uses DXVA1 which doesn't suffer from this issue but has other problems).

[Shankster]

How about FFdshow guys help MPC guys with what they want and the MPC guys fix ffdshow's vobsub implementation. :eek:

/crawling back under my rock

[tetsuo55]

The idea is to create a new subtitle renderer and with that a new subtitle format.
This new renderer should be a drop in replacement for the current stuff in ffdshow/mpc-hc
the lead dev for that project is looking for help though, he needs help with brainstorming and coding. (PM me if you are willing and able to help out)

[kieranrk]

It always makes me laugh when people say that codec packs are evil and break your system. While that may be true for ancient stuff such as Nimo and ACE, it is just blatantly untrue for any modern codec pack such as CCCP or K-Lite. In fact, for example K-Lite is equipped with special functionality to fix things. This includes detection (and removal) of several known bad codecs that cause crashes, memory leaks and other stability issues. It includes detection (and removal) of incorrect codec references in the registry (left-over stuff). It includes detection of some known audio and DirectShow issues. It includes a proper uninstaller. It adjusts what it installs based on what is already installed. Everything can be customized. Etc. Etc.


One could go on for pages and pages as to why K-lite has major problems...I'm not going to do it because it will turn this thread into a flamefest. CCCP is ffdshow and mpc-hc so is not comparable.

[madshi]

We believe that software should be open source, that is the main reason for not including closed source filters.
So, you make your decisions based on politics, instead of quality?

[clsid]

One could go on for pages and pages as to why K-lite has major problems...I'm not going to do it because it will turn this thread into a flamefest. CCCP is ffdshow and mpc-hc so is not comparable.
Your last comment just proves how little you know about this. K-Lite has different variants, of which the Standard one is very comparable to CCCP in terms of contents.
I love to hear about those so-called problems, but since you don't actually use the pack, I doubt you will be able to come up with much valid stuff.

[ikarad]

The idea is to create a new subtitle renderer and with that a new subtitle format.
This new renderer should be a drop in replacement for the current stuff in ffdshow/mpc-hc
the lead dev for that project is looking for help though, he needs help with brainstorming and coding. (PM me if you are willing and able to help out)

Why don't you include subtitle renderer from ffdshow in mpc-hc because your subtitle renderer is very buggued with a bad blu-ray subs support (with ffdshow it is nearly perfect under vista and seven (even if it doesn't work under xp) thanks to the amazing work of Albain)?

[tetsuo55]

So, you make your decisions based on politics, instead of quality?As you might have noticed there is a ton of politics in both these programs.
I hope everyone will move to a new scientifically best best model instead, but it will remain based on open-source.
being/using open source is one of mpc-hc's core ideoligies and one of our unique selling points as a dshow based player.

Why don't you include subtitle renderer from ffdshow in mpc-hc because your subtitle renderer is very buggued with a bad blu-ray subs support (with ffdshow it is nearly perfect under vista and seven (even if it doesn't work under xp) thanks to the amazing work of Albain)?they are incompatible unfortunately, we need devs that want to work on this type of changes.

[ikarad]

@Albain,
there is a bug with blu-ray sub support with xp sp3 (under vista it works very well).

The subs are divide in two parts and are not completed
http://nsa14.casimages.com/img/2010/04/20/mini_100420050119368677.jpg (http://www.casimages.com/img.php?i=100420050119368677.jpg)


I try with YUY2 or RGB 32 and there is the same problem.
I try with all versions of ffdshow (3350 to 3368) where it works under vista and there is the same problem udner xp sp3.

I notice some crashs but I think that crashs come from this problem .

same example that I give you before
example
part 1
http://www.zshare.net/info.html?7219...12ca4e70f4959e
part2
http://www.zshare.net/info.html?7222...cb0c45e37239c1
part3
http://www.zshare.net/info.html?7222...fa215e89fcfdf2

part 4
http://www.zshare.net/info.html?7287...1fac7d6620cda5

part 5
http://www.zshare.net/info.html?7289...835142beff854c
part 6
http://www.zshare.net/info.html?7290...3fe6974fb1237c

partie 7
http://www.zshare.net/info.html?7215...a94c0fc61987f5

[kieranrk]

Your last comment just proves how little you know about this. K-Lite has different variants, of which the Standard one is very comparable to CCCP in terms of contents.
I love to hear about those so-called problems, but since you don't actually use the pack, I doubt you will be able to come up with much valid stuff.

No I would never pollute my PC with K-lite. Why do you think Microsoft started putting native codecs in Windows 7? Because normal people, many of whom I know have installed codec packs such as K-lite which lead to crashes and conflicts beyond belief.

For a start in K-lite Basic/Standard you include:

Haali Video Renderer [version 1.10.120.15]
Hasn't been updated for ages. Is it that fundamental to watching videos that it deserves to be in the Basic pack?

In Standard:

MPEG-2 (Cyberlink) [version 8.4.0.1014]
And whats wrong with ffdshow's mpeg-2 decoder?

And to summarise what would be pages and pages for the other variants:

You take the Adobe Acrobat Reader approach to codecs. Include as much obscure stuff in the codec pack as possible just like Adobe includes all sorts of obscure plugins. Add so much bloat so that the 0.00001% of occasions when such a thing is needed it's there. There is also ridiculous amounts of duplication for no reason at all: ACM codecs, VFW codecs and even more obscure formats. How many AAC/AC3/MPEG-2 etc decoders do we really need according to you?

[Midzuki]

And whats wrong with ffdshow's mpeg-2 decoder?

http://forum.doom9.org/showthread.php?t=148056

for example. :)

[kieranrk]

http://forum.doom9.org/showthread.php?t=148056

for example. :)

That's a subtitling issue though, not an mpeg-2 issue.

[STaRGaZeR]

I'm sure all of you have your own opinions about codec packs but this thread is about ffdshow :rolleyes:

Did anyone review this (http://forum.doom9.org/showthread.php?p=1392884#post1392884) patch/build?

[Midzuki]

That's a subtitling issue though, not an mpeg-2 issue.

Well, the stupidly-resized subtitles are produced by the mpeg-2 decoder from ffdshow, whenever it is allowed to connect to the DVD-Navigator.

And yes, the non-independent "raw video filter" has the same issue.

[clsid]

No I would never pollute my PC with K-lite. Why do you think Microsoft started putting native codecs in Windows 7? Because normal people, many of whom I know have installed codec packs such as K-lite which lead to crashes and conflicts beyond belief.Windows 7 includes more codecs primarily to give users a better out-of-box experience. For example, the H.264 support was essential for Media Center, since H.264 is now also used in digital TV broadcasts.
Btw, K-Lite is on the official compatible application list of Windows 7, added on the initiative of Microsoft themselves!
Crashes are just nonsense. Pretty much all crash reports I deal with end up being caused by faulty drivers. The component in the pack that is the least stable is actually MPC-HC.
For a start in K-lite Basic/Standard you include:
Haali Video Renderer [version 1.10.120.15]
Hasn't been updated for ages. Is it that fundamental to watching videos that it deserves to be in the Basic pack?
It's not fundamental. It is normally bundled with Haali splitter, which is also included, so it is included to keep things similar to standalone installers. It is very small compressed and has no ill effects when not used.
In Standard:
MPEG-2 (Cyberlink) [version 8.4.0.1014]
And whats wrong with ffdshow's mpeg-2 decoder?Lots of things unfortunately.
You take the Adobe Acrobat Reader approach to codecs. Include as much obscure stuff in the codec pack as possible just like Adobe includes all sorts of obscure plugins. Add so much bloat so that the 0.00001% of occasions when such a thing is needed it's there. There is also ridiculous amounts of duplication for no reason at all: ACM codecs, VFW codecs and even more obscure formats. How many AAC/AC3/MPEG-2 etc decoders do we really need according to you?Things are included based on the needs of the users, which are millions of very satisfied ones btw. You just looked at the contents lists, not at the actual installers themselves. Then you would have noticed that not everything actually gets installed. Choices must be made between redundant items. Redundance in contents is there when there are good reasons for it, like known limitations in items, or significant groups of users that prefer specific alternative items. Items that the average user does not need are disabled by default.
Native codecs and popular third party codecs (like CoreAVC) are automatically detected so that the installer can adapt itself. So by default no MPEG-2 decoder will be installed on Vista Home Premium because that already contains a good decoder.
The fact that you don't need many things does not mean that the same applies to others. That is why the installers are fully customizable and there are different sized packs. What you call bloat is what others call essential functionality.
Even when a user decides to install more than he/she actually needs, then there is no harm other than a few MB of wasted space.

[kieranrk]

That is why the installers are fully customizable and there are different sized packs. What you call bloat is what others call essential functionality.

I did look at the installers and they still allow duplicates in terms of VFW, ACM and the likes. Xvid/ffdshow too. It's clearly bloat under the misguided attempt to second guess what people want or you feel that implementations are incomplete based on conjecture and not real evidence.

You haven't revealed specifically what is wrong with ffmpeg's mpeg-2 decoder or any of the other open source decoders and tetsuo55 has also said that you have a "secret collection of bugreports" whatever that means. You do not choose to share these bugreports so it's pretty clear that you have no idea what you are doing or think people are psychic.

You are doing a disservice to people by offering so much unnecessary cruft in the codec-pack then using the installer to patch over any potential conflicts. This is a poor way of offering software to end-users.

Even when a user decides to install more than he/she actually needs, then there is no harm other than a few MB of wasted space.

Security holes, longer time for directshow to connect to pins.

Anyway the overall point is that neither mpc-hc nor ffdshow should end up like k-lite because it's a terrible example to follow.

[clsid]

I did look at the installers and they still allow duplicates in terms of VFW, ACM and the likes. Xvid/ffdshow too. It's clearly bloat under the misguided attempt to second guess what people want or you feel that implementations are incomplete based on conjecture and not real evidence.You need to be more specific about what you think are duplicates. VFW/ACM codecs are primarily for applications that do not use DirectShow. DirectShow apps can use them as fallback, but that is obviously not their main purpose.
I don't second guess what people want or need, it is based on years of experience and actually talking to people. That is why I am also engaging in this discussion with you. Your opinion is also valued like everyone elses, even when I may disagree on certain points.
You haven't revealed specifically what is wrong with ffmpeg's mpeg-2 decoder or any of the other open source decoders and tetsuo55 has also said that you have a "secret collection of bugreports" whatever that means. You do not choose to share these bugreports so it's pretty clear that you have no idea what you are doing or think people are psychic.There is no secret list of bug reports. There is just lots of info in my head gathered from forums like this. I actually do forward important info to the appropriate developers. So you are really really out of line here.
As for specific mpeg-2 related issues, I suggest you search this topic and have a look at the ffdshow forum and bugtracker. For example a video corruption issue can be found a few pages back.
Security holes, longer time for directshow to connect to pins.Which holes? Please mention specific ones, instead of hypothetical ones. Did you know some holes in MPC were fixed long ago partly due to my efforts? I guess not.
The number of filters installed by the average pack is just a fraction of what is already part of Windows. Filter enumeration pin connection in done in the order of milliseconds.

[clsid]

I should add that the philosophy is of course to install as little as possible. That is why the smaller variants are the ones that are recommend for the average user. But the reality is that most want maximum functionality and do not care about a bit of bloat. They only care about working playback, which is what the pack provides.

[Mixer73]

While I have no problem with codec packs for the less expert user, I personally have found that they lead to less stable machines over time. So I choose to install applications or codecs specifically and very carefully.

I also have seen machines recently that were totally unable to play videos in WMP and even PowerDVD, but MPC-HC worked 100%. Its able to play even when system codecs are borked, and if we strip internal filters, this will no longer be the case.

YMMV, and yes, clsid's criticism that I do not use the packs means that my comment could well be out of date; but if I can install 2-3 things I specifically need, then I don't see a need to have a pack.

Hell, I won't even install Haali Splitter on my machines, its got far too many bugs that seem to affect me. I remux everything back to MP4 because that works perfectly!

Maybe I'm strange.

[_xxl]

n Standard:
MPEG-2 (Cyberlink) [version 8.4.0.1014]
And whats wrong with ffdshow's mpeg-2 decoder?
ffdshow's libavcodec decoder crashes with some mpeg2 samples, for example.

[lych_necross]

Maybe I'm strange.
No you're not strange. Discussing codec packs in a ffdshow thread is strange, off topic, and counter productive. :readrule:

[tetsuo55]

ffdshow is itself a kind of open source codecpack

[Mr VacBob]

ffdshow's libavcodec decoder crashes with some mpeg2 samples, for example.

Where are these?

[starla]

How about opening a separate thread for the code packs are evil discussion and keeping this thread in technical details of ffdshow?

br,
tourettes / MediaPortal

ps. I wont even say aloud my opinion on codec packs (I have seen enough damage different codec packs have done. Like unregistering filters that they think aren't needed...)

[clsid]

Like unregistering filters that they think aren't neededSuch as?

[dann23]

I also don't like codec packs. So can we keep this thread clean?

[namaiki]

Just curious, but what is the minimum amount of keys that will tell ffdshow video to output only RGB formats with dithering?

I tried the following, but it seems to only enable dithering.


Windows Registry Editor Version 5.00

[-HKEY_CURRENT_USER\Software\GNU\ffdshow]

[HKEY_CURRENT_USER\Software\GNU\ffdshow\default]
"dithering"=dword:00000001
"outYV12"=dword:00000000
"outYUY2"=dword:00000000
"outYVYU"=dword:00000000
"outUYVY"=dword:00000000

[adam777]

Hello all,
Revision 3368 breaks Hebrew subtitles - words are not displayed in correct order, letters inside words display correctly.

[albain]

Hi all,

very intense dicussions :-)

As clsid said we on ffdshow team don't have the time and the manpower to work on MPC filters and MPC in general

Tetsuo, you are the one who proposed this idea to import ffdshow inside MPC HC a while ago. After all it may be better to design interfaces or improve existing ones in ffdshow to make a better integration

As a media center user, I am not a MPC user so I still think that the codec efforts should be unified into one place.
Besides, directshow support will be dropped in the future (the question is not if but when) and we have to focus on migrating to media foundation (or rather bring support for the 2 media engines). So a lot of work is needed on our side too
Without forgetting support for new ffmpeg codecs (wma pro, ...)


About DXVA, I don't understand the posts before : we have the same implementation as MPC HC : DXVA 1 & 2 are supported
I don't understand why ffdshow would be slower, unless you enable postprocessing filters
Also, DXVA2 brings some postprocessing features (DXVA HD) that may be interesting.

But once again, I don't like DXVA : the picture is blurry compared to software decoding, and with multithreaded decoding AND postprocessing, you have the best results even if you don't have a (very) recent CPU on HD content

@adam777 : this has to be my fault, I replaced the SRT parser by the SSA parser which brings of course more features but there was a bug that seems still to be not fixed

[clsid]

Restructuring the code of ffdshow is something I would fully support. It should become more modular/layered with nice interfaces between parts. The code that interfaces with the various external libraries could also use some good cleaning/rewrite.

[SamuriHL]

And allow it all to build in VS 2010? :D

[adam777]

@adam777 : this has to be my fault, I replaced the SRT parser by the SSA parser which brings of course more features but there was a bug that seems still to be not fixed

Ah... OK. I'll keep an eye to see when things gets fixed. :thanks:

[tal.aloni]

But once again, I don't like DXVA : the picture is blurry compared to software decoding

an internal post processing beta that I made allowed me to grab frames from the DXVA GPU Buffer, and compare them to software decoding. my Radeon 4550 output was identical to H.264 software decoding. (results may vary based on the GPU obviously)

p.s.
I agree that it's harder to tame DXVA: worse postprocessing, and it doesn't work well with overlay filter. (overlay filter is the best way for me to avoid tearing)

[Sebastiii]

an internal post processing beta that I made allowed me to grab frames from the DXVA GPU Buffer, and compare them to software decoding. my Radeon 4550 output was identical to H.264 software decoding. (results may vary based on the GPU obviously)

p.s.
I agree that it's harder to tame DXVA: worse postprocessing, and it doesn't work well with overlay filter. (overlay filter is the best way for me to avoid tearing)

It's cool and good idea :)

[tetsuo55]

an internal post processing beta that I made allowed me to grab frames from the DXVA GPU Buffer, and compare them to software decoding. my Radeon 4550 output was identical to H.264 software decoding. (results may vary based on the GPU obviously)

p.s.
I agree that it's harder to tame DXVA: worse postprocessing, and it doesn't work well with overlay filter. (overlay filter is the best way for me to avoid tearing)maybe you could come on IRC and we could help you fix those tearing issues.

[_xxl]

Tested with ffdshow 3368, mkv 720p h.264 sample.
ffmpeg-mt compiled with -mmmx
User: 0s, kernel: 0s, total: 0s, real: 11s, fps: 2228.0, dfps: 149.2
ffmpeg-mt compiled with -mmmx -msse -mfpmath=sse
User: 0s, kernel: 0s, total: 0s, real: 11s, fps: 1782.4, dfps: 149.6
ffmpeg-mt compiled with -mmmx -msse -mfpmath=sse -msse2 -msse3 -mssse3
User: 0s, kernel: 0s, total: 0s, real: 11s, fps: 2228.0, dfps: 149.4

[tetsuo55]

So it doesn't hurt, nice.

[tetsuo55]

I did some more research, and it seems that every benchmark ever posted here seems to prove that the higher the sse version used, the higher the performance.

I also found 2 benchmarks that indicate msvc is actually faster than gcc/icc (all posted here on doom9)

http://forum.doom9.org/showthread.php?p=1363969#post1363969
http://forum.doom9.org/showthread.php?p=1335779&highlight=dfps#post1335779
http://forum.doom9.org/showthread.php?p=973342&highlight=dfps#post973342
http://forum.doom9.org/showthread.php?p=943149#post943149
http://forum.doom9.org/showthread.php?p=1217662#post1217662
http://forum.doom9.org/showthread.php?p=998532&highlight=dfps+msvc#post998532

Quick conclusion, it doesnt matter what you use to compile ffmpeg, the best way to get a tiny bump in performance to is to build it with all the optimisation switches your cpu supports.

[clsid]

Those benchmarks were for MSVC/GCC builds of ffdshow.ax! For that we obviously prefer MSVC. What matters for decoding performance is libavcodec.dll and for that GCC is the best by a huge margin.

The numbers from _xxl's benchmark show that it is of little benefit to enable additional compiler optimizations for libavcodec. The difference is less than 1% and within the margin of error. Compatibility with older systems is much more important than an insignificant performance difference.

[tetsuo55]

Those benchmarks all tests libavcodecs performance by running a h264 video with the internal decoder, they clearly specify when the whole project was built with gcc, icc or msvc. [i have been brainwashed to believe gcc/icc are always faster than msvc so i have a hard time believing these facts myself]

its true that the improvements fall within the margin of error scale, but over all the benchmarks posted its consistently faster by 0.1% to 2%.
I agree that the default stable builds should support the lowest os/cpu available. In the case of MPC-HC that means "SSE"

[nm]

Those benchmarks all tests libavcodecs performance by running a h264 video with the internal decoder, they clearly specify when the whole project was built with gcc, icc or msvc. [i have been brainwashed to believe gcc/icc are always faster than msvc so i have a hard time believing these facts myself]

its true that the improvements fall within the margin of error scale, but over all the benchmarks posted its consistently faster by 0.1% to 2%.
Hmm. Looking at your links, I only see one old benchmark (http://forum.doom9.org/showthread.php?p=998532#post998532) where MSVC is faster and two (this (http://forum.doom9.org/showthread.php?p=1363969#post1363969) and that (http://forum.doom9.org/showthread.php?p=1217662#post1217662)) where it's slower?

[fastplayer]

The numbers from _xxl's benchmark show that it is of little benefit to enable additional compiler optimizations for libavcodec.
In this particular case, ffmpeg-mt is used for H.264 decoding which itself avoids floating point ops like the plague. That makes "-mfpmath=sse" completely pointless.

[tetsuo55]

Hmm. Looking at your links, I only see one old benchmark (http://forum.doom9.org/showthread.php?p=998532#post998532) where MSVC is faster and two (this (http://forum.doom9.org/showthread.php?p=1363969#post1363969) and that (http://forum.doom9.org/showthread.php?p=1217662#post1217662)) where it's slower?
The two you linked to seem to use ICL for compiling it seems, so its not a true comparison to clean msvc.

Also to give msvc a fighting chance against gcc it will have to be compiled with: /O2 /GL /GS- /GA /arch:SSE2 /NXCOMPAT /DYNAMICBASE

(replace arch with whatver the lowest sse version is you want to support, equal to gcc)
I don't recommend /GS- for release builds, use /GS instead.

[nm]

The two you linked to seem to use ICL for compiling it seems, so its not a true comparison to clean msvc.
Ok, but could you point out those results where MSVC is faster?

[tetsuo55]

Ok, but could you point out those results where MSVC is faster?Only in the old benchmark you mentioned.

These older benchmarks that are missing a lot of information all seem to prove that no matter what compiler/settings you use the end result will be slower or faster within a margin of -2% to +2%.

We need good benchmarks, where all the variables are named.
*Exact compiler versions
*Switches used/disabled
*builds provided for verification on other systems
*tested on both amd/intel
*Each test run 5 times, with the average used as the final result
*etc...

Also nobody tested PGO which can result in anything from 0 to 40% more performance.

It should not matter at all for most of ffmpeg though as its mostly hand optimised with assembly.

[clsid]

For compiling libavcodec MSVC has no fighting chance whatsoever, no matter which settings you use. That is a pure and simple fact. I will repeat the reason again: all optimized code is excluded from the build. Compiler optimized C code simply can't win from the hand-written assembly code. It can't even come close.

And instead of spending lots of time to squeeze of % performance gain out of the other parts of the code, I think the effort should instead be spend on fixing bugs. More and more people are using DXVA. Also since computing power is doubling roughly every 18 months, performance is becoming a non-issue.

The only other CPU cycle hungry part in MPC is the subtitle renderer. That should be optimized by rewriting the code. For example to avoid animation overhead for static subtitles.

[tetsuo55]

For compiling libavcodec MSVC has no fighting chance whatsoever, no matter which settings you use. That is a pure and simple fact. I will repeat the reason again: all optimized code is excluded from the build. Compiler optimized C code simply can't win from the hand-written assembly code. It can't even come close.the same is true for gcc no?

[nm]

the same is true for gcc no?

No, the inline assembly in FFmpeg/libavcodec is written for GCC and compilers that support that particular syntax. MSVC doesn't.

[tetsuo55]

if thats true then i can understand the difference, however we include a assembly compiler ourselves, why do we do that then?

We have done some quick tests, hope to get a clear scientific test out tonight.
Results so far (libmpeg2)

MSVC 2008 SP1, dfps: 294.3
GCC 4.5.0, dfps: 209.5

Very strange results, we are investigating to make sure it isn't a mistake in our test environment. (both tests where run 8 times)
We are going to optimise the gcc switches further, but the current settings should already beat msvc... (and msvc uses sub-optimal switches which we intend to optimise too)

[_xxl]

Hmm. Looking at your links, I only see one old benchmark (http://forum.doom9.org/showthread.php?p=998532#post998532) where MSVC is faster and two (this (http://forum.doom9.org/showthread.php?p=1363969#post1363969) and that (http://forum.doom9.org/showthread.php?p=1217662#post1217662)) where it's slower?
That was back in 2007 when I compiled ffdshow.ax using GCC, now is not possible. Please stop this nonsense, ffdshow and all except ffmpeg(mt), mplayer, x264 should be compiled with msvc and the rest by GCC. What is so hard to understand? Forcing SSE will require SSE capable CPU's so older CPU's will fail. For speed's less then 10% is pointless to force SSE when compiling.
FFMPEG, MPLAYER and X264 will never compile with MSVC.
Ask FFMPEG, MPLAYER and X264 developers and you get your answer.
libmpeg2 has custom code that isn't in original version from mplayer to support MSVC and win64.
Original version should only be compiled by GCC, but ffdshow version only by MSVC.
ffdshow has a lot of custom code inside most libs.

[dann23]

Forcing SSE will require SSE capable CPU's so older CPU's will fail. For speed's less then 10% is pointless to force SSE when compiling.

less than 10% could make the difference for the cpu to stay in P3 state not P0 and even Pentium III has SSE. It just stupid to maintain support for such old processors.

[clsid]

The difference is <1% for libavcodec so we don't need to engage in discussion whether SSE is useful or not. All the performance critical parts are already hand-optimized to use MMXext/SSE/SSE2/SSSE3/etc whenever possible.

Libmpeg2 is compiled with MSVC by default in ffdshow because that is the fastest. In case of libmpeg2 the code contains inline assembly with a different syntax, one that MSVC does understand. In this case GCC has less optimized code to work with, hence the worse performance.

[XhmikosR]

I managed to build libmpeg2 with ICL 11 (O3, nothing else changed). And it is quite faster than MSVC (nothing changed in project settings) on my machine. The video I tested was a 1280x720 mpg @59.940fps @ 16Mbps, remuxed to avi with mplayer with the following cmd:

mencoder input.mpg -o output.avi -ovc copy -oac copy -of avi -ffourcc mpg2

MSVC 2008 SP1
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 291.7, dfps: 291.3
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 292.9, dfps: 292.3
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 291.8, dfps: 291.3
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 292.9, dfps: 292.3
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 290.1, dfps: 289.7
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 292.1, dfps: 291.6

ICL 11
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 315.4, dfps: 314.9
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 316.2, dfps: 315.4
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 315.8, dfps: 315.2
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 315.1, dfps: 314.7
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 315.8, dfps: 315.3
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 315.1, dfps: 314.2

Here (http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca33cdc61d280511ab) are the compiled dlls for anyone that may be interested.

[albain]

Beware of benchmarks results when you compare the benefits of SSE with each other :

1/ we have very few intrinsics methods defined for SSE3 and SSE4. We could add more
2/ even though we don't use them. Very few classes are templatized with more than MMX level. A few have both MMX and SSE2, and I have not seen anywhere SSE3 and 4

Simply because this is the double amount of work when you want to templatize a new series of instructions to do the same task

@Tal, I am surprised about what you said on DXVA vs software decoding results. I have a radeon 5750 and I find the picture blurrier in DXVA mode.
The only thing that could explain (apart maybe gpu drivers settings) is the fact that I use RGB32 mode against NV12 in DXVA mode

@All : I refounded the subtitles section which became a mess, please tell what you think me about it. Also I reverted the modification that affected Hebrew subtitles.

[clsid]

The subtitle format options are now suboptions of the Embedded option. What about non-embedded subtitle files? It may be useful to have an option for disabling that as well.
Maybe have seven checkboxes:
1) Embedded 2) External
3) Bluray (enabled if embedded is checked)
4) VobSub (enabled if embedded OR external is checked)
5) SSA (enabled if embedded OR external is checked)
6) Closed Captions (enabled if embedded is checked)
5) Text (enabled if embedded OR external is checked) (currently not yet present)

I would also suggest renaming "Text subtitles with SSA tags support". People now may think that it is another kind of subtitles format and check it. Something like "Process SSA tags in text subtitles" would be better, but maybe anyone else has can come up with a better description for it.

[ikarad]

@All : I refounded the subtitles section which became a mess, please tell what you think me about it. Also I reverted the modification that affected Hebrew subtitles.

@Albain
1) Have you see this problem with blu-ray subs and xp?
@Albain,
there is a bug with blu-ray sub support with xp sp3 (under vista it works very well).

The subs are divide in two parts and are not completed
http://nsa14.casimages.com/img/2010/04/20/mini_100420050119368677.jpg (http://www.casimages.com/img.php?i=100420050119368677.jpg)


I try with YUY2 or RGB 32 and there is the same problem.
I try with all versions of ffdshow (3350 to 3368) where it works under vista and there is the same problem udner xp sp3.

I notice some crashs but I think that crashs come from this problem .

same example that I give you before
example
part 1
http://www.zshare.net/info.html?7219...12ca4e70f4959e
part2
http://www.zshare.net/info.html?7222...cb0c45e37239c1
part3
http://www.zshare.net/info.html?7222...fa215e89fcfdf2

part 4
http://www.zshare.net/info.html?7287...1fac7d6620cda5

part 5
http://www.zshare.net/info.html?7289...835142beff854c
part 6
http://www.zshare.net/info.html?7290...3fe6974fb1237c

partie 7
http://www.zshare.net/info.html?7215...a94c0fc61987f5

2) Bug with 3371 :I try with 3371 and blu-ray subs are not displayed under xp and vista (rgb32 and yuy2)


3) I try the new versio nand maybe it would be interesting to explain each functions in subtitle section because there are many options and I don't know if I must use or not.

For example I don't know what means these options (in black circle)
http://nsa15.casimages.com/img/2010/04/23/mini_100423080302246904.jpg (http://www.casimages.com/img.php?i=100423080302246904.jpg)
http://nsa15.casimages.com/img/2010/04/23/mini_100423080409854437.jpg (http://www.casimages.com/img.php?i=100423080409854437.jpg)
http://nsa15.casimages.com/img/2010/04/23/mini_100423080725534897.jpg (http://www.casimages.com/img.php?i=100423080725534897.jpg)
http://nsa14.casimages.com/img/2010/04/23/mini_100423080748835235.jpg (http://www.casimages.com/img.php?i=100423080748835235.jpg)

[STaRGaZeR]

I like the new options, and I like clsid suggestions too. Also I'd rename BluRay to Blu-ray. Nitpick, yes, but that's the format's correct name :p

@albain, I'm trying to fix an "issue" with SSA/ASS subtitles, but I don't have enough knowledge to do it properly. Currently ffdshow searches for "PlayResX" and "PlayResY", and if one is missing it'll use the video resolution to calculate the subtitles' coordinates. However there are many old scripts that only have "PlayResY", and MPC/VSFilter work with that only. Obviously people create the scripts with MPC/VSFilter in mind, so ffdshow puts the subs in "incorrect" places. I've determined (by testing, so I could be totally wrong) that if Y is set but X is not, MPC/VSFilter will assume a 4/3 aspect ratio and use Y=Y and X=Y*4/3. If Y is missing but X is present it's the opposite. I've made a patch that corrects this partially, but this is what I can't fix: with the patch, if X is set and Y is not, ffdshow now puts the subs where they belong, but at the incorrect font size. Can you help me with this?

Patch: http://www.mediafire.com/?ymgznzjzeu3

[albain]

@Ikarad : sorry I can't help you, I don't have XP.
Otherwise :
- Subtitles section : checkboxes are here to enable or not the support of the different subtitle formats listed here
- Font section : only applicable to text subtitles (srt/ssa)
- Bluray/DVD subtitles positioning : if no coordinates are set then use the horizontal and vertical placement in the placement section

@clsid : for now it is not possible to disable distinct external subtitle formats, so if the subtitles filter is enabled, then all the external subtitle formats which are supported are enabled
As you said : one option per subtitles format, and 2 options : embedded and external is the right way to go.

Otherwise this is right that vobsub (DVD subs) are supported in both embedded and external, as well as srt/ssa.
But not bluray (not yet), only embedded for now.

[ikarad]

@Ikarad : sorry I can't help you, I don't have XP.


:( never fix for xp I think. I hope one day anybody will fix this problem.

But I notice this problem with 3371, blu-ray subs are not displayed under vista.

[adam777]

@All : I refounded the subtitles section which became a mess, please tell what you think me about it. Also I reverted the modification that affected Hebrew subtitles.

Doesn't seem to work.
Hebrew subtitles still displayed wrong, with "Text subtitles with SSA tags support" on and off.
Nothing is selected - No bluray, DVD etc., simply trying to play simple .srt subtitles.
Rev. 3367 still the last to work under x64.

[rica]

What is very interesting to me, expectations of everybody are too much for a coder.
Any open source coder must have three PCs?

1) XP 32 on one disk, XP 64 on another disk.
2) Vista 32 on one disk, Vista 64 on another disk.
3) 7 32 on one disk, 7 64 on another disk.

And how?
6 OS, 6 HDDs and 3 PCs? (I never stated the sub-brunches btw.)
Any donations?

[lych_necross]

I think its time to consider phasing out support for XP (focus should be placed on newer versions of Windows). Users of XP should really consider upgrading to 7.

[Shankster]

I think its time to consider phasing out support for XP (focus should be placed on newer versions of Windows). Users of XP should really consider upgrading to 7.

And before XP users go apeshit, ponder this:
You are now angry because there is no attention towards xp, cause all focus is on vista/7.

Now flip the table and imagine how it would feel if you couldnt get problems on XP fixed because of Win98 users.

At some point you are gonna have to let go of that 10 year old OS.

[ikarad]

I think its time to consider phasing out support for XP (focus should be placed on newer versions of Windows). Users of XP should really consider upgrading to 7.
Why don't support old os ?
And I don't speak about win 95 or 98 but xp which is still a very used os.

Xp isn't win 95 or 98 and xp is more use than vista or seven.

Xp has beeen replaced by vista only since two years and for most users xp stay the main os. I don't think that we can consider xp like an old os.

After I understand that albain can't fix this problem because He doens't have xp

[lych_necross]

Why don't support old os ?
And I don't speak about win 95 or 98 but xp which is still a very used os.
At some point, a developer needs to stop supporting old OSes. If a program functions correctly under Win7 but not XP, it could be said that the problem is with XP. It takes a lot of time to develop workarounds for old OSes. Instead of spending time writing workarounds, I would prefer that same time be spend on developing new features. Besides, even Microsoft has discontinued support (http://www.microsoft.com/windows/support/endofsupport.mspx) for all versions of Windows up to XP SP1 (XP SP2's support will be discontinued on July 13th, 2010 (http://windowsteamblog.com/blogs/windowsvista/archive/2010/02/24/end-of-support-for-windows-2000-windows-xp-sp2-and-windows-vista-rtm.aspx)).

[sacharja]

They only care about working playback, which is what the pack provides.
Actually, no. #3370 is unusable in its current state. As several people mentioned now, it's crashing all the time while seeking fast (with a player that is capable of fast seeking, like ZoomPlayer). I'm a little bit afraid that such an essential thing is still not fixed after over a month.

[dann23]

At some point, a developer needs to stop supporting old OSes. If a program functions correctly under Win7 but not XP, it could be said that the problem is with XP. It takes a lot of time to develop workarounds for old OSes. Instead of spending time writing workarounds, I would prefer that same time be spend on developing new features. Besides, even Microsoft has discontinued support (http://www.microsoft.com/windows/support/endofsupport.mspx) for all versions of Windows up to XP SP1 (XP SP2's support will be discontinued on July 13th, 2010 (http://windowsteamblog.com/blogs/windowsvista/archive/2010/02/24/end-of-support-for-windows-2000-windows-xp-sp2-and-windows-vista-rtm.aspx)).

If so then drop support for old processor. :D

[madshi]

At some point, a developer needs to stop supporting old OSes.
"At some point"? True. But that description fits anything between now and eternity.

Look, there are probably many people out there who have a stable and well working XP HTPC. Why forcing them to upgrade to win7, if XP works great for them? They would have to buy a win7 upgrade, reinstall the OS and all the DirectShow filters and everything, search for a reasonably well working GPU driver, etc etc. They might even have to upgrade hardware. And after going through all that trouble - what would they gain? Nothing at all, if XP does all that they need. They would even risk replacing their stable and well working setup with a new setup with new problems. So IMHO it's *MUCH* too early to drop XP support. And I don't see the need to drop XP support, anyway.

Win9x was a whole different world. It wasn't even remotely similar to XP if you looked behind the facade. XP and win7 are very very similar inside.

[tetsuo55]

I don't see why ffdshow needs to drop/add support for anything.
Since its just a transform filter and post processing filter there is little benefit in supporting a specific platform.

Over at MPC-HC the story is different, as Windows7 offers a LOT of features that benefit playback.

[ikarad]

At some point, a developer needs to stop supporting old OSes. If a program functions correctly under Win7 but not XP, it could be said that the problem is with XP. It takes a lot of time to develop workarounds for old OSes. Instead of spending time writing workarounds, I would prefer that same time be spend on developing new features. Besides, even Microsoft has discontinued support (http://www.microsoft.com/windows/support/endofsupport.mspx) for all versions of Windows up to XP SP1 (XP SP2's support will be discontinued on July 13th, 2010 (http://windowsteamblog.com/blogs/windowsvista/archive/2010/02/24/end-of-support-for-windows-2000-windows-xp-sp2-and-windows-vista-rtm.aspx)).

sp3 is still supported for one year or more.
Why don't support an old os when It's the main os for the majority of users?

I prefer a stable version of ffdshow for each OS used that add features never completly fucntionnal.


As I said above, XP isn't an useless OS like win 98. It's the main OS for many users.

[albain]

We don't intend to drop xp support as its architecture is not that far from newer OS and is still spread a lot

I just hope that someone on xp not necessarily with developer skills but able to build ffdshow in debug mode will be able to narrow down that problem, at least just enough to have a lead

About Hebrew subtitles, I will look after that : do you have an idea why the subtitles are garbled ?
I wish we could drop the srt engine and use the substation alpha one which brings a lot of features

@Stargazer : I'll look into your patch and let you know

[Px]

What is very interesting to me, expectations of everybody are too much for a coder.
Any open source coder must have three PCs?

1) XP 32 on one disk, XP 64 on another disk.
2) Vista 32 on one disk, Vista 64 on another disk.
3) 7 32 on one disk, 7 64 on another disk.

And how?
6 OS, 6 HDDs and 3 PCs? (I never stated the sub-brunches btw.)
Any donations?
There is no need for 6 HDD and 3 PCs, in time when I was actively testing/benchmarking different things, I have 5 OS-es installed on signle 200GB HDD:
C - 30 GB, main XP x64
D - 10 GB, XP x32 for game benchmarking
E - 10 GB, XP x32 for video editing
F - 10 GB, XP x32 for other experiments
G - 10 GB, W2K Prof for video capturing (drivers for XP for my tuner was less stable, then for W2K)
H - 130 GB, main storage for games, documents and other stuff
Later I bought 2nd HDD, 320 GB, and install there Vista.
So I don't see any problems to install as many OS-es as needed on single PC even with 1 HDD

[albain]

The problem is not technical but the time to spend on this kind of problem.

Personally I don't have the time to work on this and I am not motivated as well to play the DJ with OS swapping.
I guess that clsid and Tal are in the same position.

As I said, we are very few coders, and help from contributors in this forum would be highly appreciated, as it has been the case to put in place HD audio bitstream, DXVA,...

Otherwise I have brought the changes to subtitles section with the new checkboxes:
http://img17.imageshack.us/img17/3060/ffdshowsubs.png (http://img17.imageshack.us/i/ffdshowsubs.png/)

My next moves : bluray subtitles files support, fix the subtitle scalers that don't work in RGB32.
Stargazer last patch too

[Sebastiii]

Always :) thx you :)
Seb.

[clsid]

An easy way to test different OSes is to run them in a virtual machine. For example VirtualBox.

[Midzuki]

About Hebrew subtitles, I will look after that : do you have an idea why the subtitles are garbled ?
I wish we could drop the srt engine and use the substation alpha one which brings a lot of features

Hebrew and Arabic text lines are written and read "RTL". For some reason, ffdshow is "RTL"ing word-by-word, and not line-by-line. BTW, MaestroSBT has the same (unsolved) issue.

[STaRGaZeR]

@Stargazer : I'll look into your patch and let you know

This is what I have so far. VSFilter and ffdshow at the same time.

1. Without the patch, PlayResX=unset, PlayResY=768. ffdshow is at the bottom right corner.

http://thumbnails10.imagebam.com/7785/e8b65e77843528.gif (http://www.imagebam.com/image/e8b65e77843528)

2. With the patch, PlayResX=unset, PlayResY=768. Good.

http://thumbnails3.imagebam.com/7785/92168977843529.gif (http://www.imagebam.com/image/92168977843529)

3. With the patch, PlayResX=1024, PlayResY=unset. Position matches VSFilter, but the font size is wrong this time. This is what I can't figure out.

http://thumbnails24.imagebam.com/7785/cc0e5477843532.gif (http://www.imagebam.com/image/cc0e5477843532)

4. With the patch, PlayResX=1024, PlayResY=768. This one, being identical to 2., confirms that VSFilter, and ffdshow with the patch, assume the 4/3 ratio if one of them is missing.

http://thumbnails24.imagebam.com/7785/00b04f77843535.gif (http://www.imagebam.com/image/00b04f77843535)

[amango]

In FFDShow I can force bob deinterlacing to the output. This is also a good way to double the framerate. This is quite handy if you use Mediaportal because the speed of the gui is bound to the frame rate of the video.

There is however no option for this in the DXVA decoder, could someone add this? :)

[adam777]

Well, the bad news is Hebrew subtitles are still no go in rev. 3373, however, I think I realized something that might prove useful.
It seems that when there are 2 lines under the same subtitle index, that later (bottom) line is displayed to the right, while the earlier (top) line is displayed to the left.
It should be the other way around.

That is, if an index exist:
14
00:02:12,741 --> 00:02:14,657
3 2 1
6 5 4
it's displayed as (remember Hebrew is read right to left) - 3 2 1 6 5 4, while it should be displayed as 6 5 4 3 2 1

[tal.aloni]

In FFDShow I can force bob deinterlacing to the output. This is also a good way to double the framerate. This is quite handy if you use Mediaportal because the speed of the gui is bound to the frame rate of the video.

There is however no option for this in the DXVA decoder, could someone add this? :)

Not technically possible AFAIK.

[tal.aloni]

An easy way to test different OSes is to run them in a virtual machine. For example VirtualBox.

actually, if you want to test DXVA and such, than iSCSI is your best friend. I'm booting xp or win7 using gPXE.

[Keiyakusha]

In FFDShow I can force bob deinterlacing to the output. This is also a good way to double the framerate.
If you mean forcing bob on progressive footage, than I think you loosing half of the vertical resolution by doing that. Copying every frame once should be better and probably not slower. However no way to do that with DXVA...
And for interlaced content I think there should be some way to do bob with GPU, however don't know if it compatible with DXVA and I don't have interlaced content at all, so not sure.

[tal.aloni]

@Tal, I am surprised about what you said on DXVA vs software decoding results. I have a radeon 5750 and I find the picture blurrier in DXVA mode.
The only thing that could explain (apart maybe gpu drivers settings) is the fact that I use RGB32 mode against NV12 in DXVA mode

exactly,
by default, Radeon outputs YCrCb 4:4:4 over HDMI, which for 16-235 NV12 content (including DXVA), might result in wrong levels (depends on monitor settings).

when selecting RGB 4:4:4 as HDMI output colorspace, you can either let the radeon convert NV12 to RGB internally (with correct levels), or let ffdshow handle the colorspace conversion, which will result in higher cpu usage, and according to some, better quality.

in any case, the PQ benefits of not using DXVA are only the benefits of using ffdshow's (or other renderer) colorspace conversion, at least for the radeon.

Tal

[tal.aloni]

Well, the bad news is Hebrew subtitles are still no go in rev. 3373

Shalom Adam, I've just tested some hebrew srt subs with 3373,
EDIT: I see it now, when two lines are involved, the first line comes second.

Tal

[STaRGaZeR]

Found another "issue": all the ASS scripts I have are always blurred with ffdshow, while they're not with VSFilter. I've found these lines in TsubtitleText.cpp:

if (fontSettings.blur || (w->props.version >= TsubtitleParserSSA::ASS && lf.lfHeight > int(37 * gdi_font_scale))) // FIXME: messy. just trying to resemble vsfilter.
prefs.blur=true;

That if is always true here. I don't know what the author was trying to resemble in VSFilter, but I'd just enable blur if the setting is enabled in the Font section, like this:

if (fontSettings.blur)
prefs.blur=true;

BTW, what is "gdi_font_scale"? I don't understand this variable and it's used everywhere.

[madshi]

in any case, the PQ benefits of not using DXVA are only the benefits of using ffdshow's (or other renderer) colorspace conversion, at least for the radeon.
Don't forget chroma upsampling.

[XhmikosR]

@albain: I just tried r3374 and it crashes every time I try to use the subtitles. The video is a 1920x1080p WVC1 and the subtitles is a BD .sup file. If you cannot reproduce the crash let me know. I'll try to upload a sample.

Also, someone should update the ICL project files.

[fastplayer]

SRT subs don't load anymore. Even though there's a checkbox showing that the file is loaded, nothing is displayed. Checking that checkbox (in the subtitle context menu) actually enables subtitles.
SRT subtitles with 2 lines are displayed in one line. "Split long subtitle lines" has no effect.

[albain]

Found another "issue": all the ASS scripts I have are always blurred with ffdshow, while they're not with VSFilter. I've found these lines in TsubtitleText.cpp:

if (fontSettings.blur || (w->props.version >= TsubtitleParserSSA::ASS && lf.lfHeight > int(37 * gdi_font_scale))) // FIXME: messy. just trying to resemble vsfilter.
prefs.blur=true;

That if is always true here. I don't know what the author was trying to resemble in VSFilter, but I'd just enable blur if the setting is enabled in the Font section, like this:

if (fontSettings.blur)
prefs.blur=true;

BTW, what is "gdi_font_scale"? I don't understand this variable and it's used everywhere.

Indeed, this test should be removed. I guess this is Haruhiko's code.

Haruhiko tried a while ago to increase the resolution of the glyphs but that resulted into choppy playback. I think it would be good to digg into this to get better subtitles

I committed bluray subtitles files support by the way. I have not tested it a lot but it seems to work so far.
Unlike the other subtitles formats that load all the content of the file in memory (even vobsubs), I chose to have only 10 seconds ahead to reduce memory usage. As a result the subtitles file remains opened and read all the time but I don't think this is a problem

[XhmikosR]

OK, here (http://www.mediafire.com/?wikozzzxknz) is a sample and here (http://www.mediafire.com/?ojhvmnvtmqm) are the BD subtitles which cause the crash with r3374.

[Casshern]

DTS Spdif passthrough problems with m2ts files:

1) The DTS core does get picked up when playing the file from the beginning
2) I had MPC-HC set to remember last position, so it essentially seeked before playing -> this results in the core not being picked up
3) When seeking sound gets totally screwed, only multiple reseeking helps to get it back

Here is a small sample (which makes seeking hard, but try to play it NOT from the beginning to replicate)

http://www.mediafire.com/file/mztm2kmzu4u/testdtscore.m2ts

A workaround for this issue is to disable the "connect as pcm first" option under the output options. This may not work for all soundcards, but on my gear there seem to be no side effects as of yet. I hope this helps for others with this problem.

[fastplayer]

SRT subs don't load anymore. Even though there's a checkbox showing that the file is loaded, nothing is displayed. Checking that checkbox (in the subtitle context menu) actually enables subtitles.
SRT subtitles with 2 lines are displayed in one line. "Split long subtitle lines" has no effect.

Forgot to mention that hitting the "Reload" button in the subtitle filter options caused MPC to crash. This is fixed now in rev 3376. Subs still don't load automatically though.

[FreeFall]

System Setup

Windows 7 Professional x64
Zoomplayer 7 Home Max 7.10 Alpha 3
Gabest Mpeg Splitter - standalone filter svn 1809 x32
Haali splitter - 27-03-2010
FFDShow rev 3374 x32
EVR renderer output


Blu-ray Disc Sword Of The Stranger

Subtitles don't display with rev 3371 - 74, working with 3370. Using FFDShow decoder subs colour is orange, with the Microsoft decoder subs are the correct colour (blue).

Sample http://www.mediafire.com/?txzygwnnxmt Subs http://www.mediafire.com/?lkwttbz4ndd


DVD subtitles inside mkv container don't display after working at first. Sample Drunken Master, subs work up until the end of the first fight scene and won't display in the dojo scene until you turn them off and on again.

Sample http://www.mediafire.com/?l223ejwjywu Subs http://www.mediafire.com/?dhthkm0jmx1


Thanks
FreeFall

[hoborg]

@albain:

-Fixed audio/subtitles streams switching from keyboard
:thanks:

[STaRGaZeR]

Albain, this is not the first time I've seen a post like this (http://forum.doom9.org/showthread.php?p=1394100#post1394100), does ffdshow really need a limitation in the number of format changes? Besides, the 3 second trick you added in r3335 doesn't work here. Commenting

if (audioParserData.nbFormatChanges>=4 || (m_tStart / 10000 / 1000 > 3))
return codecId;

out fixes all my issues, and I bet it will fix most problems with audio switching for other users too.

[XhmikosR]

More compiler comparison. This time I compared the speed difference between MSVC, ICL and GCC for xvidcore.

XviD, 25fps, 1778 Kbps, 704x400 (25000 frames)

ICL 11.1.065 /O3
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 760.6, dfps: 759.5
User: 33s, kernel: 0s, total: 33s, real: 33s, fps: 753.8, dfps: 751.3
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 762.0, dfps: 760.6
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 764.2, dfps: 762.4
User: 32s, kernel: 0s, total: 32s, real: 33s, fps: 757.7, dfps: 756.6
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 760.9, dfps: 759.9


MSVC 2008 SP1
User: 50s, kernel: 0s, total: 50s, real: 50s, fps: 496.5, dfps: 495.7
User: 49s, kernel: 0s, total: 50s, real: 50s, fps: 499.5, dfps: 499.1
User: 50s, kernel: 0s, total: 50s, real: 50s, fps: 497.5, dfps: 496.9
User: 50s, kernel: 0s, total: 50s, real: 50s, fps: 498.5, dfps: 497.5
User: 49s, kernel: 0s, total: 50s, real: 50s, fps: 498.8, dfps: 497.2
User: 49s, kernel: 0s, total: 50s, real: 50s, fps: 498.9, dfps: 498.0


GCC 4.5.0
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 763.1, dfps: 761.7
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 760.2, dfps: 758.8
User: 32s, kernel: 0s, total: 32s, real: 33s, fps: 758.8, dfps: 757.3
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 761.7, dfps: 761.0
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 762.4, dfps: 761.3
User: 32s, kernel: 0s, total: 32s, real: 32s, fps: 760.9, dfps: 760.2
=====================================================================


=====================================================================
DivX, 29.97fps, 3917 Kbps, 1280x720 (10000 frames)

ICL 11.1.065 /O3
User: 51s, kernel: 0s, total: 51s, real: 51s, fps: 195.7, dfps: 195.3
User: 50s, kernel: 0s, total: 51s, real: 51s, fps: 196.0, dfps: 195.9
User: 50s, kernel: 0s, total: 51s, real: 51s, fps: 195.8, dfps: 195.7
User: 50s, kernel: 0s, total: 51s, real: 51s, fps: 196.0, dfps: 195.7
User: 50s, kernel: 0s, total: 51s, real: 51s, fps: 195.9, dfps: 195.6
User: 50s, kernel: 0s, total: 51s, real: 51s, fps: 195.7, dfps: 195.4


MSVC 2008 SP1
User: 68s, kernel: 0s, total: 68s, real: 68s, fps: 145.9, dfps: 145.7
User: 68s, kernel: 0s, total: 68s, real: 68s, fps: 146.0, dfps: 145.7
User: 69s, kernel: 0s, total: 69s, real: 69s, fps: 143.4, dfps: 143.3
User: 68s, kernel: 0s, total: 68s, real: 69s, fps: 145.0, dfps: 144.6
User: 68s, kernel: 0s, total: 68s, real: 68s, fps: 146.2, dfps: 145.7
User: 68s, kernel: 0s, total: 68s, real: 68s, fps: 145.6, dfps: 145.1


GCC 4.5.0
User: 50s, kernel: 0s, total: 51s, real: 51s, fps: 195.6, dfps: 195.5
User: 51s, kernel: 0s, total: 51s, real: 51s, fps: 195.4, dfps: 195.3
User: 51s, kernel: 0s, total: 51s, real: 51s, fps: 195.7, dfps: 195.5
User: 51s, kernel: 0s, total: 51s, real: 51s, fps: 195.4, dfps: 194.7
User: 51s, kernel: 0s, total: 51s, real: 51s, fps: 194.8, dfps: 194.7
User: 51s, kernel: 0s, total: 51s, real: 51s, fps: 195.2, dfps: 195.0
=====================================================================


=====================================================================
XviD, 24fps, 16.1 Mbps, 1280x720 (2000 frames) (Big Buck Bunny)

ICL 11.1.065 /O3
User: 15s, kernel: 0s, total: 15s, real: 15s, fps: 131.4, dfps: 130.8
User: 15s, kernel: 0s, total: 15s, real: 15s, fps: 131.2, dfps: 130.8
User: 15s, kernel: 0s, total: 15s, real: 15s, fps: 131.4, dfps: 131.0
User: 15s, kernel: 0s, total: 15s, real: 15s, fps: 131.1, dfps: 131.0
User: 15s, kernel: 0s, total: 15s, real: 15s, fps: 131.6, dfps: 130.8
User: 15s, kernel: 0s, total: 15s, real: 15s, fps: 131.4, dfps: 130.8


MSVC 2008 SP1
User: 22s, kernel: 0s, total: 22s, real: 22s, fps: 87.2, dfps: 87.1
User: 22s, kernel: 0s, total: 22s, real: 23s, fps: 87.2, dfps: 86.9
User: 22s, kernel: 0s, total: 22s, real: 22s, fps: 87.5, dfps: 87.2
User: 22s, kernel: 0s, total: 22s, real: 23s, fps: 87.2, dfps: 86.8
User: 22s, kernel: 0s, total: 22s, real: 22s, fps: 87.4, dfps: 87.2
User: 22s, kernel: 0s, total: 22s, real: 23s, fps: 87.2, dfps: 86.9


GCC 4.5.0
User: 14s, kernel: 0s, total: 14s, real: 14s, fps: 134.2, dfps: 134.1
User: 14s, kernel: 0s, total: 14s, real: 14s, fps: 134.5, dfps: 134.1
User: 14s, kernel: 0s, total: 14s, real: 14s, fps: 134.0, dfps: 133.4
User: 14s, kernel: 0s, total: 14s, real: 14s, fps: 134.4, dfps: 133.8
User: 14s, kernel: 0s, total: 14s, real: 14s, fps: 134.8, dfps: 134.2
User: 14s, kernel: 0s, total: 14s, real: 14s, fps: 134.0, dfps: 133.7

And I wonder, what was the reason for changing xvidcore's compiler from GCC to MSVC? Also it's obvious that ICL is superior to MSVC. The compiled dlls are here (http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335caf38cabe7f356913a) for anyone interested. If you do try to replicate my results, don't use a 320x240 XviD video. Use something bigger with high bitrate.

And another libmpeg2 benchmark


mpeg2 remuxed to avi with mencoder, 59.940fps, 15.7 Mbps, 1280x720

MSVC 2008 SP1
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 291.2, dfps: 290.3
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 292.6, dfps: 292.0
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 290.7, dfps: 290.6
User: 29s, kernel: 0s, total: 29s, real: 30s, fps: 288.2, dfps: 287.4
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 289.8, dfps: 289.2
User: 29s, kernel: 0s, total: 29s, real: 29s, fps: 292.0, dfps: 291.5

ICL 11.1.065 /O3
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 312.9, dfps: 311.9
User: 27s, kernel: 0s, total: 27s, real: 28s, fps: 308.4, dfps: 307.5
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 308.7, dfps: 308.4
User: 27s, kernel: 0s, total: 27s, real: 28s, fps: 308.6, dfps: 307.5
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 310.3, dfps: 309.4
User: 27s, kernel: 0s, total: 27s, real: 27s, fps: 311.3, dfps: 310.0

GCC 4.5.0
User: 40s, kernel: 0s, total: 40s, real: 41s, fps: 210.8, dfps: 210.4
User: 40s, kernel: 0s, total: 40s, real: 40s, fps: 211.6, dfps: 210.8
User: 40s, kernel: 0s, total: 40s, real: 40s, fps: 213.0, dfps: 212.5
User: 40s, kernel: 0s, total: 40s, real: 40s, fps: 212.4, dfps: 211.9
User: 40s, kernel: 0s, total: 40s, real: 40s, fps: 212.7, dfps: 212.2
User: 40s, kernel: 0s, total: 40s, real: 40s, fps: 212.5, dfps: 211.7

The compiled dlls are here (http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca33cdc61d280511ab). GCC is so much slower for libmpeg2 than MSVC, but still ICL is faster.

[tetsuo55]

So these benchmarks seem to point out that all the non-gcc optimised stuff might as well be built with icl.

A simple change that can give up to 40% speed improvement over msvc(on my system the difference varied between 40% and 50% for xvid)

This clearly helps, xxl or clsid could you update the projectfile?
We did some testing and almost all of ffdshow can be build with icl and appears to be faster across the board.
Since the way we did it was a bit hacky we dont have scientific results.

[dimitrik]

But once again, I don't like DXVA : the picture is blurry compared to software decoding, and with multithreaded decoding AND postprocessing, you have the best results even if you don't have a (very) recent CPU on HD content



in any case, the PQ benefits of not using DXVA are only the benefits of using ffdshow's (or other renderer) colorspace conversion, at least for the radeon.
Tal


Pardon my ignorant question, but I don't understand - aren't h.264 decoders supposed to be bit-identical? This PQ argument seems to imply that they are not? :confused:

[dimitrik]

By the way I found something strange in the ffdshow h.264 DXVA subtitles filter.

The letterbox option doesn't seem to work (it works fine with software decoding).

Is it a bug or is it just not possible under DXVA?

[Keiyakusha]

Pardon my ignorant question, but I don't understand - aren't h.264 decoders supposed to be bit-identical? This PQ argument seems to imply that they are not? :confused:

Short answer: after the stream is decoded, there is other things that need_to/may be done, before you will see picture on your screen.
EDIT: also bugs are possible ^_^

[tal.aloni]

The letterbox option doesn't seem to work (it works fine with software decoding).

Is it a bug or is it just not possible under DXVA?

not possible under DXVA.

[clsid]

what was the reason for changing xvidcore's compiler from GCC to MSVC?Who says it has changed? I for example always use GCC for xvidcore.dll

[XhmikosR]

Yeah, right. You maybe, I didn't because xxl insisted in the past that msvc is "better" and added xvidcore in the solution file which overwrote my previous gcc compilation, so I removed xvidcore build with gcc from my scripts.
Anyway, how about the ICL? Don't you think it's time to update to ICL 11? And build everything with ICL.

[Delerue]

Any chance to introduce mjpeg video and twos/sowt audio decoding since they're already supported by FFMPEG? Samples here:

mjpb: http://red.cachefly.net/video/milkgirls1080p.mov
twos: http://www.fileshack.com/file.x/17838/ATI+Radeon+HD+3000+'Ping+Pong'+Tech+Demo+Video

Thanks!

[STaRGaZeR]

Yeah, right. You maybe, I didn't because xxl insisted in the past that msvc is "better" and added xvidcore in the solution file which overwrote my previous gcc compilation, so I removed xvidcore build with gcc from my scripts.
Anyway, how about the ICL? Don't you think it's time to update to ICL 11? And build everything with ICL.

You disable vectorization (/Qvec) in ICL11 because it breaks the Info&CPU tab, right? Does this affect speed in the decoders or image filters?

[tal.aloni]

Hi Guys,
I've tested some performance optimizations that Tetsuo55 suggested for the generic x86 build (only for the main executable for now), the results might be interesting:
User: 5s, kernel: 0s, total: 5s, real: 50s, fps: 881.6, dfps: 97.0 ==>> 3337 - current SVN
User: 5s, kernel: 0s, total: 5s, real: 50s, fps: 909.7, dfps: 97.4 ==>> 3339 - linker optimizations (/NXCOMPAT /DYNAMICBASE)
User: 4s, kernel: 0s, total: 4s, real: 50s, fps: 1105.1, dfps: 97.3 ==>> 3339 - linker optimizations (/NXCOMPAT /DYNAMICBASE) + /O2

can someone please explain what dfps means?


--- ffdshow_2008.vcproj Sun Apr 25 20:22:23 2010
+++ ffdshow_2008.vcproj Mon Apr 26 22:52:40 2010
@@ -285,7 +285,7 @@
/>
<Tool
Name="VCCLCompilerTool"
- AdditionalOptions="/MP"
+ AdditionalOptions="/MP /O2"
Optimization="2"
InlineFunctionExpansion="2"
EnableIntrinsicFunctions="true"
@@ -325,6 +325,7 @@
/>
<Tool
Name="VCLinkerTool"
+ AdditionalOptions="/NXCOMPAT /DYNAMICBASE"
RegisterOutput="false"
IgnoreImportLibrary="true"

[clsid]

Anyway, how about the ICL? Don't you think it's time to update to ICL 11? And build everything with ICL.I will continue to make generic builds and occasionally ICL10.1 builds. I will not upgrade to ICL 11.1 unless anyone can show me it has any real benefit compared to ICL 10.1.

My generic builds use MSVC for ffdshow.ax, GCC for libavcodec.dll ffmpegmt.dll libmplayer.dll xvidcore.dll ff_x264.dll, ICL10 for various other DLLs, and MSVC for the remaining small files.

[clsid]

User: 5s, kernel: 0s, total: 5s, real: 50s, fps: 881.6, dfps: 97.0 ==>> 3337 - current SVN
User: 5s, kernel: 0s, total: 5s, real: 50s, fps: 909.7, dfps: 97.4 ==>> 3339 - linker optimizations (/NXCOMPAT /DYNAMICBASE)
User: 4s, kernel: 0s, total: 4s, real: 50s, fps: 1105.1, dfps: 97.3 ==>> 3339 - linker optimizations (/NXCOMPAT /DYNAMICBASE) + /O2

can someone please explain what dfps means?
How many times did you run these benchmarks? Because the differences are so small that they fall within the margin of error. You can get similar differences with multiple runs with the same build. It also now indicates that /O2 is rather pointless. But since most of the work is done outside of ffdshow.ax any compiler effects on ffdshow.ax are dampened. So to better analyze the effects of the additional switches, more work should be done by ffdshow, for example by enabling some processing filters.

[XhmikosR]

You disable vectorization (/Qvec) in ICL11 because it breaks the Info&CPU tab, right? Does this affect speed in the decoders or image filters?
Yes, I disable Qvec for ffdshow.ax because of that problem. The filters have vectorization enabled and decoders and image filters are not even built with ICL. But from my tests, libmpeg2 is faster when compiled with ICL. And so xvidcore is almost the same as when compiled with gcc when MSVC completely fails. Also I have managed to build a complete ICL build (except from ffmpeg, ffmpeg-mt, libmplayer, x264 and xvidcore). I can post that build after I tweak the settings a little bit.

I will continue to make generic builds and occasionally ICL10.1 builds. I will not upgrade to ICL 11.1 unless anyone can show me it has any real benefit compared to ICL 10.1.

My generic builds use MSVC for ffdshow.ax, GCC for libavcodec.dll ffmpegmt.dll libmplayer.dll xvidcore.dll ff_x264.dll, ICL10 for various other DLLs, and MSVC for the remaining small files.
Well how about the obvious, that you are using a compiler which is no longer available for download? Or is it somewhere hidden?

The ffdshow ICL project files should not be dependent of other project files. I mean, having to compile ffdshow with MSVC and after that ffdshow with ICL is not the best thing. Also, like I showed you above libmpeg2 has a significant speed gain when compiled with ICL. If I could measure other filters I'm pretty sure they would be faster. But that's a speculation only. (for the time being)

You wanted numbers in order to change something. I posted valid numbers from various tests.

[STaRGaZeR]

Yes, I disable Qvec for ffdshow.ax because of that problem. The filters have vectorization enabled and decoders and image filters are not even built with ICL. But from my tests, libmpeg2 is faster when compiled with ICL. And so xvidcore is almost the same as when compiled with gcc when MSVC completely fails. Also I have managed to build a complete ICL build (except from ffmpeg, ffmpeg-mt, libmplayer, x264 and xvidcore). I can post that build after I tweak the settings a little bit.

If you disable vectorization in ffdshow.ax you're disabling it in all the image processing filters, like deinterlacers, etc. That's what I'm asking, if you have done any tests to determine if for example yadif is slower using Qvec. It should be interesting, I've not made the change to ICL11 yet just because of this Qvec thingy, I don't know if ICL10 without Qvec is faster or slower than ICL11 with Qvec in the key parts of ffdshow (yadif, sharpen, etc.)

Just for the record, now that I'm working with the subs I use MSVC builds, and with deband+resize+sharpen+subtitles+RGB32HQ they have measurably higher CPU consumption (up to 15%) with the specs in my sign than my ICL10.1 builds, but I've not benched them. I imagine than in slower computers the difference will be bigger.

I'm interested in that build BTW.

The ffdshow ICL project files should not be dependent of other project files. I mean, having to compile ffdshow with MSVC and after that ffdshow with ICL is not the best thing.

I agree 100% with this.


@Albain, I've fixed my patch. Now the check is done in the parser, so any changes are applied to the entire chain.

- Fix for subtitles when movie dimensions are missing or incomplete in SSA/ASS scripts. Now ffdshow behaves like VSFilter in this regard.
- Fix for blur being (almost) always activated with SSA/ASS subs, now it responds to the Blur setting in the Font section.

Patch: http://www.mediafire.com/?dyhwwehyijz
ICL10.1 build: http://www.mediafire.com/?2zjerzwz3mz

[follz20]

Just a quick question: Why does ffdshow say the output for all DTSHD MA tracks as 96khz, 8 channel @ 1536 kbps?

I'm sure this is probably addressed in this thread, but at 576 pages long it's a little hard to find!

Cheers ;)

[XhmikosR]

If you disable vectorization in ffdshow.ax you're disabling it in all the image processing filters, like deinterlacers, etc. That's what I'm asking, if you have done any tests to determine if for example yadif is slower using Qvec. It should be interesting, I've not made the change to ICL11 yet just because of this Qvec thingy, I don't know if ICL10 without Qvec is faster or slower than ICL11 with Qvec in the key parts of ffdshow (yadif, sharpen, etc.)

Just for the record, now that I'm working with the subs I use MSVC builds, and with deband+resize+sharpen+subtitles+RGB32HQ they have measurably higher CPU consumption (up to 15%) with the specs in my sign than my ICL10.1 builds, but I've not benched them. I imagine than in slower computers the difference will be bigger.

I'm interested in that build BTW.

I agree 100% with this.


Hmm you are right. Keep in mind, that maybe there's another solution to this problem. I mean, I found out by trial and error that vectorization breaks ffdshow in Info & CPU tab. I haven't benchmarked it but I could do it, although I'm not quite sure how to measure the performance of the internal filters. Maybe I could just use the cpu cycles as a comparison.

EDIT: OK, For the video part I can still use timecodec. I'll post my results soon, but from my initial tests I see that indeed ICL 11 even with Qvec- (vectorization disabled) is ~15% faster than MSVC.

RESULTS

ffdshow r3379 deband+sharpening+resize to 1920x1200

Big Buck Bunny 720p x264 (ffmpeg-mt)
ICL 11.1.065 Qvec-
User: 46s, kernel: 0s, total: 46s, real: 62s, fps: 99.6, dfps: 74.2
User: 45s, kernel: 0s, total: 46s, real: 61s, fps: 99.9, dfps: 74.3
User: 46s, kernel: 0s, total: 46s, real: 62s, fps: 98.2, dfps: 73.2
User: 45s, kernel: 0s, total: 45s, real: 62s, fps: 100.7, dfps: 73.6
User: 47s, kernel: 0s, total: 47s, real: 62s, fps: 97.3, dfps: 73.5
User: 46s, kernel: 0s, total: 46s, real: 62s, fps: 98.2, dfps: 73.5

MSVC 2008 SP1
User: 58s, kernel: 0s, total: 58s, real: 73s, fps: 78.8, dfps: 62.5
User: 56s, kernel: 0s, total: 56s, real: 72s, fps: 81.6, dfps: 63.2
User: 55s, kernel: 0s, total: 55s, real: 72s, fps: 82.3, dfps: 63.3
User: 58s, kernel: 0s, total: 58s, real: 73s, fps: 78.4, dfps: 62.5
User: 56s, kernel: 0s, total: 56s, real: 72s, fps: 81.1, dfps: 63.3
User: 57s, kernel: 0s, total: 57s, real: 72s, fps: 80.2, dfps: 63.2


Big Buck Bunny 720p XviD (libavcodec)
ICL 11.1.065 Qvec- (vectorization disabled)
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 71.9, dfps: 59.3
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 70.3, dfps: 59.5
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 71.6, dfps: 59.6
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 73.3, dfps: 59.3
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 70.9, dfps: 59.5
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 72.9, dfps: 58.9

ICL 11.1.065 (vectorization enabled)
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 72.1, dfps: 59.6
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 71.6, dfps: 59.7
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 73.5, dfps: 59.6
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 69.6, dfps: 59.7
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 73.7, dfps: 59.7
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 71.0, dfps: 59.6

ICL 10.1 r3356 clsid
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 70.5, dfps: 59.6
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 70.3, dfps: 59.4
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 71.9, dfps: 59.8
User: 27s, kernel: 0s, total: 27s, real: 33s, fps: 72.4, dfps: 59.4
User: 27s, kernel: 0s, total: 28s, real: 33s, fps: 71.1, dfps: 59.8
User: 28s, kernel: 0s, total: 28s, real: 33s, fps: 70.9, dfps: 59.6

MSVC 2008 SP1
User: 32s, kernel: 0s, total: 32s, real: 38s, fps: 61.1, dfps: 51.3
User: 33s, kernel: 0s, total: 33s, real: 39s, fps: 59.5, dfps: 50.9
User: 32s, kernel: 0s, total: 32s, real: 38s, fps: 60.9, dfps: 51.3
User: 32s, kernel: 0s, total: 32s, real: 38s, fps: 61.1, dfps: 51.4
User: 33s, kernel: 0s, total: 33s, real: 39s, fps: 59.9, dfps: 50.9
User: 32s, kernel: 0s, total: 32s, real: 38s, fps: 60.8, dfps: 51.4

EDIT: If someone creates some nice graphs we could update the wiki (http://ffdshow-tryout.sourceforge.net/wiki/faq:performance_with_filters) with my results which are closer to the newer processors compared to the Pentium 2 used. I will also upload my samples if go for that.

[albain]

@Albain, I've fixed my patch. Now the check is done in the parser, so any changes are applied to the entire chain.

- Fix for subtitles when movie dimensions are missing or incomplete in SSA/ASS scripts. Now ffdshow behaves like VSFilter in this regard.
- Fix for blur being (almost) always activated with SSA/ASS subs, now it responds to the Blur setting in the Font section.

Patch: http://www.mediafire.com/?dyhwwehyijz
ICL10.1 build: http://www.mediafire.com/?2zjerzwz3mz

Thanks, I am testing it right now and I will commit it then

[Blight]

Request:
Any chance of FFDShow exposing an IAMStreamSelect interface (at least for subs)?

[STaRGaZeR]

Hmm you are right. Keep in mind, that maybe there's another solution to this problem. I mean, I found out by trial and error that vectorization breaks ffdshow in Info & CPU tab. I haven't benchmarked it but I could do it, although I'm not quite sure how to measure the performance of the internal filters. Maybe I could just use the cpu cycles as a comparison.

I agree, this is interesting to say the least. Also, can you do another test with "Big Buck Bunny 720p XviD (libavcodec)" and ICL 11.1.065 but this time without Qvec-? Apples to apples comparison with the ICL10 build, even if it breaks Info & CPU.

[XhmikosR]

I updated my previous post. The speed gain is marginal, like when using specific cpu instructions. So to summarize, there's no speed loss between ICL 10 and ICL 11. But since no one can download ICL 10 anymore, ffdshow should update its ICL project files.

[fastplayer]

EDIT: If someone creates some nice graphs we could update the wiki (http://ffdshow-tryout.sourceforge.net/wiki/faq:performance_with_filters) with my results which are closer to the newer processors compared to the Pentium 2 used. I will also upload my samples if go for that.
I still have the Calc-template of these old tests somewhere here, so just post the complete results with/without filters and I'll make the graph.

[clsid]

If you want ICL11 projects then go ahead and add them. Just don't remove the ICL10 ones.

[STaRGaZeR]

I updated my previous post. The speed gain is marginal, like when using specific cpu instructions. So to summarize, there's no speed loss between ICL 10 and ICL 11. But since no one can download ICL 10 anymore, ffdshow should update its ICL project files.

Yeah, margin of error. Can you create and share the ICL11 projects so everything can be built with ICL11?

[XhmikosR]

OK then, I'll try to clean up my ICL 11 project files in the next days.

[albain]

System Setup

Windows 7 Professional x64
Zoomplayer 7 Home Max 7.10 Alpha 3
Gabest Mpeg Splitter - standalone filter svn 1809 x32
Haali splitter - 27-03-2010
FFDShow rev 3374 x32
EVR renderer output


Blu-ray Disc Sword Of The Stranger

Subtitles don't display with rev 3371 - 74, working with 3370. Using FFDShow decoder subs colour is orange, with the Microsoft decoder subs are the correct colour (blue).

Sample http://www.mediafire.com/?txzygwnnxmt Subs http://www.mediafire.com/?lkwttbz4ndd


Hi, I have tested your sample and it works fine on my side : the subtitles are displayed when ffdshow loading external sup file or using the embedded stream.
However they are yellow and it seems to be the right color : if you open the sup file with supread you'll get the same color.
To confirm this you could try to play your bluray with arcsoft,powerdvd or an electronic bluray player

[rpm7200]

if i enable lfe crossover, ffdshow always uses 32 bit floating point. is this a bug or something?

[FreeFall]

Albain,

Thanks for looking at it, your right I just tested with powerdvd and the subs are yellow, in earlier ffdshow builds they were originally blue so I just figured something got messed up.

I forgot to mention that the subs are orange when using the YV12 or YUY2 output, using RGB32 output displays the subs yellow as you've said.

I tested this using rev 3370 from the xvidvideo website as it's the latest build that works with Blu-ray subtitles for me. I tested rev 3371 & 3373 but the subs just won't display when selected, dvd subtitles worked fine except for the problem with the Drunken Master sample.

You said the Blu-ray subtitles are working on your end so I'll wait for a new build to test.


FreeFall

[djesteban]

@dev team
Ok, so, I have this sporadic problem when playing mkv's (with mpc-hc x64 and ffdshow x64 clsid latest and older builds) where at some point in the movie, it will start dropping the framerate for the duration of a particular scene, and right when the scene change it will jump to like 40 fps for 1-2 seconds and then go back to normal. For example, I am getting a steady 23.976, then it changes to another scene and the framerate goes down to 18fps, then changes scene again, it goes to 40fps for like a second or two and then back to a steady 23.976.
I have noticed that it happens on scene where the bitrate is (or seems to me) higher than the rest of the movie. Check the stats below that I have extracted from the info box in MPC-HC during playback.
http://img291.imageshack.us/img291/9262/fpsdp.jpg

I have also uploaded a sample (http://netfolder.in/folder.php?folder_id=0nRRYNE) (a little more than a hundred MB). Listen to it from the beginning, you will see it plays ok until you get to that "grainy, desaturated scene" part. Then you should see the fps drop slightly in MPC-HC... but it's enough to be noticeable and I didn't put the sound in this example but I can tell you that it jerks off the sound also.

Be aware that this DOES NOT happen at all when I am using CoreAVC 2.0 and plays smoothly all the way.
This is only one example, but I have a couple more mkv where this problem also arises.

Hope this can be fixed :P
Thanks in advance

[hoborg]

@albain:

-Fixed audio/subtitles streams switching from keyboard
:thanks:

Just installed rev 3383 from http://xvidvideo.ru/ and...
...stil not working, pressing CTR+ALT+F4/numpad 1 simply does nothing, no OSD, no strem switched.
Only for me? :/

[clsid]

@djesteban
this might help:
Options -> Decoder options -> disable "Drop frame on delay"

[dann23]

Catalyst 10.4 brings H.264 Level 5.1 support

http://www2.ati.com/relnotes/Catalyst_104_release_notes.pdf

I tested this with ffdshow but there are some artifacts (maybe a bug in ffdshow or in driver). But the movie is watchable. But I have to choose Ignore number of reference frames. Is this how is supposed to be? Or ffdshow must be able to detect if the driver is capable to decode 5.1 profile? And when ffdshow dxva will decode mpeg2?

[djesteban]

@djesteban
this might help:
Options -> Decoder options -> disable "Drop frame on delay"

Yeah, but that will just drop frames... do you know why it doesn't drop any frames with CoreAVC but doesn't play back smoothly with FFdshow?

[albain]

Request:
Any chance of FFDShow exposing an IAMStreamSelect interface (at least for subs)?

It already does

[albain]

Just installed rev 3383 from http://xvidvideo.ru/ and...
...stil not working, pressing CTR+ALT+F4/numpad 1 simply does nothing, no OSD, no strem switched.
Only for me? :/

Why do you have 3 activation keys ? Normally this is 2

For ex I mapped audio stream switching to "K" and when I do control+alt+K it switches audio stream

[albain]

Albain,

Thanks for looking at it, your right I just tested with powerdvd and the subs are yellow, in earlier ffdshow builds they were originally blue so I just figured something got messed up.

I forgot to mention that the subs are orange when using the YV12 or YUY2 output, using RGB32 output displays the subs yellow as you've said.

I tested this using rev 3370 from the xvidvideo website as it's the latest build that works with Blu-ray subtitles for me. I tested rev 3371 & 3373 but the subs just won't display when selected, dvd subtitles worked fine except for the problem with the Drunken Master sample.

You said the Blu-ray subtitles are working on your end so I'll wait for a new build to test.


FreeFall

You're right, in YUV mode they are orange I can reproduce that. This apart I don't have any other issues

[hoborg]

Why do you have 3 activation keys ? Normally this is 2

For ex I mapped audio stream switching to "K" and when I do control+alt+K it switches audio stream

I am not sure if i understand it correctly.
Default FFDShow activation keys are CTRL+ALT + mapped key.
For example CTRL+ALT+O to show/hide OSD, CTRL+ALT+S to show/hide subtitles. I have no problems with this (working fine).
But CTRL+ALT+F4 or CTRL+ALT+Nunpad1 simply does nothing for me, i tryed remap CTRL+ALT+F4 to CTRL+ALT+K, but still nothing (is there some OSD message about strem swithched?)

Here is my settings: (audio switcher/subtitles is enabled, testing Samurai Champaloo sample)
http://hobring.esero.net/saf/ffdshow/streams.png

[onomatopellan]

You're right, in YUV mode they are orange I can reproduce that. This apart I don't have any other issues
I have the same problem. Since rev3371 I can't see PGS bluray subtitles. Instead it shows NullTextRenderer at the top of filter list. :confused:
http://i43.tinypic.com/bi6on8.png

[XhmikosR]

@devs: I have updated libsamplerate to v.0.1.7 locally. The increase in the installer's size is ~800KB. Test binaries (x86 and x64) and the patch can be found here (http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335cae32d40843051dd14). The changelog for libsamplerate since 0.1.2 mentions:

# Version 0.1.3 (Mar 23 2008) Huge quality improvements to two best SINC based converters.
# Version 0.1.4 (Jul 02 2008) Fix segfault when using extremely low conversion ratios.
# Version 0.1.5 (Jan 11 2009) Optimisation resulting in dramatic throughput improvements ( See here (http://www.mega-nerd.com/erikd/Blog/CodeHacking/SecretRabbitCode/rel_0_1_5.html).).
# Version 0.1.6 (Jan 27 2009) Minor bug fix in test suite (account for rounding error on x86_64).
# Version 0.1.7 (Feb 14 2009) Fix a segfault bug. Fix compilation under MSVC.

What do you think? I mean, it's an important increase in size, but the changes sound very interesting.

[fastplayer]

Maybe you can compensate with a higher compression ratio in the installer?
LZMA2 was recently added to Inno.

[XhmikosR]

I've already tried it, no improvement unfortunately.

[clsid]

LZMA2 doesn't help much.

As usual I vote against the huge libsamplerate. Few people actually use it and most of them would probably not even hear a difference in a double-blind test.

[fastplayer]

I guess UPXing ff_samplerate.dll wouldn't make any difference too...

[tetsuo55]

the size increased because the current version was fundamentally flawed and the author had to refactor a large portion of code iirc

[Shark007]

@devs: I have updated libsamplerate to v.0.1.7 locally. The increase in the installer's size is ~800KB. Test binary based on r3387 can be found here (http://www.mediafire.com/?jugemjcjngk) and the patch is here (http://www.mediafire.com/?un0gju2itn0). The changelog for libsamplerate since 0.1.2 mentions:



What do you thing? I mean, it's an important increase in size, but the changes sound very interesting.

could you compile an x64 version for testing please?
Thanks in advance . . . (providing just the single dll would suffice)

[XhmikosR]

@Shark007: I updated my previous (http://forum.doom9.org/showthread.php?p=1396106#post1396106) post (both the binaries and patch). Keep in mind I haven't tested the x64 version at all since I don't have 64bit Windows.

[Sarasa]

I think I have a bug with the Resize & Aspect option

video size that don't work correctly : 400x300 / 480x352 / 512x384 / 576x432 / 640x480 / 640x496 / 704x512 / 1008x762
No problem with : 640x360 / 704x388 / 704x396 / 704x400 / 720x400 / 1008x560


Here is a test done with version 3387 and two video clip (640x480 / 704x396).
The parameter for Border are left with default option & setting > Lock / Bicubic
if I put this resize parameter
http://i43.tinypic.com/30963j7.jpg

The clip 1 is not resized right, has you can see the image has a black border.
http://i40.tinypic.com/6edizc.jpg
but clip 2 is resized correctly
http://i39.tinypic.com/2a9prbo.jpg

With this resize parameter
http://i41.tinypic.com/15e7hnq.jpg
It resize correctly the two clip
http://i41.tinypic.com/msbxy1.jpg
http://i39.tinypic.com/vxnec4.jpg


I'm doing something wrong or forgetting a option ? :confused:

[adam777]

Well, although I haven't seen anything in the changelog, I decided I'll give rev. 3390 a try, perhaps the issue mentioned here - http://forum.doom9.org/showthread.php?p=1394378#post1394378, is gone.
The good news is the issue is not presented anymore, the bad news is it's because subtitles are not displayed at all :-)
Back to rev. 3367.

[XhmikosR]

@devs: I have updated all the mingw libs (libgcc.a, libmingwex.a, libws2_32.a) and the boost lib (r61712). I've tested MSVC x86/x64 Release builds x86 ICL11 Release build and I didn't face any problems. Although I had to add BOOST_NO_RTTI in the Preprocessor definitions for ff_vfw in order to get the ICL11 x86 build to compile. From Debug builds, I've only tested MSVC Debug x86.

Are the mingw libs in the SVN specially compiled? Because I see that they have a big difference in size compared to my mingw ones.

PS. I still wait for libsamplerate. Only clsid has replied so far and he is against the update.

[STaRGaZeR]

Well, although I haven't seen anything in the changelog, I decided I'll give rev. 3390 a try, perhaps the issue mentioned here - http://forum.doom9.org/showthread.php?p=1394378#post1394378, is gone.
The good news is the issue is not presented anymore, the bad news is it's because subtitles are not displayed at all :-)
Back to rev. 3367.

Can you send me a script or a video with no displayed subtitles?

About libsamplerate I say go with the update, better quality is always better. There are a lot of things that people use less than resampling that could be removed from ffdshow yet they're there :p

[namaiki]

http://i43.tinypic.com/30963j7.jpg
The clip 1 is not resized right, has you can see the image has a black border.
http://i40.tinypic.com/6edizc.jpg
but clip 2 is resized correctly
http://i39.tinypic.com/2a9prbo.jpg

In both clips, there are black bars on the top and bottom of the video area which is 16:9. You can see and assume that clip 2 is taking up the entirety of the available video area which is resized 1280x720 pixels.

In clip one, there is also black bars in the video on the left and right within the 16:9 area in the centre of the screen touching the inside of the screen, so now there are black bars left and right from ffdshow and top and bottom because there is nothing there. If you put the video into a window (not full screen), you will see that there are only black bars on the left and right.


With this resize parameter
http://i41.tinypic.com/15e7hnq.jpg
It resize correctly the two clip
http://i41.tinypic.com/msbxy1.jpg
http://i39.tinypic.com/vxnec4.jpg

Now the video area is 4:3 and is taking up the whole screen. There is no area that is not video area even if it is black.

In the second clip, the black bars on the top and bottom are part of the video and put there by ffdshow to keep the original aspect of the video. If the second video was resized to 1440x1080 without aspect ratio correction/black bars, the aspect would be 4:3 and not the original 16:9.

TL;DR the aspect ratio of the clips is being kept.

[STaRGaZeR]

Albain, r3374 causes subtitles to disappear on a seek. This happens at least with subtitles readed from file. The only solution is to reload them. The culprit is:

if (subs) subs.onSeek();

Removing this line fixes the issue. Maybe that's why adam777 can't see any subs now.

[FreeFall]

Albain,

I just tested rev 3383 and 3387, just playing the original Blu-ray disc Sword Of The Stranger with zoom player but the subtitles are not displayed. Build 3370 is the last working version for me, I don't know if anything changed with the way the subs are displayed after 3370 or if the problem is with zoom player.

You can download a trial copy of Zoom Player Home MAX here http://www.inmatrix.com/files/zoomplayer_download.shtml.
Here is the update http://www.mediafire.com/?zldjdyomz2w which fixes a problem where ffdshow's subtitles were being disabled, copy and replace original .exe in zoom player install folder.

Untick the option Disable FFDShow subtitle rendering when no subtitle files/streams are found (see image). I'm using the Win7 Direct Show Filter Tweaker 3.1 to change video renderers to FFDShow.

System Setup:

Windows 7 Professional 64bit
GeForce GTX 280 - Forceware 197.25
AnyDVD HD - 6.6.3.4
Zoomplayer 7 Home Max - 7.10 Alpha 3
Gabest Mpeg Splitter - standalone filter svn 1824 32bit
Haali splitter - 27-03-2010
EVR renderer, YV12 Output


Let me know if you need any more info.

Thanks

FreeFall

[Sebastiii]

Albain, r3374 causes subtitles to disappear on a seek. This happens at least with subtitles readed from file. The only solution is to reload them. The culprit is:

if (subs) subs.onSeek();

Removing this line fixes the issue. Maybe that's why adam777 can't see any subs now.

Hi, cool,

I have to test, i have a strange issue, and i try to figure out, but on m2ts and last FFDshow, the sub out pin don't want to connect to input pin of FFDshow, i have enable subtitle.

If i use old version (3364) but not all old version the connect is made, i use a quick test by graphstudio to look this.

Also, i have see that you have work for SSA positionning.
I have a question about .srt file, it's normal that i don't succeed to have the same (font, position) that i obtain with directvobsub and MPC-HC ?

Another think, but i thik that all issue come from me lol.
If i have a movie with in the same directory an .srt with part of the name of the movie, the sub is load automatly with the bad .srt

Example : movie -> sample.mkv and sub in directory : sample2.srt
Agin in MPC-HC or Directvobsub, it take only subs with the match name and don't load sample2.srt.

Thank you :)
Cheers,
Seb.

[ikarad]

Albain,

I just tested rev 3383 and 3387, just playing the original Blu-ray disc Sword Of The Stranger with zoom player but the subtitles are not displayed. Build 3370 is the last working version for me, I don't know if anything changed with the way the subs are displayed after 3370 or if the problem is with zoom player.

You can download a trial copy of Zoom Player Home MAX here http://www.inmatrix.com/files/zoomplayer_download.shtml.
Here is the update http://www.mediafire.com/?zldjdyomz2w which fixes a problem where ffdshow's subtitles were being disabled, copy and replace original .exe in zoom player install folder.

Untick the option Disable FFDShow subtitle rendering when no subtitle files/streams are found (see image). I'm using the Win7 Direct Show Filter Tweaker 3.1 to change video renderers to FFDShow.

System Setup:

Windows 7 Professional 64bit
GeForce GTX 280 - Forceware 197.25
AnyDVD HD - 6.6.3.4
Zoomplayer 7 Home Max - 7.10 Alpha 3
Gabest Mpeg Splitter - standalone filter svn 1824 32bit
Haali splitter - 27-03-2010
EVR renderer, YV12 Output


Let me know if you need any more info.

Thanks

FreeFall


same problem but with MPC-HC and ffdshow since 3371

[onomatopellan]

Those who can't see the Bluray subtitles, can you also see THIS (http://forum.doom9.org/showpost.php?p=1396094&postcount=11541) on the top of your filter list?

[FreeFall]

Onomatopellan,

I don't see NullTextRenderer anywhere on my filter list.


Ikarad, Onomatopellan,

Thanks for the feedback it's good to know I'm not the only one having trouble with builds newer than 3370.


FreeFall

[ikarad]

Those who can't see the Bluray subtitles, can you also see THIS (http://forum.doom9.org/showpost.php?p=1396094&postcount=11541) on the top of your filter list?

I don't have Nulltext renderer in my filter list.

But I notice that the problem is more important that I think.

Even in mkv video the subs are not displayed.

With 3371, only bluray subs were not displayed. With 3390 all subs are not displayed.

It's like if support of all subs in ffdshow was disabled by ffdshow team.

[Sarasa]

@namaiki

So the problem is the Aspect Ration conversion.

Let's see what are the AR of my tests clips, before > after conversion to 1280x720 (using FFDSHOW OSD for the info)

400x300 => 4/3 > 16/9 *; 480x352 => 15/11 > 16/9 *; 512x384 => 4/3 > 16/9 *
576x432 => 4/3 > 16/9 *; 640x480 => 4/3 > 16/9 *; 640x496 => 40/31 > 16/9 *
704x512 => 11/8 > 16/9 *; 1008x762 => 63/47 > 16/9 *
* ffdshow had a big black border around them.

640x360 => 16/9 > 16/9; 704x388 => 176/96 > 16/9 *
704x396 => 16/9 > 16/9; 704x400 => 44/25 > 16/9 *
720x400 => 9/5 > 16/9 *; 1008x560 => 9/5 > 16/9 *
* Add only a black border up/down

The thing that disturb me, is why FFdshow add a big border around the video in some AR conversion and don't try to use more of the surface O_o
And yes I know, some of my clips have strange AR -_-;

---------------


I have also another problem, but with sub this time ^^;

Spec of Clip : MKV > AVC1 + Vorbis + ASS sub
FFDSHOW > ffmpeg-mt h264 + Tremor

Using DirectVobSub 2.39 > http://i41.tinypic.com/wbucl5.jpg

Using ffdshow (3390) option activate in subtitles (placement/text/font/bluray default option)
> http://i41.tinypic.com/2dtn41e.jpg
If maintain outside text inside the picture "enable" > http://i39.tinypic.com/2ce4sp0.jpg
If maintain outside text inside the picture "Disable" > http://i41.tinypic.com/14afea.jpg

I have attached the ASS (remove txt at the end), if it can help :)

[Sebastiii]

I don't have Nulltext renderer in my filter list.

But I notice that the problem is more important that I think.

Even in mkv video the subs are not displayed.

With 3371, only bluray subs were not displayed. With 3390 all subs are not displayed.

It's like if support of all subs in ffdshow was disabled by ffdshow team.

Hi,
I didn't have the connection between Pin sub too on last FFDshow after 3370 -> Not sun at all too.

[7084] TtextInputPin::ReceiveConnection
[7084] TtextInputPin::ReceiveConnection
[7084] TffdshowBase::Constructor
[7084] TffdshowDecVideo::Constructor
[7084] TffdshowDecVideoOutputPin::Constructor
[7084] TffdshowDecVideoDXVA constructor
[7084] Join filter graph
[7084] TffdshowVideoInputPin::ReceiveConnection
[7084] TtextInputPin::ReceiveConnection
[7084] Removed from filter graph
[7084] TffdshowDecVideo::Destructor
[7084] TffdshowDecVideoOutputPin::Destructor
[7084] TffdshowBase::Constructor
[7084] TffdshowDecVideo::Constructor
[7084] TffdshowDecVideoOutputPin::Constructor
[7084] Join filter graph
[7084] TffdshowDecVideo::Destructor
[7084] TffdshowDecVideoOutputPin::Destructor


This is that i have when i try so connect PIN.
Maybe is wrong here.

On old revision, when Blu-ray sub works, On my Avatar BD, i have some subtitle not display correctly (it appear on Upper Left corner of the screen) but not all the time -> I think is related between normal sub and n'avi sub (Italic or someting like that).

But like i can't test on recent version maybe it's solved.

I have make a clean installation to see if it's not a setting that broke something -> But not :)

Cheers,
Seb.

Edit : : Link to sample sub display not correctly rev 3370 (http://www.mediafire.com/download.php?zo0trzmgznl)

[adam777]

Albain, r3374 causes subtitles to disappear on a seek. This happens at least with subtitles readed from file. The only solution is to reload them. The culprit is:

if (subs) subs.onSeek();

Removing this line fixes the issue. Maybe that's why adam777 can't see any subs now.

I can confirm that in rev. 3390 with .srt subs, seeking cause subtitles to disappear.
Also, while subtitles are displayed, they are still gabbled as described here - http://forum.doom9.org/showthread.php?p=1394378#post1394378

[Atak_Snajpera]

Image corruption while using ffmpeg-mt with more than 1 thread. (libavcodec and DVXA are also ok)
http://img265.imageshack.us/img265/5303/avatarblurayffmpegmtcor.png

Tested with ffdshow r3370 xxl and r3390 from http://www.xvidvideo.ru

This must be some very old bug because it occurs even in very old revisions as well.

Remuxed BD sample http://www.sendspace.com/file/f9s4te

[STaRGaZeR]

OK, guys try this build. Included fixes:

- Subtitles disappearing after a seek
- Audio switching resulting in silence
- Improve the subtitles GUI a bit (subtitles and placement tab)

http://www.mediafire.com/?tndzdokwqdj

EDIT: clsid, the ICL10 solution is broken again, same issue as the last time (ffdshow excluded from build, and set to x64).

[Sebastiii]

@STaRGaZeR :

Sorry, i can't connect subtitle PIN (Mpc Mpeg Splitter) to FFDshow in put subtitle pin :(

Thanks,
Seb.

[STaRGaZeR]

But that wasn't supposed to fix that, will see if I can do something but I doubt it, albain is your man :)

Just asking, but have you checked this?

http://thumbnails24.imagebam.com/7870/97fa7778698837.gif (http://www.imagebam.com/image/97fa7778698837)

Man the list of bugs these changes have caused is huge, maybe just revert for the time being until they're ironed out?

BTW, some answers I forgot :p

Also, i have see that you have work for SSA positionning.
I have a question about .srt file, it's normal that i don't succeed to have the same (font, position) that i obtain with directvobsub and MPC-HC ?

.srt files contain plain text subtitles, that means you have only text, with no formatting. You're the one to select how you want to display that text: font, color, etc. You can configure all this stuff in the Font and Text sections of ffdshow's subtitles filter, MPC-HC has similar options.

Another think, but i thik that all issue come from me lol.
If i have a movie with in the same directory an .srt with part of the name of the movie, the sub is load automatly with the bad .srt

Example : movie -> sample.mkv and sub in directory : sample2.srt
Agin in MPC-HC or Directvobsub, it take only subs with the match name and don't load sample2.srt.

Yeah, I've noticed that too. I don't know if that's a bug or a feature, so I'd not touch it :p

[Sebastiii]

No pb :)

Ok, i try on my side (my little side) i compare what change and it's maybe related with .sup fil, i have change 2 line of code but i didn't see that i'm on rev 3377 -> but it works.
So something maybe missing : the patch has made : PGSandSUP_test.patch (http://sebinternet.free.fr/PGSandSUP_test.patch)
Seb :)

Edit : : ffdshow_rev3392_20100501.exe with my change (test only) (http://sebinternet.free.fr/ffdshow_rev3392_20100501.exe)

[STaRGaZeR]

Are you having this problem with embedded subtitles or with external .pgs/.sup files?

See my previous post BTW :p

[STaRGaZeR]

Sarasa, if you don't want ffdshow to proccess aspect ratio internally deselect "Process pixel aspect ratio internally".

I have also another problem, but with sub this time ^^;

Spec of Clip : MKV > AVC1 + Vorbis + ASS sub
FFDSHOW > ffmpeg-mt h264 + Tremor

Using DirectVobSub 2.39 > http://i41.tinypic.com/wbucl5.jpg

Using ffdshow (3390) option activate in subtitles (placement/text/font/bluray default option)
> http://i41.tinypic.com/2dtn41e.jpg
If maintain outside text inside the picture "enable" > http://i39.tinypic.com/2ce4sp0.jpg
If maintain outside text inside the picture "Disable" > http://i41.tinypic.com/14afea.jpg

I have attached the ASS (remove txt at the end), if it can help :)

Please upload that ass to mediafire, or better yet, PM me the sample, I'm interested in this one.

[Sebastiii]

@STaRGaZeR,

Are you having this problem with embedded subtitles or with external .pgs/.sup files?

Yes i have set the same option in ffdshow, it's works with external sub (.srt) but not in .sup, but i'm not sure about my sample with .sup.


.srt files contain plain text subtitles, that means you have only text, with no formatting. You're the one to select how you want to display that text: font, color, etc. You can configure all this stuff in the Font and Text sections of ffdshow's subtitles filter, MPC-HC has similar options.


Ok i understand, for setup font etc. but if i set this normal sub is ok but when sub is to be just a little up (like i see in MPC-HC subtitle renderer) i see it very up in FFDshow (surely something to setup but i don't succeed).

A screen is better than word :

Normal positionning :
http://sebinternet.free.fr/MPC-HC_renderer.png

Not Normal positionning :
http://sebinternet.free.fr/FFDshow_Subtitle.png

Link to sample .srt : SRT File (http://sebinternet.free.fr/sample-positionning.srt)

Big thanks :)
Seb

[Sarasa]

@STaRGaZeR

Send it to you by PM :)

[STaRGaZeR]

@Sebastii

Ah, I see. There's a \pos token in that line. Can you send me a few seconds of video around that line? Also, did r3383 change anything?

Yes i have set the same option in ffdshow, it's works with external sub (.srt) but not in .sup, but i'm not sure about my sample with .sup.

And you say that with your patch it works? If I'm reading it correctly what you've done is remove the suppgs check, and suppgs is set when you check the "Blu-ray subtitles" checkbox in the subtitles section, so if you have it enabled your patch should have no effect :confused:

[Sebastiii]

@Sebastii

Ah, I see. There's a \pos token in that line. Can you send me a few seconds of video around that line? Also, did r3383 change anything?


This the sample lenght 50 seconds (the subline appear arround 40 seconds)
sample_position-001.mkv (http://sebinternet.free.fr/sample_position-001.mkv)

I think that r3383 didn't change something :(


And you say that with your patch it works? If I'm reading it correctly what you've done is remove the suppgs check, and suppgs is set when you check the "Blu-ray subtitles" checkbox in the subtitles section, so if you have it enabled your patch should have no effect :confused:

I try to be more clear lol :)

So with my patch : all revision after 3370, i can't get connection between Output and Input pin of FFDshow (for sub and i think the connection is needed only for embedded sub.)

With with all later revision, i can play and see sample with .srt and surely .sup (but the one i have, i'm not sure it's a good .sup/sample).

So i look what change between the version that break the connection and i see that albain has add suppgs (for external .sup).

So i remove it and the connection between pin comeback and now for sure if a sample with .sup doesn't work anymore.

Thank you,
Seb.

[albain]

Hello guys,

I've been busy these last few days, could someone confirm me what's wrong exactly ?

I saw in the last posts that bluray subtitles in external files are not loaded ?

They work fine on my side : please make sure that the .sup extension is in the list of subtitles extension in ffdshow subtitles settings.
I added it but it will be set only on new fresh ffdshow setup.
Maybe the setting should be reset by the setup

Also, keep in mind that there is still a bug in MPC m2ts splitter that does not detect the PGS subtitles (embedded) until it has parsed the first one it meets (which occurs about 10 seconds before it is displayed).
As nobody on MPC team seems to have found the cause, I'll try to have a look on MPC source code.

As I said, I just remade a test and everything works fine : embedded/external subtitles, bluray or others (srt/ssa).

The bugs that are still there are : color problems in YUV mode with PGS subtitles, and some subtitles scalers that don't work correctly in RGB mode

[Sebastiii]

Salut Damien,

I have made some test again, .sup and .srt (external :) ) working on last FFDshow.

But for me (and on my 2 PCs) with rev upper than 3370 the connection between Output PIN subtitle of splitter and Input PIN of FFDshow Subtitle doesn't connect anymore.

I have change the add that you have made in TtextInputPin.cpp :
//(mtIn->subtype==MEDIASUBTYPE_HDMV_PGS && suppgs) || //Bluray subs
//(mtIn->subtype==MEDIASUBTYPE_NULL && mtIn->formattype == FORMAT_SubtitleInfo && suppgs) || //Bluray subs
(mtIn->subtype==MEDIASUBTYPE_HDMV_PGS) || //Bluray subs
(mtIn->subtype==MEDIASUBTYPE_NULL && mtIn->formattype == FORMAT_SubtitleInfo) || //Bluray subs

I have delete this : && suppgs and after the connection is establish and i can see and select sub.
I have another issue about the position on some subs (display in left upper corner) but it's another think to solve later lol.
I don't know why the connection isn't establish when you add the support .sup.

Also, keep in mind that there is still a bug in MPC m2ts splitter that does not detect the PGS subtitles (embedded) until it has parsed the first one it meets (which occurs about 10 seconds before it is displayed). As nobody on MPC team seems to have found the cause, I'll try to have a look on MPC source code.

I have talk a little with Casimir666 and he said me, that on some .m2ts that didn't detect sub, hte splitter read BDMV\CLIPINF\*.clpi to detect subtitle stream, audio stream etc. and if we have a .m2ts alone, the sub appear after it has parsed the first one.

But, like i answer him, is that Mediainfo detect all stream directly and quickly (because he suppose that mediainfo analyse the whole stream and after it give all information but on my test when i mediainfo .m2ts that not working direclty, i didn't see my hard drive read the whole file so i suppose that it take it correctly).

Maybe it's possible to implement the detection source of mediainfo in Mpeg splitter source ?

Thank you damien and all guys :)
Seb.

[STaRGaZeR]

The bugs that are still there are : color problems in YUV mode with PGS subtitles, and some subtitles scalers that don't work correctly in RGB mode

Add to that:

- SRT subs: subtitles with 2 lines are displayed in one line.
- The input pin of ffdshow doesn't accept Blu-ray subs because suppgs is always 0 regardless of the blu-ray subs checkbox in the subs options. I checked this with a debug build.
- Subtitles disappear after a seek, I've found the cause to be this line in TimgFilterSubtitles.cpp: if (subs) subs.onSeek();. Removing it fixes the problem.
- Hebrew subtitles broken (adam777's report).
- Sebastii's issues as reported here (http://forum.doom9.org/showthread.php?p=1396354#post1396354).

[SamuriHL]

http://support.microsoft.com/kb/981738

Not sure if this is helpful to anyone but I thought it was interesting.

[FreeFall]

Sebastiii,

I Just tested your patched build and Blu-ray subtitles are working again, good work.


Albain,

For me internal Blu-ray subtitles were not working, with Sebastiii's patch they are working again.


FreeFall

[Sebastiii]

Sebastiii,

I Just tested your patched build and Blu-ray subtitles are working again, good work.


Albain,

For me internal Blu-ray subtitles were not working, with Sebastiii's patch they are working again.


FreeFall

FreeFall good to know, but with this build, you can't use .sup subtitle file and when we seek sub disappear.
I try something else to see :) and with the .patch from STaRGaZeR.
I hope albain nail it lol :) (i know it does)
Seb.

[Sebastiii]

http://support.microsoft.com/kb/981738

Not sure if this is helpful to anyone but I thought it was interesting.

Hi,

Good to know too, i have a strange thing with MediaPortal and MPC Video codec (black screen a few second) maybe it help.

Thanks,
Seb

[Sebastiii]

Hi,

I have made change, i hope Albain can look this :
Blu-ray sub is reenable and .sup too. (But on my test only the first one appear for .sup) if someone can test it :)

Patch : SUP_SEEK_Build3.patch (http://sebinternet.free.fr/SUP_SEEK_Build3.patch)
Build : ffdshow_rev3393_20100501_build3.exe on Free (http://sebinternet.free.fr/ffdshow_rev3393_20100501_build3.exe) and ffdshow_rev3393_20100501_build3.exe on Mediafire (http://www.mediafire.com/download.php?bwy2mkyfv2j)

Thx,
Seb.

[STaRGaZeR]

Sebastii, the problem with your patch (the input pin part) is that you're effectively disabling the Blu-ray checkbox in the subtitles section. Uncheck the checkbox and you'll see the subs still work when they shouldn't :p
The real problem here is that suppgs is always 0, that's why when you remove it from the if the subs work.

[Sebastiii]

Hi,

Yes it's not a good patch :)

Otherwise, could you test rev3390 (the no patched version) with .sup, on my side i can only show the first, with suprip i can confirm that all sub is here.
Thx :)

[albain]

Hi,

I have fixed the problem with embedded subtitles.

I was also able to reproduce the seeking problem with sup files, I am on it

[Sebastiii]

You're great :)
Seb :)

Edit: Rev 3394 bluray connection subtitle is back :) works like a charm with PGS embedded sub :) Thanks

[albain]

Second problem fixed too : external bluray subtitles
Some (or all) were not displayed

[Sebastiii]

i'm testing right now :)

Edit : Sorry, i have tested on my sample with external .sup PGS file, only the first subs appear.
I try something else to see.
Thanks,
Seb.

[onomatopellan]

Tried to install rev3395 from xhmikosr and get this error:
"This build of fdshow requires a CPU with SSE2 extension support. Your CPU does not have those capabilities." :(
Would be SSE2 a requisite from now on?

[XhmikosR]

Use a non ICL11 build.

[SamuriHL]

Hi,

Good to know too, i have a strange thing with MediaPortal and MPC Video codec (black screen a few second) maybe it help.

Thanks,
Seb

Yea, I don't know what exactly it's for, but, I thought I'd throw it out there. I haven't installed it yet myself.

[Midzuki]

Feature request:

Stereo DTS can be flagged as "Left Total + Right Total",
however ffdshow sees no difference between plain 2.0 DTS and "matrix-encoded" stereo DTS. It would be nice if

1) ffdshow didn't ignore the "Lt + Rt" flag ;)

2) ffdshow automagically applied DPL2 upmix to such streams :cool:

[Sebastiii]

Yea, I don't know what exactly it's for, but, I thought I'd throw it out there. I haven't installed it yet myself.

I have try but didn't change something on MP.
Don't know what to say :)
Seb.

[onomatopellan]

Use a non ICL11 build.
Ok, I'll do that. :thanks:

[STaRGaZeR]

Second problem fixed too : external bluray subtitles
Some (or all) were not displayed

I've found the code that causes one of the SRT problems (subtitles with 2 lines are displayed in one line): the changes in prepareKaraoke in r3368. Maybe fixing this also fixes Hebrew subtitles, but I don't have any Hebrew subs to test. Hope this helps.

[albain]

I've found the code that causes one of the SRT problems (subtitles with 2 lines are displayed in one line): the changes in prepareKaraoke in r3368. Maybe fixing this also fixes Hebrew subtitles, but I don't have any Hebrew subs to test. Hope this helps.

Thanks for inquiring

Isn't it due to the fact that the SSA engine is used instead of SRT one ?

Have you tried to uncheck the "Support for SSA tags inside srt subtitles" ?

[adam777]

I've found the code that causes one of the SRT problems (subtitles with 2 lines are displayed in one line): the changes in prepareKaraoke in r3368. Maybe fixing this also fixes Hebrew subtitles, but I don't have any Hebrew subs to test. Hope this helps.

Well, the problem with Hebrew subtitles as far as I could see is 2 lines subtitles displayed as one line, in a reverse order, so I suppose that display 2 lines as 2 lines might solve it. :thanks:

[STaRGaZeR]

Thanks for inquiring

Isn't it due to the fact that the SSA engine is used instead of SRT one ?

Have you tried to uncheck the "Support for SSA tags inside srt subtitles" ?

Nope:

- SSA engine, prepareKaraoke disabled --> Good
- SSA engine, prepareKaraoke enabled --> Bad
- SRT engine, prepareKaraoke disabled --> Good
- SRT engine, prepareKaraoke enabled --> Bad

I force the engines by commenting out everything except processSSA or processHTML respectively in format. I enable/disable prepareKaraoke commenting the first if, so return is reached first.

[Jeremy Duncan]

Hi, kurt_pruenner updated LibMpeg2 in revision 1194.
The affected files were:
For TvideoCodecLibmpeg2.cpp

19 #include "stdafx.h"
20 #include "IffdshowDecVideo.h"
21 #include "TvideoCodecLibmpeg2.h"
22 #include "Tdll.h"
23 #include "Tconfig.h"
24 #include "libmpeg2/include/attributes.h"
25 #include "libmpeg2/libmpeg2/mpeg2_internal.h"
26 #include "libavcodec/avcodec.h"
27 #include "ffdebug.h"
28 #include "dsutil.h"
29 #include "cc_decoder.h"

For TvideoCodecLibmpeg2.h

4 #include "TvideoCodec.h"
5 #include "libmpeg2/include/mpeg2.h"

This particular patch made the Libmpeg2 video decoder ablse to see the root menu when using Avisynth in ffdshow that used buffers.
For the mvtools frame doubler I can't see the root menu in Windows 7 64 bit premium.
I suspect that since libmpeg2 has been updated since kurt_pruenner updated it that the files he patch have since been updated and his modifications have been erased.

Once before kurt_pruenner updated libmpeg2 the libmpeg2 was updated and kurt_pruenner updated the files again so they acted before.
So since this error has happened before when the libmpeg2 was updated it could have happened again.
I would install the src files for the ffdshow trunks: 1193, 1194, and the newest version and compare diff files and merge the 1194 into the newest version but I'm lazy and it's not really my job as I'm not a trusted developer here, so I ask you kind people to make the newest libmpeg2 act like it did when Kurt updated it. :)

Oh yeah, once you make the libmpeg2 act like kurt made it then commit the patch to the svn, don't just post the diff file here. Please commit the diff files to the svn and make a new build with the changes. :thanks:

[clsid]

I would be useful if you could narrow down the revision were it got broken. The changes haven't been undone.
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/src/codecs/TvideoCodecLibmpeg2.cpp?view=log

[JaylumX]

I have read most of the post before and most of it goes over my head. Are you saying that with the current ffdshow build i can get the aliens language to display without having to enable English subs. That is how far i have got with Avatar. The Navis language appears just like i remember in the cinema with the correct formatting but also the English text. How do i get rid of just the English

Cheers

[_xxl]

I would be useful if you could narrow down the revision were it got broken. The changes haven't been undone.
That revision contains only cosmetics changes.

[_xxl]

Apply push button is not working properly.

[fastplayer]

Apply push button is not working properly.
Confirmed. Just switching between various filters without making any changes causes the "Apply" button to go into "active" state instead of being disabled. IIRC, this started with rev3371.

[_xxl]

External subtitles .srt are not working.

[fastplayer]

External subtitles .srt are not working.
Old news! (http://forum.doom9.org/showthread.php?p=1394859#post1394859) :)

[STaRGaZeR]

It's fixed here, please test: http://www.mediafire.com/?d52mhhtznyj

[albain]

Yes, apply button and external/embedded subtitles are now fixed and should behave equally

I am attacking the problem raised with karaoke methods that affect the subtitles rendering (2 lines rendered on 1)

[fastplayer]

It's fixed here, please test: http://www.mediafire.com/?d52mhhtznyj
Now they load automatically but as soon as you try to switch subtitles in the context menu, they disappear and you can't bring em up again.
The two-line issue seems to be fixed, though.

Edit:
To clarify the issue:
If you have movie.avi and movie.srt, ffdshow loads automatically and everything's fine.
If you have another subtitle in the same directory as the movie file and try to switch to it, subs get disabled.
If you have movie.avi, sub1.srt and sub2.srt in the same directory and choose sub1.srt in the context menu, subs are fine but if you try to switch to sub2.srt, subs disappear. It seems as if ffdshow can handle only one sub at a time. In the context menu you can only choose the very first sub/entry of the list, nothing else.

Edit2:
Stargazer, did you apply some patch to your build? rev3397 from xhmikosr behaves differently (doesn't auto-load at all).

Edit3:
Switching between rev3395 and rev3397 completely disabled subtitling.
Going back to rev3370 fixes everything.

Edit4:
Ignore "Edit3" :D

[Sebastiii]

Yes, apply button and external/embedded subtitles are now fixed and should behave equally

I am attacking the problem raised with karaoke methods that affect the subtitles rendering (2 lines rendered on 1)

Thanks :)

[djesteban]

@dev team
Ok, so, I have this sporadic problem when playing mkv's (with mpc-hc x64 and ffdshow x64 clsid latest and older builds) where at some point in the movie, it will start dropping the framerate for the duration of a particular scene, and right when the scene change it will jump to like 40 fps for 1-2 seconds and then go back to normal. For example, I am getting a steady 23.976, then it changes to another scene and the framerate goes down to 18fps, then changes scene again, it goes to 40fps for like a second or two and then back to a steady 23.976.
I have noticed that it happens on scene where the bitrate is (or seems to me) higher than the rest of the movie. Check the stats below that I have extracted from the info box in MPC-HC during playback.
http://img291.imageshack.us/img291/9262/fpsdp.jpg

I have also uploaded a sample (http://netfolder.in/folder.php?folder_id=0nRRYNE) (a little more than a hundred MB). Listen to it from the beginning, you will see it plays ok until you get to that "grainy, desaturated scene" part. Then you should see the fps drop slightly in MPC-HC... but it's enough to be noticeable and I didn't put the sound in this example but I can tell you that it jerks off the sound also.

Be aware that this DOES NOT happen at all when I am using CoreAVC 2.0 and plays smoothly all the way.
This is only one example, but I have a couple more mkv where this problem also arises.

Hope this can be fixed :P
Thanks in advance

Yeah, but that will just drop frames... do you know why it doesn't drop any frames with CoreAVC but doesn't play back smoothly with FFdshow?

@clsid
Actually, I was able to fix this issue by choosing ffmpeg-mt instead of libavcodec in ffdshow's video decoder configuration. Performance wise, it's now as good as when I use CoreAVC.
Was I suppose to always use ffmpeg-mt? Like, is it suppose to be that much better than libavcodec?

[nm]

Was I suppose to always use ffmpeg-mt? Like, is it suppose to be that much better than libavcodec?

ffmpeg-mt is an experimental branch of libavcodec that includes frame-based multithreading. It's faster if you have a multicore CPU.

[XhmikosR]

Can someone share their opinion on the mingw libs/boost update (http://forum.doom9.org/showthread.php?p=1396238#post1396238) and the libsamplerate update (http://forum.doom9.org/showthread.php?p=1396106#post1396106)?

[adam777]

@albain
Rev. 3398 solves the word wrapping problem with multi-lines subtitles, which also solves the Hebrew subtitles problem (it was probably the same problem all along).
However, the subtitles still disappear on seek (see http://forum.doom9.org/showthread.php?p=1396538#post1396538).
All the best, Adam.

[tetsuo55]

Can someone share their opinion on the mingw libs/boost update (http://forum.doom9.org/showthread.php?p=1396238#post1396238) and the libsamplerate update (http://forum.doom9.org/showthread.php?p=1396106#post1396106)?commit please :)

[markanini]

Can someone share their opinion on the mingw libs/boost update (http://forum.doom9.org/showthread.php?p=1396238#post1396238) and the libsamplerate update (http://forum.doom9.org/showthread.php?p=1396106#post1396106)?

Re: libsamplerate update:
YAY!:)

[fastplayer]

Can someone share their opinion on the mingw libs/boost update (http://forum.doom9.org/showthread.php?p=1396238#post1396238) and the libsamplerate update (http://forum.doom9.org/showthread.php?p=1396106#post1396106)?
I can't comment on MinGW but Boost should only be updated to the latest stable release 1.42.0, not to the latest from trunk. We don't really have to play beta tester for the Boost library especially since Haruhiko - who integrated it into ffdshow - is not around anymore...
As for libsamplerate, I'm against the update. The "price" is too high IMO.

[XhmikosR]

Find which rev exactly is the boost 1.42 release and I can happily update to that. I cannot find out the exact revision number of the 1.42 release that's why I just updated to r61712 taken from here http://svn.boost.org/svn/boost/branches/release

[fastplayer]

Find which rev exactly is the boost 1.42 release and I can happily update to that. I cannot find out the exact revision number of the 1.42 release that's why I just updated to r61712 taken from here http://svn.boost.org/svn/boost/branches/release
You're right, rev61724 is about to get released as 1.43.0:
https://svn.boost.org/trac/boost/browser/branches/release

[XhmikosR]

OK, I'll update to that revision.

EDIT: I'm actually getting a new warning:
Warning 170 warning C4512: 'boost::detail::future_object_base::relocker' : assignment operator could not be generated h:\progs\ffdshow_boost\src\boost\thread\future.hpp 143 ff_vfw
Other than that I had to add BOOST_NO_RTTI in the Preprocessor definitions for ff_vfw in order to get the ICL11 x86 build to compile.

[fastplayer]

Can you try with 1.42.0?
http://sourceforge.net/projects/boost/files/boost/1.42.0/boost_1_42_0.7z/download

[XhmikosR]

Exactly the same thing, same warning and I also need to add BOOST_NO_RTTI in the Preprocessor definitions for ff_vfw ICL11.

[albain]

@albain
Rev. 3398 solves the word wrapping problem with multi-lines subtitles, which also solves the Hebrew subtitles problem (it was probably the same problem all along).
However, the subtitles still disappear on seek (see http://forum.doom9.org/showthread.php?p=1396538#post1396538).
All the best, Adam.

Very interesting news, this means (if confirmed) that we can drop the subviewer (srt) engine to use only the SSA one.

About the remaining problem, you were too fast, I fixed this in revision after it (3399). Please tell me if this is fixed for you

Otherwise, there is an annoying problem with ffdshow dialogs : the apply button is ungreyed as soon as you select another dialog (with no modifications in the old one).
I started to look after this issue with no luck for now

@developers
I tried to build an ICL11 version of ffdshow 64 bits with no luck. It seems that MMX intrinsics are not compatible. The _m64 register datatype seems to be faulty. This means that we have to move forward to SSE2 instructions or higher, but apart the big refactoring work, what about AMD intrinsics ?

[adam777]

Very interesting news, this means (if confirmed) that we can drop the subviewer (srt) engine to use only the SSA one.

About the remaining problem, you were too fast, I fixed this in revision after it (3399). Please tell me if this is fixed for you


Confirmed.
All is well now with Rev. 3402 :)
Thanks again for all your hard work, glad I could be of some assistance :thanks:

[albain]

Confirmed.
All is well now with Rev. 3402 :)
Thanks again for all your hard work, glad I could be of some assistance :thanks:

Good, even with "support for SSA tags..." checked ?

[fastplayer]

Otherwise, there is an annoying problem with ffdshow dialogs : the apply button is ungreyed as soon as you select another dialog (with no modifications in the old one).
I started to look after this issue with no luck for now
I tracked it down to rev3377. That's when the issue started appearing.

Maybe this change in TffdshowBase.cpp (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/src/TffdshowBase.cpp?r1=3377&r2=3376&pathrev=3377)?

[adam777]

Good, even with "support for SSA tags..." checked ?

Yep. :)

[albain]

I tracked it down to rev3377. That's when the issue started appearing.

Maybe this change in TffdshowBase.cpp (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/src/TffdshowBase.cpp?r1=3377&r2=3376&pathrev=3377)?

Nice catch, besides it was my changes :thanks:

[fastplayer]

Au contraire! Thank you for your diligence!
I guess working on subtitles is not exactly the most exciting part about ffdshow which in turn makes your contributions even more appreciated :)

[Sebastiii]

+1 :)

So cool :)

Damien, i send you a note :)
Cheers,
Seb.

[STaRGaZeR]

So many fixed things.

/me likes.

[jeremy33]

Hi,

I have a problem with this file http://www.mediafire.com/?mzaj2zm4jxw and ffdshow resizer.

The aspect ratio isn't correct. It must be 16/9 (1.78) and it is like 16/6 (2.60).

Without resizer it's good.

[hoborg]

Hi.
From MPC-HC:

Revision 1820 - Directory Listing
Modified Wed Apr 28 23:55:20 2010 UTC (6 days, 6 hours ago) by Aleksoid

Correct : Level@5.1 check for Ati(remove check for video width);
Add : MPCVIdeoDecoder, disable DXVA(H.264) for SD video(width < 1280);

Revision 1819 - Directory Listing
Modified Wed Apr 28 21:21:42 2010 UTC (6 days, 9 hours ago) by Spec-Chum

Added H.264 L5.1 DXVA support for ATI cards with drivers > v8.14.1.6105 (Catalyst 10.4)

Can this be inplemented in to FFDshow DXVA decoder too? (compatibility check for ATI card)

[fastplayer]

In rev3401 a new option has been added:
"Override those shadow settings (ignore SSA shadow settings)"
Which settings exactly? All of em (including font size, color etc.) or just the "shadow mode" itself?

[albain]

In rev3401 a new option has been added:
"Override those shadow settings (ignore SSA shadow settings)"
Which settings exactly? All of em (including font size, color etc.) or just the "shadow mode" itself?

Only shadow mode & size for now

Hi.
From MPC-HC:

Revision 1820 - Directory Listing
Modified Wed Apr 28 23:55:20 2010 UTC (6 days, 6 hours ago) by Aleksoid

Correct : Level@5.1 check for Ati(remove check for video width);
Add : MPCVIdeoDecoder, disable DXVA(H.264) for SD video(width < 1280);

Revision 1819 - Directory Listing
Modified Wed Apr 28 21:21:42 2010 UTC (6 days, 9 hours ago) by Spec-Chum

Added H.264 L5.1 DXVA support for ATI cards with drivers > v8.14.1.6105 (Catalyst 10.4)

Can this be inplemented in to FFDshow DXVA decoder too? (compatibility check for ATI card)

Done

[hoborg]

Done

:thanks:

[fastplayer]

Only shadow mode & size for now
Then I suggest renaming this option to
"Override shadow mode and size (ignore SSA shadow settings)".
This should make it clearer what is actually being overridden.

After a quick test of SRT subtitle handling, I can say that it is looking very good: 2-line subs, wrapping, switching and seeking are all working without any fuss. I'll test later more thoroughly. :cool:

[STaRGaZeR]

@albain

More subs fixes:

- Fix for SSA position tags in SRT subs.
- Fix for line wrapping when position tags are used inside SRT subs.
- Fix for incorrect colors when using the \xc tag.

Sebastii uploaded a nice SRT script full of SSA tags, it's great for testing: http://forum.doom9.org/showthread.php?p=1396430#post1396430

Patch: http://www.mediafire.com/?tgxvij5nyto
Build: http://www.mediafire.com/?jrmnwu0munm

Guys, please test and report if you find any bugs. You need "Process SSA tags in text subtitles" checked. I haven't found any problems with SRT/SSA/ASS.

[albain]

@albain

More subs fixes:

- Fix for SSA position tags in SRT subs.
- Fix for line wrapping when position tags are used inside SRT subs.
- Fix for incorrect colors when using the \xc tag.

Sebastii uploaded a nice SRT script full of SSA tags, it's great for testing: http://forum.doom9.org/showthread.php?p=1396430#post1396430

Patch: http://www.mediafire.com/?tgxvij5nyto
Build: http://www.mediafire.com/?jrmnwu0munm

Guys, please test and report if you find any bugs. You need "Process SSA tags in text subtitles" checked. I haven't found any problems with SRT/SSA/ASS.

Thank you, I'll commit your patch. Do you have a sourceforge user ?

@clsid : could you add commit rights to stargazer ?

[clsid]

Done!
(I assume 'stargazer' is the correct SF.net user ID?)

[STaRGaZeR]

stargazer69 actually, plain stargazer was already in use :p

Thanks!

[fastplayer]

^Done! :)

[horvathd]

I noticed some time ago that when I start play a video using ffdshow and I right-click on the video decoder icon on the tray then among the filters one is called is GradFun. But when I open the configuration dialog with a double click the the same filter is called DeBand. After I close the configuration window and right-click again on the icon then the filter name remains DeBand.

I hope it's understandable what I trying to say.

PS.: I'm using now the rev3402 of ffdshow.

[fastplayer]

^Yep, confirmed with rev3406.

[rpm7200]

libdts' s name changed libdca. but ffdshow is showing libdts in info and cpu tab. sorry bad very bad english.

[Sebastiii]

Hi Albain,

I have try last SVN 3416 and the sub isn't good,
This is 2 screenshot @ 18 seconds and 23 seconds.

The sub are cut and some like @ 12 and 17 seconds are not displayed (That is says ?).

Thanks :) a little more fix and go :)
Bon courage :)

[fastplayer]

The "Reload" button in the subtitle options doesn't seem to work. With "Watch for file changes" disabled, a during playback modified subtitle doesn't refresh/reload when clicking the button.

@Sebastiii:
http://fr.wikipedia.org/wiki/Compression_d'image

[Sebastiii]

Hi fastplayer :)

Thank you :) i didn't see that png it's non compressed image :( i think idiot and i'm thinking it's like jpg.
I have change and now it's better thank you again :)
Seb.

[fastplayer]

Hi fastplayer :)

Thank you :) i didn't see that png it's non compressed image :( i think idiot and i'm thinking it's like jpg.
I have change and now it's better thank you again :)
Seb.
PNG uses lossless compression whereas JPEG is lossy.
If you want to reduce the size of the PNG file, you can use an optimizer like OptiPNG (http://optipng.sourceforge.net/). The image will still retain its quality but redundant data will be stripped out.

As for posting screenshots in forums, I suggest using thumbnails because this doesn't break forum layout and keeps the bandwidth use down to a minimum :)

[fastplayer]

Here's a tricky one:
ffdshow randomly loads one of my CPU cores with 100%. Unfortunately, this is not easily reproducible but I noticed that it happens when SRT subtitles are involved (maybe even with other formats). The issue occurs with AVI, MP4 and MKV files.
Here's a screenshot of Process Explorer showing that ffdshow.ax is the culprit:
http://i43.tinypic.com/111rupg.png

My setup:
Athlon64 X2 3800+
Radeon HD4670 (CAT10.4)
Windows 7 32-bit
ffdshow 3415, YV12 output (happens with HQRGB32 too)
MPC-HC 1856, with ffdshow audio/video decoder set up as external filters
madVR v0.12 (happens with EVR-CP/VMR-9 too)
Haali Media Splitter 1.10.120.15 (happens with MPC's internal splitters too)

PS:
If necessary, I can upload a dump of mpc-hc.exe.

[Sebastiii]

PNG uses lossless compression whereas JPEG is lossy.
If you want to reduce the size of the PNG file, you can use an optimizer like OptiPNG (http://optipng.sourceforge.net/). The image will still retain its quality but redundant data will be stripped out.

As for posting screenshots in forums, I suggest using thumbnails because this doesn't break forum layout and keeps the bandwith use down to a minimum :)

Thanks :)

[moonrak]

My first question in this forum.
Where do you guys download versions newer than the ones posted in "http://www.xvidvideo.ru/ffdshow-tryouts-project-x86-x64/"?
I see that some here have rev 3416 while in xvidvideo the latest one is rev 3415.
:)
moonrak

[clsid]

The official site:
http://sourceforge.net/projects/ffdshow-tryout/files/

[STaRGaZeR]

fastplayer, do you know in which revision the problem started?

[Sebastiii]

Here's a tricky one:
ffdshow randomly loads one of my CPU cores with 100%. Unfortunately, this is not easily reproducible but I noticed that it happens when SRT subtitles are involved (maybe even with other formats). The issue occurs with AVI, MP4 and MKV files.
Here's a screenshot of Process Explorer showing that ffdshow.ax is the culprit:
http://i43.tinypic.com/111rupg.png

My setup:
Athlon64 X2 3800+
Radeon HD4670 (CAT10.4)
Windows 7 32-bit
ffdshow 3415, YV12 output (happens with HQRGB32 too)
MPC-HC 1856, with ffdshow audio/video decoder set up as external filters
madVR v0.12 (happens with EVR-CP/VMR-9 too)
Haali Media Splitter 1.10.120.15 (happens with MPC's internal splitters too)

PS:
If necessary, I can upload a dump of mpc-hc.exe.

Did you try with graphstudio too ?
i didn't remark this but i look at to see too :)
Seb.

[fastplayer]

fastplayer, do you know in which revision the problem started?
I wish I knew. :(
I think it started around the time when Haruhiko made some architectural changes to the subtitle engine (a year ago?). The issue was so rare that I didn't bother to report it but in the past 6-8 weeks the frequency, with which this issue occurs, increased. Now it's downright annoying...
Unfortunately, trying to reproduce it, is pure coincidence. A video+sub combo that exhibits this issue one time, won't next time.
Another interesting observation I made when investigating it, was that when it occurs and I start Process Explorer, CPU load would drop to normal levels. At first, that totally peed me off because I couldn't pinpoint the reason for the unusual CPU load. So I made it an habit to start Process Explorer before playing a video. :D

Anyway, I'll take a look at the dump file with MS's Debug Tools. Maybe something enlightening will come up *keeping fingers crossed*

By the way, what happened to parsing of HTML tags in SRT subs? What setting needs to be enabled?

[fastplayer]

Did you try with graphstudio too ?
i didn't remark this but i look at to see too :)
Seb.
How can this help me?
I can certainly rule out other interfering filters. It's just MPC and ffdshow. No interfering virus/spyware scanners, no fancy shell extensions etc.
Process Explorer clearly shows that ffdshow.ax is the cause of this abnormal CPU load.
I'm on Windows 7 now but it happened in XP too. Different renderers (madVR, VMR7/9, EVR-CP) as well as GPUs (X800 GTO, HD4670) all don't seem to matter, too.

[Sebastiii]

How can this help me?
I can certainly rule out other interfering filters. It's just MPC and ffdshow. No interfering virus/spyware scanners, no fancy shell extensions etc.
Process Explorer clearly shows that ffdshow.ax is the cause of this abnormal CPU load.
I'm on Windows 7 now but it happened in XP too. Different renderers (madVR, VMR7/9, EVR-CP) as well as GPUs (X800 GTO, HD4670) all don't seem to matter, too.

Maybe to force to reproduce it :)
But yes hard to nail it :(
Seb.

[STaRGaZeR]

I wish I knew. :(
I think it started around the time when Haruhiko made some architectural changes to the subtitle engine (a year ago?). The issue was so rare that I didn't bother to report it but in the past 6-8 weeks the frequency, with which this issue occurs, increased. Now it's downright annoying...
Unfortunately, trying to reproduce it, is pure coincidence. A video+sub combo that exhibits this issue one time, won't next time.
Another interesting observation I made when investigating it, was that when it occurs and I start Process Explorer, CPU load would drop to normal levels. At first, that totally peed me off because I couldn't pinpoint the reason for the unusual CPU load. So I made it an habit to start Process Explorer before playing a video. :D

Anyway, I'll take a look at the dump file with MS's Debug Tools. Maybe something enlightening will come up *keeping fingers crossed*

By the way, what happened to parsing of HTML tags in SRT subs? What setting needs to be enabled?

Can't help you with that, I hope that you can at least fin the cause.

The parsing of HTML tags is always enabled now, that's why the option was removed. However ffdshow still checks for the option before doing the actual parsing, this check should be removed because some people had it disabled and can't activate it now that the option is missing :D

[hoju3508]

Has this issue been addressed?

:thanks:

Image corruption while using ffmpeg-mt with more than 1 thread. (libavcodec and DVXA are also ok)
http://img265.imageshack.us/img265/5303/avatarblurayffmpegmtcor.png

Tested with ffdshow r3370 xxl and r3390 from http://www.xvidvideo.ru

This must be some very old bug because it occurs even in very old revisions as well.

Remuxed BD sample http://www.sendspace.com/file/f9s4te

[fastplayer]

The parsing of HTML tags is always enabled now, that's why the option was removed. However ffdshow still checks for the option before doing the actual parsing, this check should be removed because some people had it disabled and can't activate it now that the option is missing :D
Put it on your ToDo list because I can't find this reg key :D

[fastplayer]

Has this issue been addressed?

:thanks:
Nope but it has been reported upstream (http://forum.doom9.org/showthread.php?t=154365).

[STaRGaZeR]

Put it on your ToDo list because I can't find this reg key :D

Try with this build: http://www.mediafire.com/?ny42mtiyuyt

I assume you had it deactivated, this one doesn't perform any checks. It should work.

[fastplayer]

Try with this build: http://www.mediafire.com/?ny42mtiyuyt

I assume you had it deactivated, this one doesn't perform any checks. It should work.
Yep, it works now :)

Anyway, trying to debug the dump file was a waste of time:
- I need symbol files for all EXE/DLL files involved.
- mpc-hc.exe didn't crash and is not the root cause but ffdshow is running "inside" of MPC's process... *argh*

Any other ideas are welcome!

[adam777]

Hello all,
Some annoying problem I've noticed with rev. 3421 (probably has something to do with the SSA engine being the default one now) - subtitles tags inside srt subtitles are displayed as simple text, and not being translated.
That is, the following subtitle:
1
00:00:00,530 --> 00:00:02,610
<i>Blah Blah Blah</i>

is displayed as:
<i>Blah Blah Blah</i>

Any setting I might have missed?
Thanks, Adam.

[STaRGaZeR]

Try r3422 :D

[moonrak]

Try r3422 :D

Where is r3422?
moonrak

[adam777]

Try r3422 :D

Oh... look at that, right in front of my eyes in that SVN log :thanks:

@moonrak - http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1

[moonrak]

Oh... look at that, right in front of my eyes in that SVN log :thanks:

@moonrak - http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1

So we have the "official" site, the russian (xvidvideo.ru) and mediafire also!
Anywhere else you would like to confess?
:)
moonrak

[Sebastiii]

Damien :) you're the best :)

All subtitle BD and .sup it's working very great en rev 3425

Thank you so much :)
Cheers,
Seb.

:thanks: :thanks:

[STaRGaZeR]

Only one thing missing here, everything else is fixed!

Original, 1920x1080:

http://thumbnails14.imagebam.com/7976/e8084c79757179.gif (http://www.imagebam.com/image/e8084c79757179)

Resized, 1680x944:

http://thumbnails13.imagebam.com/7976/da200a79757178.gif (http://www.imagebam.com/image/da200a79757178)

Position is correct, size is not. Almost there!

[STaRGaZeR]

Guys, I need you opinion here.

Currently the autosize function in the subtitles filter does indeed autosize the subs, but based on a formula that doesn't make any sense to me. For me autosize means same perceived size regardless of resolution. It seems VSFilter agrees with me. I've made a new version that does that: if autosize is checked you'll see constant sized subs, regardless of resolution, like you do with VSFilter/MPC. If you want the subs to look the same in both, install this build and change these settings:

- Placement - vertical position --> set it to 93%
- Font --> enable autosize, change shadow mode to classic shadow, set font size to 18. I've not determined VSFilter's outline or shadow size yet, so help me with this.

My goal is that people activating ffdshow subtitle filter gets the same experience, using default settings, compared to VSFilter/MPC. The above settings will be the default ones if you like this change.

Build: http://www.mediafire.com/?2injn40wynd

[Sebastiii]

Hi,

It's a good idea and try it :)
i think that is the good way to do :)
Seb.

[Sebastiii]

Great, it works very good :)

So this is my setup :

http://thumbnails3.imagebam.com/7978/f5f75079770675.gif (http://www.imagebam.com/image/f5f75079770675)

Outline : 2
Shadow : 3

I have see only a thing, is that on .srt (legend of..) @ 12 seconds
In MPC-HC i have 4 lines and here 3 (the middle line is long but we can read all the words).

http://thumbnails6.imagebam.com/7978/5aa92879771449.gif (http://www.imagebam.com/image/5aa92879771449)

:)

[STaRGaZeR]

Yup, ffdshow uses smaller borders for wrapping calculation, that's in the TODO list too :D

[fastplayer]

For me autosize means same perceived size regardless of resolution.
You mean that the size stays "constant" (independent of resolution) when, for example, switching between SD and HD, right?

[dann23]

You mean that the size stays "constant" (independent of resolution) when, for example, switching between SD and HD, right?

+1. I'm tired to change font size when watching SD and HD

[Sebastiii]

Hi,

Maybe i misunderstood, but that i understand is that the size stay constant. it's scale the size of sub like you have setup and if it's HD or SD, it keep to have the same overview on all movies

By my test, this is that i can see :)

But if a more explanation can be made, i'm ok :)
Seb.

[fastplayer]

Well, then I'm all in favor of making this change.

What about the situation when you maximize/full-screen an SD video and the subs "maximize" themselves too? They get all blurry and ugly. Can they be made sharper? Reducing the font size - to mitigate the bigger size - causes even more aliasing and makes the sub unwatchable.

[horvathd]

Well, then I'm all in favor of making this change.

What about the situation when you maximize/full-screen an SD video and the subs "maximize" themselves too? They get all blurry and ugly. Can they be made sharper? Reducing the font size - to mitigate the bigger size - causes even more aliasing and makes the sub unwatchable.

My solution for that is to set ffdshow filter to resize to the video to my screen resolution, then apply the subtitle filter. This way the fonts is alway sharp at full screen and also don't show any problem when watching a video in a window. The drawback is the higher CPU usage.

[DJ_Phatic]

I am not sure if i understand it correctly.
Default FFDShow activation keys are CTRL+ALT + mapped key.
For example CTRL+ALT+O to show/hide OSD, CTRL+ALT+S to show/hide subtitles. I have no problems with this (working fine).
But CTRL+ALT+F4 or CTRL+ALT+Nunpad1 simply does nothing for me, i tryed remap CTRL+ALT+F4 to CTRL+ALT+K, but still nothing (is there some OSD message about strem swithched?)

Here is my settings: (audio switcher/subtitles is enabled, testing Samurai Champaloo sample)
http://hobring.esero.net/saf/ffdshow/streams.png

I would also like to raise this issue. The next subtitle stream and audio stream shortcuts do not work with ffdshow DXVA, same settings as above though without focus is ticked.

[horvathd]

I noticed that the Overlay filter can't be hidden in the "Show / hide filters" option with rev 3406.

[namaiki]

My solution for that is to set ffdshow filter to resize to the video to my screen resolution, then apply the subtitle filter. This way the fonts is alway sharp at full screen and also don't show any problem when watching a video in a window. The drawback is the higher CPU usage.

How can you set it so that it upscales to around desktop resolution, but doesn't add black bars, and keeps aspect ratio of the video?

[horvathd]

My setting are:

http://i524.photobucket.com/albums/cc329/davhorv112/Misc/ffdshow_resize.jpg

The Resize -> Borders, and Resize -> Settings are the defaults.

[Sebastiii]

My solution for that is to set ffdshow filter to resize to the video to my screen resolution, then apply the subtitle filter. This way the fonts is alway sharp at full screen and also don't show any problem when watching a video in a window. The drawback is the higher CPU usage.

Thanks for the tip :)
I didn't set resize before sub and now whaouuuu :) it's really clear :)
Seb.

[pankov]

The resize solution is working great in software mode but if I'm not mistaken it's not supported in the DXVA version, because the decoded frame is never returned to the main memory.
I was wondering though what will happen if the subtitles are always rendered at screen resolution and then mixed with the video frame which is also resized to screen resolution but by the video card. .... now the question is - which happens first - the mixing or the resizing of the video frame?

[namaiki]

in dxva if you are using ffdshow dxva's subtitle renderer, then the mixing is done first; ffdshow dxva video doesn't have a resizing option. if you're using mpc-hc's internal subtitle renderer instead, then the subs are scaled separately.. but that is different, someone might complain :P

[Jeremy Duncan]

In post # 11600 I complained that the menu's in mpeg 2 weren't showing up like they did when leaks patch was applied. But I tried a different dvd and it worked and with a different avisynth code it works showing the root menu in every movie, so the mistake of the root menu not showing is nothing to be worried about, ffdshow behaves like it did before.

In the svn trunk revision 3358 tal_aloni allpied a patch from sET and me, but when I tried a build in sourceforge from xxl_ it didn't work properly as if the patch wasn't applied, so I made a patch in tortoise svn and looked at the difference in the code and then made a build so the patch was the same as the one I posted in this thread before and the build worked as expected, so the difference in code tal_aloni made to the patch broke it, so i will post the patch that fixes revision 3358 and hope a developer can commit it to the svn trunk asap? :-/ Here is the patch and build link (http://www.mediafire.com/?imrh4wdz4wm)

[pankov]

namaiki,
I'm using Zoomplayer so mpc-hc's render is out of my reach. I guess I'm stuck either with bad subtitles on SD/720p content or with non DXVA decoding, right?

[namaiki]

namaiki,
I'm using Zoomplayer so mpc-hc's render is out of my reach. I guess I'm stuck either with bad subtitles on SD/720p content or with non DXVA decoding, right?

What's your CPU? I would prefer quality over DXVA any day.

Maybe Zoomplayer has it's own subtitle renderer somehow?

[STaRGaZeR]

You mean that the size stays "constant" (independent of resolution) when, for example, switching between SD and HD, right?

Exactly. When resolution is higher, ffdshow adapts the font size so the subs look the same.

Well, then I'm all in favor of making this change.

What about the situation when you maximize/full-screen an SD video and the subs "maximize" themselves too? They get all blurry and ugly. Can they be made sharper? Reducing the font size - to mitigate the bigger size - causes even more aliasing and makes the sub unwatchable.

We can't do anything in this case, ffdshow prints the subs to the video, if the video is SD you'll get SD subs no matter what. This is the real advantage of MPC over VSFilter too, it prints the subs to the screen (in the renderer) instead of the video. The only solution as others have said is to resize to screen resolution first, then print the subs.

My solution for that is to set ffdshow filter to resize to the video to my screen resolution, then apply the subtitle filter. This way the fonts is alway sharp at full screen and also don't show any problem when watching a video in a window. The drawback is the higher CPU usage.

Thanks for the tip :)
I didn't set resize before sub and now whaouuuu :) it's really clear :)
Seb.

Yeah, I've been using this "trick" since forever, subs look great :)

So you guys like the new autosize option? What about the shadow and outline settings to match VSFilter? Don't use MPC for this, it blurs the subs and you can't change that.

[pankov]

What's your CPU? I would prefer quality over DXVA any day.

Maybe Zoomplayer has it's own subtitle renderer somehow?

my CPU is Q9400 @ 3.2GHz so I don't have any problems with software decoding ... at least not yet - we'll see what will happen with BD 3D. ... so I too prefer quality

And about Zoomplayer - sadly it doesn't have any subs renderer - it relies only on directshow filters

[fastplayer]

What about the shadow and outline settings to match VSFilter? Don't use MPC for this, it blurs the subs and you can't change that.
Dunno about VSFilter because I use ffdshow for subs exclusively. My settings are: gradient shadow, outline=2, shadow=3, weight=medium, size=22, placement_vertical=98%.
As for outline and shadow size, these are the only settings that make sense to me and to Sebastiii apparently (http://forum.doom9.org/showthread.php?p=1398424#post1398424) too :)
Both values - unlike the default - are subtle and give the subs a certain "depth".

[Sebastiii]

Yeah, I've been using this "trick" since forever, subs look great :)

So you guys like the new autosize option? What about the shadow and outline settings to match VSFilter? Don't use MPC for this, it blurs the subs and you can't change that.

I'm ok on new autosize :)
Seb.

[magic144]

really happy to see albain's progress...

this is the latest comment I have updated to the bugtracker re: VobSub subs muxed (embedded) in 720p mkv...
https://sourceforge.net/tracker/?func=detail&aid=2981790&group_id=173941&atid=867360

Build 3425.
Rendering and placement seems OK, but disappearing subs still happens - seems to be on seeking to an 'inopportune' timecode, or if subs are absent for a while (i.e. noone in the video is talking), they seem to stay off.

As ever, I can provide subtitle data if it would help.

[Sebastiii]

Hi,
Can you provide the file :)

All latest version 3425, i try to reproduce your pb and i didn't -> so it's pretty cool, but i want to try with your file :)
Seb.

[Taurus]

This is what I got with rev3422_20100508 from XvidVideo.RU:
http://img10.imageshack.us/img10/9885/clickl.png

Rev3406_20100505 is doing fine.

Player: Mediaplayer Classic HC rev1.3.1862.0.x86

[DigitalDeviant]

Subtitle and menu buttons are not displayed correctly in that version. It has been fixed by 3425.

[STaRGaZeR]

So here it is. In this build:

- With autosize enabled:

Font size is constant across resolutions. You can see small variations between very different resolutions like 640x360 and 1920x1080, but this is normal and matches VSFilter.
Outline is not constant across resolutions (no changes here).
Shadow is semi constant across resolutions, it'll be bigger at low res.
Changed default values to render subs like VSFilter by default.

With all of the above ffdshow now looks like VSFilter. Please test and report your findings!

http://www.mediafire.com/?xtwzz2zbmhy

[Taurus]

Subtitle and menu buttons are not displayed correctly in that version. It has been fixed by 3425.
Thank you.
Will test it asap.
Edit:
Confirmed, menu buttons working again.

[Sebastiii]

So here it is. In this build:

- With autosize enabled:

Font size is constant across resolutions. You can see small variations between very different resolutions like 640x360 and 1920x1080, but this is normal and matches VSFilter.
Outline is not constant across resolutions (no changes here).
Shadow is semi constant across resolutions, it'll be bigger at low res.
Changed default values to render subs like VSFilter by default.

With all of the above ffdshow now looks like VSFilter. Please test and report your findings!

http://www.mediafire.com/?xtwzz2zbmhy


Great have to test :)
Thank you,
Cheers,
Seb.

Edit: Works Great :) commit :)
Bye,
Seb.

[DeathWolf]

I saw that it was reported in the past already, but is there any chance that deband might be fixed for ffdshow x64?

[DigitalDeviant]

My original multiple subtitle sample plays fine now, a million thanks Albain. However, here is another problem sample I just ran into.

In addition to some fugly lines through the yellow subs at the bottom and fugly shadowing (see pic), the subtitle positions get switched around.

http://img442.imageshack.us/img442/6645/badsub.th.png (http://img442.imageshack.us/i/badsub.png/)

Sample (http://www.mediafire.com/download.php?mgztymdmdjn)

Positioning is fixed but there are some ugly lines through some subtitles and the subpic borders show some junk too. I notice it on other Bluray discs too.

Edit: It pappears with all usable scalers and with both RGB and YV12. Additionally the above sample shows a greenish tint on the subtitle outline with YV12 output.

[Delerue]

So here it is. In this build:

- With autosize enabled:

Font size is constant across resolutions. You can see small variations between very different resolutions like 640x360 and 1920x1080, but this is normal and matches VSFilter.
Outline is not constant across resolutions (no changes here).
Shadow is semi constant across resolutions, it'll be bigger at low res.
Changed default values to render subs like VSFilter by default.

With all of the above ffdshow now looks like VSFilter. Please test and report your findings!

http://www.mediafire.com/?xtwzz2zbmhy

Thanks a lot, man! It works flawlessly. Finally I don't need to use VSFilter anymore. :)

BTW, how can I use FFDShow subtitle extension with an ordinary DVD to avoid the ugly default font? I've already checked the combox 'DVD subtitles (vobsub)' and 'Close captions decondng', but the font remains the same old ugly thing. I've tried with 'The Matrix', the original first DVD version.

[clsid]

DVDs use VobSub subtitles, which are images, not text. So the font can't be changed.

[Delerue]

DVDs use VobSub subtitles, which are images, not text. So the font can't be changed.

I see. But there's a way to rip these images into a srt file (via OCR, I think), right? Do you think that FFDShow can do that in real time? Would be great.

[clsid]

I think our time could be better spend making other improvements.

[Simplex]

I want to bitstream only DTS core of DTS-MA track over SPDIF to my receiver on Win7 32bit. Using latest ffdshow I get only loud static and too fast movie playback. Perhaps I am missing something. Any help would be appreciated, if you need more info about my config, I will post it here.

[STaRGaZeR]

I see. But there's a way to rip these images into a srt file (via OCR, I think), right? Do you think that FFDShow can do that in real time? Would be great.

That would be a huge mess, just look at how OCR apps work, now imagine having to pause the movie for you to enter a letter that ffdshow doesn't recognize, over and over. Even if possible, it would cause more problems than it solves.

[SamuriHL]

I want to bitstream only DTS core of DTS-MA track over SPDIF to my receiver on Win7 32bit. Using latest ffdshow I get only loud static and too fast movie playback. Perhaps I am missing something. Any help would be appreciated, if you need more info about my config, I will post it here.

DTS core can be done. DTS-HD MA can't be bitstreamed over S/PDIF. If ffdshow is attempting to then it could be causing your problems and I'd consider it a bug.

[Delerue]

I think our time could be better spend making other improvements.

That would be a huge mess, just look at how OCR apps work, now imagine having to pause the movie for you to enter a letter that ffdshow doesn't recognize, over and over. Even if possible, it would cause more problems than it solves.

OK. You've convinced me. :)

Another suggestion: if you install a new CPU with different number of cores/threads and SIMD instructions, FFDShow won't recognize; instead, it'll keep the old values. So you have to set the new values manually or reinstall FFDShow. Wouldn't be safer if it detects these values on-the-fly? Or at least create a button inside the 'Video decoder configuration' dialog to auto-detect these informations. What you think?

[Keiyakusha]

Delerue
It sounds like you changing your CPU twice in a week :p

[Mr VacBob]

Looking at that Avatar sample with the mplayer demuxer (fails and looks the same as ffdshow) with FF_DEBUG_PICT_INFO|FF_DEBUG_STARTCODE:


[h264 @ 0x100b132a0]- packet start
[h264 @ 0x100b132a0]NAL 9 at 3/284946 length 1
[h264 @ 0x100b132a0]NAL 6 at 9/284946 length 10
[h264 @ 0x100b132a0]non-existing SPS 32 referenced in buffering period
[h264 @ 0x100b132a0]NAL 1 at 24/284946 length 66536
[h264 @ 0x100b132a0]slice:1 F mb:0 P fix pps:0 frame:35 poc:108/108 ref:1/1 qp:26 loop:0:0:0 weight:0
[h264 @ 0x100b132a0]NAL 1 at 66565/284946 length 73850
[h264 @ 0x100b132a0]slice:2 F mb:2040 P fix pps:0 frame:35 poc:108/108 ref:1/1 qp:8 loop:0:0:0 weight:0
[h264 @ 0x100b132a0]NAL 1 at 140420/284946 length 85437
[h264 @ 0x100b132a0]slice:3 F mb:4080 P fix pps:0 frame:35 poc:108/108 ref:1/1 qp:3 loop:0:0:0 weight:0
[h264 @ 0x100b132a0]NAL 1 at 225862/284946 length 58988
[h264 @ 0x100b132a0]slice:4 F mb:6120 P fix pps:0 frame:35 poc:108/108 ref:1/1 qp:3 loop:0:0:0 weight:0
[h264 @ 0x100b132a0]NAL 8 at 284858/284946 length 86
[h264 @ 0x100b132a0]pps:0 sps:0 CABAC slice_groups:1 ref:1/1 qp:26/26/-9/-3 LPAR 8x8DCT
[h264 @ 0x100b132a0]- packet end
[h264 @ 0x100b132a0]- packet start
[h264 @ 0x100b132a0]NAL 9 at 3/143772 length 1
[h264 @ 0x100b132a0]NAL 6 at 9/143772 length 10
[h264 @ 0x100b132a0]non-existing SPS 32 referenced in buffering period
[h264 @ 0x100b132a0]NAL 1 at 25/143772 length 30506
[h264 @ 0x100b132a0]slice:1 F mb:0 B fix pps:0 frame:36 poc:104/104 ref:1/1 qp:26 loop:0:0:0 weight:2 SPAT
[h264 @ 0x100b132a0]NAL 1 at 30536/143772 length 40103
[h264 @ 0x100b132a0]slice:2 F mb:2040 B fix pps:0 frame:36 poc:104/104 ref:1/1 qp:10 loop:0:0:0 weight:2 SPAT
[h264 @ 0x100b132a0]NAL 1 at 70644/143772 length 45570
[h264 @ 0x100b132a0]slice:3 F mb:4080 B fix pps:0 frame:36 poc:104/104 ref:1/1 qp:6 loop:0:0:0 weight:2 SPAT
[h264 @ 0x100b132a0]NAL 1 at 116219/143772 length 27459
[h264 @ 0x100b132a0]slice:4 F mb:6120 B fix pps:0 frame:36 poc:104/104 ref:1/1 qp:5 loop:0:0:0 weight:2 SPAT
[h264 @ 0x100b132a0]NAL 8 at 143684/143772 length 86
[h264 @ 0x100b132a0]pps:0 sps:0 CABAC slice_groups:1 ref:1/1 qp:26/26/-9/-3 LPAR 8x8DCT
[h264 @ 0x100b132a0]- packet end


In other words, the stream updates PPS frequently and the demuxer puts PPS/SPS at the end of a frame packet instead of the beginning of the next one. I think this verges on being a demuxer bug.

Would you mind running the h264 AVParser on TS packets before decoding? I think that would fix it, and any fix on the ffmpeg side would pretty much amount to running that. Well, or not starting the next thread until it checks for stray PPS, but I think that would effectively disable multithreading for this file.

[DeathWolf]

I saw that it was reported in the past already, but is there any chance that deband might be fixed for ffdshow x64?
Sorry for the bump, but I noticed one thing: if deband is turned on on the x64 version, then even paused video starts playing back.
Would deband get a constant stream of frames fed in an improper manner?

EDIT: also avisynth seems to crash in ffdshow x64

[STaRGaZeR]

EDIT: also avisynth seems to crash in ffdshow x64

Check your filters, TIVTC x64 from the Avisynth x64 thread crashes on me, Decomb works fine. It seems it's not a ffdshow issue.

[DeathWolf]

Right, it was an avisynth x64 issue. Still, deband does not exist in x64 avisynth and crashes as an ffdshow filter.

[STaRGaZeR]

In post # 11600 I complained that the menu's in mpeg 2 weren't showing up like they did when leaks patch was applied. But I tried a different dvd and it worked and with a different avisynth code it works showing the root menu in every movie, so the mistake of the root menu not showing is nothing to be worried about, ffdshow behaves like it did before.

In the svn trunk revision 3358 tal_aloni allpied a patch from sET and me, but when I tried a build in sourceforge from xxl_ it didn't work properly as if the patch wasn't applied, so I made a patch in tortoise svn and looked at the difference in the code and then made a build so the patch was the same as the one I posted in this thread before and the build worked as expected, so the difference in code tal_aloni made to the patch broke it, so i will post the patch that fixes revision 3358 and hope a developer can commit it to the svn trunk asap? :-/ Here is the patch and build link (http://www.mediafire.com/?imrh4wdz4wm)

Unless I'm missing something your patch doesn't change anything: the code is exactly the same.

Right, it was an avisynth x64 issue. Still, deband does not exist in x64 avisynth and crashes as an ffdshow filter.

It doesn't crash, it's just slow, try it with a very low res video. There's a lot of assembly code in there optimized for x86 that isn't used in the x64 version, that's why it's slow as molasses.

[DeathWolf]

It doesn't crash, it's just slow, try it with a very low res video. There's a lot of assembly code in there optimized for x86 that isn't used in the x64 version, that's why it's slow as molasses.

Sorry I did not mean crash. However turning on deband here has ill effects:
1.play video with deband enabled
2.play/pause/play quickly enoug
3.Experience video continuing playback even though it is paused.
EDIT: I don't get it with graphedit, so it has to be something linked with mpc-hc, sorry about that.

On a side note, it is my impression(from the code of deband) that there is actually no non-asm path at all... so it shouldn't even be doing anything and not be terribly slow. Did I miss something?

[STaRGaZeR]

Sorry I did not mean crash. However turning on deband here has ill effects:
1.play video with deband enabled
2.play/pause/play quickly enoug
3.Experience video continuing playback even though it is paused.
EDIT: I don't get it with graphedit, so it has to be something linked with mpc-hc, sorry about that.

On a side note, it is my impression(from the code of deband) that there is actually no non-asm path at all... so it shouldn't even be doing anything and not be terribly slow. Did I miss something?

Yup, only mmx version of those 2 functions. I'm a complete noob at this, so just ignore everything I say about it :p. I've asked for the source code of GradFun2DB x64 in the Avisynth x64 thread to see if JoshyD has added a SSE2 version of those functions or something. In the meantime you can use GradFun2DB x64 and call it a day.

[Vincent Vega]

i'm on vista64, how can i get vdubmod to prefer ffdshow vfw over xvid, divx, helix and the like?

i already enabled all codecs in ffdshow vfw config, still vdub chooses helix for yv12.. i also tried modifying vidc.yv12 in registry to ff_vfw.dll, then vdubmod just crashes..

[clsid]

First of all, use VirtualDub instead of the ancient VirtualDubMod.

If you don't need those other codecs, then one way to force use of ffdshow is to remove the entries of those other codecs from the registry.

[Vincent Vega]

i just tried virtualdub 1.9.9, it uses "internal dib decoder" for yv12. how do i disable it and use ffdshow?

my other option would be to uninstall divx, xvid and helix, but i need them for other things..

[Keiyakusha]

i just tried virtualdub 1.9.9, it uses "internal dib decoder" for yv12. how do i disable it and use ffdshow?

my other option would be to uninstall divx, xvid and helix, but i need them for other things..

In preferences uncheck "directly decode uncompressed yuv sources". But why you want that? Internal VDub's filter is fine!

[Vincent Vega]

if i uncheck "directly decode" virtualdub 1.9.9 also uses helix decoder for yv12 just as vdubmod.

i need ffdshow because for mpeg2 streams (dvds) where color primaries are not explicitly specified the only option to figure out the truth is to try to decode using both 601 and 709 and compare same frame side by side, i can do that with ffdshow on my other machine.

[Keiyakusha]

Ahh i see. I don't know how to make ffdshow vfw to be preferred over helix since I'm not using vfw and never have more than one filter for something when they interfere like this. I won't gen into details here, but if you want opinion, then you doing this in a most weird way possible... And bt601/bt709, is not something that can be done only on decoding step

[Vincent Vega]

And bt601/bt709, is not something that can be done only on decoding step

could you kindly enlighten me of other opportunities to accomplish the same task?

[Keiyakusha]

could you kindly enlighten me of other opportunities to accomplish the same task?

Well DGIndex normally takes care of that. Also if colorimetry is not defined, normally any 720p and higher is 709 and anything lower is 601. But if you know you may need to change colorimetry or you want to compare 2 variants side by side or interleaved, avisynth with colormatrix plugin should do the job. Afaik there is x64 versions of both. And no need to mess with decoders and stuff.

[Vincent Vega]

so you are basically saying, if i want to "compare 2 variants side by side" i should actually do 2 encodes and compare the results?

i still like my solution better, it allows to do only 1 encode, once i know the right coefficients.

[Keiyakusha]

nono. You can open avisynth script in virtual dub. no need to do any encodes!
You maybe even don't need a plugin...
Mpeg2Source("<<source d2v here>>") #you may use another types of source filters. FFMS2 maybe, but DSS/DSS2 not recommended.
a=last.Converttorgb(matrix="<<chose matrix you want>>")
b=last.Converttorgb(matrix="<<chose another matrix you want>>")
Stackcvertical(a,b)
And open this in virtualdub.

Something like that.

EDIT: I'm sorry, but this is my last post on this topic. ffdshow guys will be angry for this offtop ^_^

[Vincent Vega]

dude, thanks a lot, that worked excellent for me. i should dedicate more time to learning avisynth. only stackhorizontal() seemed a bit more convenient. i'll add these 3 lines (with #) to my avisynth scratchpad in megui, so i can use it whenever i need to.

[Vincent Vega]

and also the original question still remains:

are there any other ways to get the system to prefer one vfw decoder over others? i mean apart from uninstalling everything else except that one decoder, or this vidc.xxx= editing which does not seem to work so well..

[clsid]

Windows will first try the VIDC.FOURCC entry. Then it will try the other ones.

I am not sure why "vidc.yv12=ff_vfw.dll" would fail. Perhaps you forgot to enable uncompressed video in ffdshow VFW options?

[Vincent Vega]

raw video is enabled in ffdshow vfw settings.

on this problem vista machine i have default vidc.yv12=yv12vfw.dll, i'm not sure which program could have installed it. and if i change that to ff_vfw.dll vdub(mod) crashes.

on the other xp computer, i have pretty much the same software config (also realproducer, helix producer) but there is no vidc.yv12 entry in the registry and no yv12vfw.dll installed. ffdshow vfw works fine both without vidc.yv12=.. or with vidc.yv12=ff_vfw.dll specified.

so guess i need to figure out where this yv12vfw.dll troublemaker comes from..

[clsid]

Your conclusion is wrong. yv12vfw.dll is not the troublemaker.

Have you tried removing the vidc.yv12 entry like I suggested before?
Doesn't VirtualDub show which module causes the crash?

[Vincent Vega]

i dont know what happened, now everything seems to be fine, and i haven't even rebooted the computer, it's been encoding this whole time.

for the same YV12 script today:
with vidc.yv12=yv12vfw.dll - vdub(mod) uses "Helix YV12"
with vidc.yv12=ff_vfw.dll - vdub(mod) uses ffdshow
with vidc.yv12 entry deleted - vdub(mod) uses "Helix I420"

and no more crashes. :confused:

[WorBry]

Quick question about the 'Curves' option in the FFDShow decoder 'Levels' dialogue. I'm wondering if, with YV12 input, the 'curve' is modified natively in YV12 or converted to RGB and back in the process. I'm thinking it is likely a port of VD_Gradation curves or GiCoCu, since it allows for import of Photoshop acv curves, so it's probably the latter, but just wanted to be sure. Thanks.

[Keiyakusha]

MPC-HC 1883, ffdshow (icl11) revisions after 3435 (last checked - 3440), raw video filter with postprocessing (different profiles for different cases) added to MPC-HC's external filters. When playback is started, after trying to press right mouse button or open dropdown menu "play" -> ffdshow.ax is crashing.
EDIT: ffdshow settings --> tray, dialog & paths --> add tray icon menu to player's stream/languages menu option is responsible for this crash.

[Joniii]

I'm having a weird bug in latest builds. It started to happen somewhere around build 3422 if I remember it correctly.

If I install ffdshow to default location everything works fine. If I install it to custom location for example (C:\Program Files\Windows Codecs\VideoFilter\ffdshow) I get problems in Windows 7 Media Center (WMP works fine).

I have video.m2ts and video.srt in 7MC movie library, and if I start the playback (haali, ffdshow icl11 DXVA+subtitles) it works as usual but if I stop the playback and resume it sometimes later it hangs. This only happens on new builds if installed on custom location.

[clsid]

Does this happen only in 7MC or also with WMP (or MPC)?

[Keiyakusha]

MPC-HC 1883, ffdshow (icl11) revisions after 3435 (last checked - 3440), raw video filter with postprocessing (different profiles for different cases) added to MPC-HC's external filters. When playback is started, after trying to press right mouse button or open dropdown menu "play" -> ffdshow.ax is crashing.
EDIT: ffdshow settings --> tray, dialog & paths --> add tray icon menu to player's stream/languages menu option is responsible for this crash.

I'm sorry if this known, but ffdshow r3444, mpc-hc r1896. Problem still there. MPC-HC crashes when this option enabled: Enabling option tray, dialog & paths --> add tray icon menu to player's stream/languages

[clsid]

I can confirm that bug. Crash also happens when right-clicking on the tray icons.

@devs
The regular video decoder and the raw video filter currently share their settings. Wouldn't it be better to separate that?

[Midzuki]

@devs
The regular video decoder and the raw video filter currently share their settings. Wouldn't it be better to separate that?

Yep, an independent video processor would be quite an improvement.

:goodpost:

[Keiyakusha]

@devs
The regular video decoder and the raw video filter currently share their settings. Wouldn't it be better to separate that?
+1 I want this since I switched to win7 from XP!

[fastplayer]

clsid, what's the current status of ffmpeg's AAC decoder?
Is it complete enough to be integrated into ffdshow?

[clsid]

clsid, what's the current status of ffmpeg's AAC decoder?
Is it complete enough to be integrated into ffdshow?
It is still a work in progress afaik. Most code to use it is already in place in ffdshow, so if you want to try it out, you should be able to get it to work.

[Mike89]

Why all ffdshow posts in one thread? Go to post one to see latest ffdshow and then have to go past 580 pages to see if there is any posted information about it.

How about getting modern and more efficient and make a separate section just for ffdshow, then within that have separate threads for each new version? I mean as much as is still going on with the development and new tryout versions coming out so often, it would be so much more user friendly to have it organized better here.

To my understanding, ffdshow tryouts was started from members here? That being said, anyone in the know would want to come here to find the latest skinny on the versions made. Sure would be better organized for those than one humungous thread.

Just a suggestion from a user.

[clsid]

You can click on the blue arrow icons on this forum to automatically go to the first unread post in a topic.

[Shankster]

You can also just bookmark this link:
http://forum.doom9.org/showthread.php?t=120465&goto=newpost

Only works if you are logged in afaik.

[DigitalDeviant]

There is always a search thread fucntion. Also there are forums on the ffdshow-tryouts website.

[Keiyakusha]

r3446... now it crashes even if that option is off...

[clsid]

@albain

Could you have a look at the crash that Keiyakusha reported? There basically is a crash when right-clicking on the tray icon when both ffdshow video decoder and the raw video decoder are in the same filter graph.

[clsid]

Revision 3448 now stores the RAW video decoder settings separately. Same applies for its subtitle variant.

[clsid]

Can anyone confirm whether DTS passthrough is broken (since the audio parser changes some time ago)?

[TFM_TheMask]

Can anyone confirm whether DTS passthrough is broken (since the audio parser changes some time ago)?

I can because I made the bugreport.:)

[rsd78]

Can anyone confirm whether DTS passthrough is broken (since the audio parser changes some time ago)?

Not sure if it's related but I noticed that I can't get ffdshow to pass the core DTS track of an mkv containing DTS-MA over spdif (DTS light goes on but no sound). Didn't have issue getting stereo sound from the DTS-MA track on another machine.

[zummi]

Hi guys,

got some trouble with subs here. I've searched the thread for references to my gfx card with no luck... Anyway, i just got a new laptop with an integrated Intel GMA 4500MHD. When running in DXVA-mode i get lots of artifacts around the subtitles (screenshot: http://i46.tinypic.com/s1tffo.png). Without DXVA enabled everything is OK. Running Win7x64, latest ffdshow x64, MPC-HCs audiorenderer and mkvsplitter.

The CPU is only a Pentium SU4100 1,3ghz dual core, so i really need to have dxva enabled for 1080p playback. Any help/feedback would be appreciated :) Thanks.

[Shakey_Jake33]

Quick question about enabling 24-bit sound support. I have a Blu-Ray disc which I know has 24-bit 96KHz Stereo LPCM, and I've been trying to get it to play in 24-bit using MPC and ffdshow. I've enabled 24-bit 96Khz sound in the Windows sound setup itself, and ticked 24-bit integer in the ffdshow setup, but it still seems to be outputting using 16-bit integer. The only solution might be to disable 16-bit integer, but I'm assuming it's not a good idea to do that even most media uses 16-bit sound this forcing everything to 'upscale'.

Have I missed something? I'm using a Realtek HD Audio, which should support 24-bit sound.

[Midzuki]

clsid wrote:
Revision 3448 now stores the RAW video decoder settings separately. Same applies for its subtitle variant.

Improvement confirmed. :)

Also: the bug reported by Keiyakusha remains unfixed. :(

Besides: in MPC, crash also happens when, in paused mode,
the arrow "-->" is pressed. :confused:

Serious Suggestion: remove the "In Text" pins from the raw video filter !!!

http://forum.videohelp.com/attachments/1933-1274498970/ffdshow-vp-trouble.png

[Sarasa]

I have a problem with another ass subtitle

Spec of Clip : MKV > AVC1 + Vorbis + ASS sub
FFDSHOW > ffmpeg-mt h264 + Tremor
Tested Output > YUY2 & RGB32
Using ffdshow (3448) / MPCH 1922 (EVR)

The sub should appear like this
> http://i50.tinypic.com/nxvyu.jpg

But FFdshow subtitle make it appear like this
> http://i47.tinypic.com/25uqjrk.jpg


Here is the ass files ;)

[jhoff80]

Apologies if this was mentioned already. It's a problem I've had for a long time, and never got around to troubleshooting until now, so I never posted it.

Basically, I'm running Windows 7 Ultimate x64 with Windows Media Center and the x64 versions of FFDShow-tryouts. Unfortunately, pretty much any build newer than the betas just doesn't work in Windows Media Center any more.

I've already used Win7DSFilterTweaker, both to set FFDShow as the default and to disable Media Foundation.

Upon playing the file, what I'm left with is a frozen background. This happens with any file type set to use ffdshow, regardless of whether I'm using DXVA or not.

http://img94.imageshack.us/img94/5238/p5220232.th.jpg (http://img94.imageshack.us/i/p5220232.jpg/)

As I mentioned, the older betas work 100% with all of my files. I just wanted to update to take advantage of newer features.

The PC is a Core 2 Quad Q6600, 8GB RAM, and an ATI 4850 (running Catalyst 10.4), if that matters.

Also, it's definitely some sort of interaction with Media Center, because using MPC-HC with all internal filters disabled, files play correctly. Unfortunately, because of how my system is set up for recording premium cable content, anything other than Media Center is not an option though.

I even did a new install of Windows 7 today and it was an issue even with that.

[ptr727]

Hi, I am trying to enable OSD to see CPU usage and framerate during playback.

Windows 7 x64, Haali, and ffdshow DXVA, playing using Windows Media Player.

I enable the framerate option in ffdshow DXVA OSD configuration.

During playback I enable OSD from the context menu.
No OSD?

How can I see stats using OSD during playback?

P.

[namaiki]

Hi guys,

got some trouble with subs here. I've searched the thread for references to my gfx card with no luck... Anyway, i just got a new laptop with an integrated Intel GMA 4500MHD. When running in DXVA-mode i get lots of artifacts around the subtitles (screenshot: http://i46.tinypic.com/s1tffo.png). Without DXVA enabled everything is OK. Running Win7x64, latest ffdshow x64, MPC-HCs audiorenderer and mkvsplitter.

The CPU is only a Pentium SU4100 1,3ghz dual core, so i really need to have dxva enabled for 1080p playback. Any help/feedback would be appreciated :) Thanks.
Use the Microsoft DXVA filter and MPC-HC's internal subtitle renderer. Here is an ini file for your convenience. http://www.mediafire.com/file/2jnmzqzwodu/mpc-hc.ini

Put it in the same directory/folder as the MPC-HC.exe

[namaiki]

I have a problem with another ass subtitle

Spec of Clip : MKV > AVC1 + Vorbis + ASS sub
FFDSHOW > ffmpeg-mt h264 + Tremor
Tested Output > YUY2 & RGB32
Using ffdshow (3448) / MPCH 1922 (EVR)

The sub should appear like this
> http://i50.tinypic.com/nxvyu.jpg

But FFdshow subtitle make it appear like this
> http://i47.tinypic.com/25uqjrk.jpg


Here is the ass files ;)

Which component is rendering the subtitles?

[Sarasa]

Which component is rendering the subtitles?For the first screen ? It's DirectVobsub.


Also here is the option I used in ffdshow

In subtitle > every think check except heuristic search
Placement > default option
Text > just "Split long subtitle line" - smart wrapping, line are evenly broken
Font > default option
Bluray&dvd > default option

In MPCH (1922) I use EVR for the output. I have added DirectVobsub as an external filter, that way I can block it when I use FFdshow for the subtitle.

[Midzuki]

Yet another serious suggestion: move ALL subtitling functions to a standalone "ff_subs.ax". :cool:

[STaRGaZeR]

I have a problem with another ass subtitle

I'll take a look at it, but not in a while. I'm very busy right now.

[Keiyakusha]

Hi STaRGaZeR!
About subtitles:
MPC-HC's (http://imgur.com/DWowD.png) output, which is good. VSFilter obviously looks the same.
ffdshow's (http://imgur.com/2itYn.png) subtitle renderer.
All subtitles settings in ffdshow I believe on defaults. However I did tried change settings but without luck.

But... Fun begins here:
MPC-HC (http://imgur.com/TnSu6.png) vs. FFDshow (http://imgur.com/542yb.png)

Subtitles itself: here (http://s2.sigmirror.com/files/78028_dxmkd/ffsubsbug.7z). Have fun! ^_^
EDIT: reuploaded to another host.

Albain
About crash I reported before. In addition to everything else now I noticed that ffdshow also crashes if I press space button, which is play/pause in MPC-HC...

[Keiyakusha]

Revision 3448 now stores the RAW video decoder settings separately. Same applies for its subtitle variant.
Thank you!
Is it possible to access their settings through shortcut just like directshow filter settings?

Serious Suggestion: remove the "In Text" pins from the raw video filter !!!
Wait! Here is my (almost)usual fillter (http://imgur.com/EcXdA.png) chain. How to get subs working if that pin removed? There is ffdshow subtitles filter (http://imgur.com/42shL.png) but then we won't get postprocessing. In any case having that pin doesn't do any harm.

Yet another serious suggestion: move ALL subtitling functions to an independent "ff_subs.ax". :cool:
Currently ffdshow subtitles filter is a part of ffdshow.ax. Why you need separate ax file? What is the benefits?

[Midzuki]

@ Keiyakusha:

it's possible to get ffdshow video processing together with "DirectVobSub", so I don't see how it shouldn't be possible to apply the same to ffdshow's subtitle filter.

As it should be obvious from my post on the page before,
the video processor connects to the subpicture pin of Microsoft's DVD-Navigator, and applies the following resizing "ratiotinium":

video resolution = 352x480, AR = 4:3,
therefore width of displayed video =
352 x (640/352) ;

subtitle picture resolution = 720x480, AR = 4:3,
therefore width of displayed subpicture =
720 x (640/352) ;

[Brazil2]

Revision 3448 now stores the RAW video decoder settings separately. Same applies for its subtitle variant.
Great news! Thanks for that :)



Subtitles itself: here (http://www.mediafire.com/?2iwogmyi2wm).
The archive seems to be corrupted, I've downloaded it 3 times -after clearing browser cache- and always got the same error.

[Keiyakusha]

Midzuki

Sorry I don't really understand your answer. Here is my questions in somewhat different form:
1) How to use ffdshow subtitles + postprocessing + 3rd party decoder at the same time if that pin removed?
2) Why these pins should be removed to begin with? No one is forced to use them if they are not needed.
3) Why separate .ax file for subtitles filter is better?

[STaRGaZeR]

Hi STaRGaZeR!
About subtitles:
MPC-HC's (http://imgur.com/DWowD.png) output, which is good. VSFilter obviously looks the same.
ffdshow's (http://imgur.com/2itYn.png) subtitle renderer.
All subtitles settings in ffdshow I believe on defaults. However I did tried change settings but without luck.

But... Fun begins here:
MPC-HC (http://imgur.com/TnSu6.png) vs. FFDshow (http://imgur.com/542yb.png)

Subtitles itself: here (http://www.mediafire.com/?2iwogmyi2wm). Have fun! ^_^

Albain
About crash I reported before. In addition to everything else now I noticed that ffdshow also crashes if I press space button, which is play/pause in MPC-HC...

Yup, the first one is because rotation is not implemented in ffdshow, and the kanji subs are rotated 90º. Sarasa reported it earlier. I have a semi-working patch here, but nowhere near usable. I will try to finish it when I have time. Dunno about the second, will look at it.

Your file is corrupted BTW.

[Keiyakusha]

Hmm. I tried to reupload it but it seems mediafire have some problems. Here (http://s2.sigmirror.com/files/78028_dxmkd/ffsubsbug.7z) uploaded it to another place.

[STaRGaZeR]

Got it. One thing, do not use ffdshow and MPC at the same time. This can be the cause of the second issue. First try it with ffdshow alone (you have to disable the MPC subtitle renderer, disabling the subs once loaded is not enough), then with MPC alone.

[Keiyakusha]

I did exactly like you said.
When I checked subtitles in ffdshow, filter graph looked like this (http://i.imgur.com/EcXdA.png) (yes, this is the same picture I posted for Midzuki, don't be confused). Autoload subtitles in MPC was disabled beforehand.
When I checked subs in MPC, internal renderer was ON and ffdshow wasn't present at all in the chain.
I also tried following filter chain:
haali -> coreavc -> vsfilter -> video renderer
and in looks exactly as MPC internal renderer.

[leeperry]

BTW, in the latest builds the ffdshow DS properties show the stream being processed and fill the screen with shortcuts to all the available filters/profiles...is there to make it simple as it was before? it was only showing "ffdshow audio/video filter"...now I have to scroll all the way up to reach "setup" :o

[STaRGaZeR]

I did exactly like you said.
When I checked subtitles in ffdshow, filter graph looked like this (http://i.imgur.com/EcXdA.png) (yes, this is the same picture I posted for Midzuki, don't be confused). Autoload subtitles in MPC was disabled beforehand.
When I checked subs in MPC, internal renderer was ON and ffdshow wasn't present at all in the chain.
I also tried following filter chain:
haali -> coreavc -> vsfilter -> video renderer
and in looks exactly as MPC internal renderer.

Yeah, confirmed. Can you try if it happened before r3353? Just a guess here, but the lines in this script are so freaking huge (seriously, what the hell) maybe ffdshow needs more memory than the harcoded max.

[Keiyakusha]

@ clsid and devs.
Hi. First of all, just now I noticed that enabling/disabling checkboxes on this page (http://imgur.com/uGmi5.png) of RAW filter also affects subtitle filter. Is that how it supposed to be?

Second. We discussed about removing pins a bit more, so here is the situation:

We have some source with subtitles.
We want to use RAW filter as postprocessor.
But we don't want to use ffdshow subtitles in this postprocessor.
However RAW filter stealing subpicture connection.

Making separate options for RAW and subtitle filters should fix this issue, since when we disable all chechboxes in subtitle settings page, subpictures can't connect to raw filter (which is what we want).

Another solution will be adding an option to RAW filter, which will be called: "do not allow subpicture connections".

@ STaRGaZeR
Just tried r3351. Well I won't say it looks better, weird artifacts happens sometimes, bit it looks different, yes. Here is screenshot (http://imgur.com/etX0S.png).

[STaRGaZeR]

Nah, that's not it then.

[albain]

@albain

Could you have a look at the crash that Keiyakusha reported? There basically is a crash when right-clicking on the tray icon when both ffdshow video decoder and the raw video decoder are in the same filter graph.

Sorry, I was away, this is fixed in revision 3452

[jhoff80]

Apologies if this was mentioned already. It's a problem I've had for a long time, and never got around to troubleshooting until now, so I never posted it.

Basically, I'm running Windows 7 Ultimate x64 with Windows Media Center and the x64 versions of FFDShow-tryouts. Unfortunately, pretty much any build newer than the betas just doesn't work in Windows Media Center any more.

I've already used Win7DSFilterTweaker, both to set FFDShow as the default and to disable Media Foundation.

Upon playing the file, what I'm left with is a frozen background. This happens with any file type set to use ffdshow, regardless of whether I'm using DXVA or not.

http://img94.imageshack.us/img94/5238/p5220232.th.jpg (http://img94.imageshack.us/i/p5220232.jpg/)

As I mentioned, the older betas work 100% with all of my files. I just wanted to update to take advantage of newer features.

The PC is a Core 2 Quad Q6600, 8GB RAM, and an ATI 4850 (running Catalyst 10.4), if that matters.

Also, it's definitely some sort of interaction with Media Center, because using MPC-HC with all internal filters disabled, files play correctly. Unfortunately, because of how my system is set up for recording premium cable content, anything other than Media Center is not an option though.

I even did a new install of Windows 7 today and it was an issue even with that.

I figured it out. The "Connect as PCM first" option in the audio decoder passthrough breaks things in Windows Media Center. Disabling that makes it work perfectly.

[Keiyakusha]

Sorry, I was away, this is fixed in revision 3452

Yep, seems to be fixed. Thank you! :)

EDIT: however if "add tray icon menu to player's stream/languages menu" is active and i'm trying to turn on/off subtitles - MPC-HC with ffdshow crashes.
EDIT2: or better to say this option now has no effect. Tray icon menu always there. If I selecting subtitles in tray icon menu - it works.
Also I noticed another problem, if I'm trying to select mplayer temporal noise reducer, libavcodec crashes...

[DigitalDeviant]

Positioning is fixed but there are some ugly lines through some subtitles and the subpic borders show some junk too. I notice it on other Bluray discs too.

Edit: It pappears with all usable scalers and with both RGB and YV12. Additionally the above sample shows a greenish tint on the subtitle outline with YV12 output.

The garbage lines appear fixed now except when there are two subtitles on screen at the same time. It's getting much closer to perfect, thanks albain.

[leeperry]

so I guess this is non-negotiable? :(

seriously: http://thumbnails15.imagebam.com/8179/def72a81781406.gif (http://www.imagebam.com/image/def72a81781406)

It doesn't happen for ffdshow audio...so why do you force it for ffdshow video?

After you've spent a lot of time working on automatic profiles, you couldn't care less about all this...all it does is filling my HTPC screen w/ useless stuff, and require one more click to reach "setup" on the very top of the screen :o

It'd be really nice if you could add an option to disable all those shorcuts, in order to make it the way it was before...hope you will consider it :thanks:

[Sarasa]

I'll take a look at it, but not in a while. I'm very busy right now.Thank's and take your time ;)

[clsid]

@ clsid and devs.
Hi. First of all, just now I noticed that enabling/disabling checkboxes on this page (http://imgur.com/uGmi5.png) of RAW filter also affects subtitle filter. Is that how it supposed to be?

Second. We discussed about removing pins a bit more, so here is the situation:

We have some source with subtitles.
We want to use RAW filter as postprocessor.
But we don't want to use ffdshow subtitles in this postprocessor.
However RAW filter stealing subpicture connection.

Making separate options for RAW and subtitle filters should fix this issue, since when we disable all chechboxes in subtitle settings page, subpictures can't connect to raw filter (which is what we want).
The raw filter and the raw subtitle filter currently share their settings.

Please explain why you are having both in the same graph, because I see no reason for it. In fact, it is just very inefficient to do so.

[clsid]

so I guess this is non-negotiable? :(

seriously: http://thumbnails15.imagebam.com/8179/def72a81781406.gif (http://www.imagebam.com/image/def72a81781406)
I am not sure what the rationale was behind this change. Maybe Albain can explain. Maybe for selecting subs?

Perhaps we could show just an on/off for subtitles in there, instead of showing all the filters.

[Keiyakusha]

Please explain why you are having both in the same graph, because I see no reason for it. In fact, it is just very inefficient to do so.
No its not about two in the same graph. Midzuki saying that there is situation when subtitles connected to raw filter when we don't want this and proposed remove intext pins from raw filter. And I was saying that this can't/shouldn't be done, and that if that really happens - there should be another ways to solve that (without affecting subtitles filter).
Also having the same options kind of defeats the purpose of having separate subtitles filter at all.

By the way there probably can be confusion because it is not obvious enough that here (http://imgur.com/Xioqf.png), if checkbox that marked red is unchecked, this options marked blue still has effect on intext pin behavior and thats why he says that RAW filter "stealing" connection.

so I guess this is non-negotiable? :(
You probably haven't seen MPC-HC (http://imgur.com/VLBVQ.png) yet. :)

[Thunderbolt8]

so I guess this is non-negotiable? :(

seriously: http://thumbnails15.imagebam.com/8179/def72a81781406.gif (http://www.imagebam.com/image/def72a81781406)

It doesn't happen for ffdshow audio...so why do you force it for ffdshow video?

After you've spent a lot of time working on automatic profiles, you couldn't care less about all this...all it does is filling my HTPC screen w/ useless stuff, and require one more click to reach "setup" on the very top of the screen :o

It'd be really nice if you could add an option to disable all those shorcuts, in order to make it the way it was before...hope you will consider it :thanks:+1 from me

[leeperry]

I am not sure what the rationale was behind this change. Maybe Albain can explain. Maybe for selecting subs?

Perhaps we could show just an on/off for subtitles in there, instead of showing all the filters.
I dunno, but it takes one more click to reach "setup"...I liked it the way it was before tbh :o

one click > the ffdshow config panel opens: perfect!

but this new layout is good if you're still messing around w/ the settings, so a simple checkbox to enable all those shortcuts would be really great....best of both worlds!

[clsid]

No its not about two in the same graph. Midzuki saying that there is situation when subtitles connected to raw filter when we don't want this and proposed remove intext pins from raw filter. And I was saying that this can't/shouldn't be done, and that if that really happens - there should be another ways to solve that (without affecting subtitles filter).
Also having the same options kind of defeats the purpose of having separate subtitles filter at all.No pins will be removed.

[leeperry]

oh and BTW, I know I still believe in Santa Claus...but if there were a way you guys could look into this bug report, that'd be really sweet: http://sourceforge.net/tracker/index.php?func=detail&aid=2725991&group_id=173941&atid=867360

because I posted a sample a little while ago, and some AC3 files are 5.1 but seem to be tagged as stereo..so using the ffdshow built-in AC3 decoders, it will trigger the stereo automatic profile :(

using AC3Filter fixes the problem altogether, but ffdshow is unable to detect the number of channels...for some reason :o

[leeperry]

anyway, I tried the new layout...it doesn't seem to work for me :o

I've got the Avisynth filter enabled and ticked, if I click on it...it enables the "Offset" filter :confused:

if I click on "Overlay", it flips the picture?! it looks like none of those things are doing what they are supposed to for me..

[Midzuki]

No pins will be removed.

Does that mean the end-user will be allowed to control their (mis)behavior { i.e., enable/disable them } ?

[Keiyakusha]

Midzuki
Do you still see the problem if you disable all options that marked blue on this picture (http://i.imgur.com/Xioqf.png)? Because I think if everything is disabled there, RAW filter shouldn't except connections to in-text pin. If it still accepts them, maybe this is a bug.

[Midzuki]

@ Keiyakusha:

For the time being, I can only confirm the old BUG still is present in r3448. Yes, I downloaded r3452 , and I should have checked if that BUG remains unsolved, however I uninstalled r3452 as soon as I found out a brand-new problem — the raw video filter "automagically resizes" the clips to an incorrect aspect ratio, even if it has all of its checkboxes unticked.

EDIT:

Just a "wild guess" — maybe the new BUG happens only when the monitor is set to the "outdated" :rolleyes: 800x600 resolution.

[Sebastiii]

The garbage lines appear fixed now except when there are two subtitles on screen at the same time. It's getting much closer to perfect, thanks albain.

Hi, :)

On my test 3453 fix it :)
Thank you Albain as usual :)
Seb.

[DigitalDeviant]

Hi, :)

On my test 3453 fix it :)
Thank you Albain as usual :)
Seb.

On this (http://www.mediafire.com/?mgztymdmdjn)sample? I don't have the setup to compile it here to test so I'll have to wait until someone makes a build available. If it doesn't work I'll make a more detailed post since after I demuxed and examined the subs the problem isn't exactly what I thought it was.

Either was, thank you Albain.

[Sebastiii]

hi :)

Try just for try lol : ffdshow_rev3453_20100525-Debug.exe (http://sebinternet.free.fr/ffdshow_rev3453_20100525-Debug.exe)
:)
Yes it's on this sample :)
Cheers,
Seb.

[DigitalDeviant]

hi :)

Try just for try lol : ffdshow_rev3453_20100525-Debug.exe (http://sebinternet.free.fr/ffdshow_rev3453_20100525-Debug.exe)
:)
Yes it's on this sample :)
Cheers,
Seb.

Thanks. It'll be a while until I get home from work and can test but this has made my day a whole lot better. Well, this and cake...

[Eliminateur]

yesterday i installed win7 x64, installed ffdshow x64, mpchc x64 and all the hoopla only to find that i had 700S of jitter, 99% dropped frames or increasing linear jitter and 80% cpu usage on a 640x480 divx avi files!
After talking to some people i isolated the problem to the deband filter which i always used on 32b xp, they also told me that it's been broken since a logn time ago on x64 and it's even reported.

is there any plan on porting the deband filter to x64 correctly?
or if it's so broken, why allow it to be enabled, put a warning or in x64 make it do nothing

[Snowknight26]

Any chance of preventing lossless H.264 from being loaded with the DXVA filter?

[rack04]

I noticed in recent builds (don't know exactly which version) that if I right click while a video is playing in MPC-HC, select filters, select ffdshow video decoder, and select any filter it always selects the filter that is two space below the one that is selected. For example, I click on "subtitles" then "OSD" is selected.

[DigitalDeviant]

hi :)

Try just for try lol : ffdshow_rev3453_20100525-Debug.exe (http://sebinternet.free.fr/ffdshow_rev3453_20100525-Debug.exe)
:)
Yes it's on this sample :)
Cheers,
Seb.

It's definitely better but not gone.

http://img25.imageshack.us/img25/6645/badsub.th.png (http://img25.imageshack.us/i/badsub.png/)

If you look at the "had died" part there is still some junk. I don't get the subtitle format. When I view them in SupRip or BDSup2Sub it's not 2 separate subpics as I assumed it would be; it's one big one but the bottom lines aren't the same color as it's displayed during playback. MPC-HC's internal subtitle renderer displays it correctly however.

[yair]

hi, im looking for motion targa (MTGA) support for ffdshow,
on what shoulder should i cry on :)
http://forum.doom9.org/showthread.php?t=154781
targa spec (http://local.wasp.uwa.edu.au/~pbourke/dataformats/tga/)

tags: tagra, motion tagra, MTGA, directshow, libavcodec

[Sebastiii]

It's definitely better but not gone.

http://img25.imageshack.us/img25/6645/badsub.th.png (http://img25.imageshack.us/i/badsub.png/)

If you look at the "had died" part there is still some junk. I don't get the subtitle format. When I view them in SupRip or BDSup2Sub it's not 2 separate subpics as I assumed it would be; it's one big one but the bottom lines aren't the same color as it's displayed during playback. MPC-HC's internal subtitle renderer displays it correctly however.

Effectively, i'm too quick, i have the same result :) (it's closed to perfect :) )

[albain]

I am not sure what the rationale was behind this change. Maybe Albain can explain. Maybe for selecting subs?

Perhaps we could show just an on/off for subtitles in there, instead of showing all the filters.

This has always been like that, Milan implemented it this way

Actually there was no subtitle/audio streams inside this menu at first but ffdshow filters. I only added subs/audio after, but with a different group id.
This is the implementation of IAMStreamSelect interface which is the interface now used by all the players to display and select streams and maybe chapters too (Haali)

I chose to keep it that way. Maybe there is a bug on the selection part, not on audio/subs but on internal streams, I have never tried.

Anyway, we can choose to remove ffdshow internal filters from the list (I don't have any objection) or to tell other player developers to take into account the group ID of the stream (0 for video, 1 for audio, 2 or 99 for subtitles, and other values for other types) as they should do.

It's definitely better but not gone.

http://img25.imageshack.us/img25/6645/badsub.th.png (http://img25.imageshack.us/i/badsub.png/)

If you look at the "had died" part there is still some junk. I don't get the subtitle format. When I view them in SupRip or BDSup2Sub it's not 2 separate subpics as I assumed it would be; it's one big one but the bottom lines aren't the same color as it's displayed during playback. MPC-HC's internal subtitle renderer displays it correctly however.

Indeed, I have not seen that one. I had to rewrote all the PGS parsing because all the existing PGS parsers are buggy. This problem is tricky because the picture is cut into several parts, here in the midle of the yellow sub (and not between the white and yellow), so I have to find and tune how/where to start each new part

[DigitalDeviant]

Well, obviously it doesn't render the subtitle unreadable so it's not a critical bug. I've only seen this one title with subs like this and if it is too distracting MPC-HC will play it correctly.

Now I have a quick question. How does ffdshow handle forced subtitles in Blu-ray? I had someone ask me this and couldn't answer them and I don't have any samples to test it myself.

[obieobieobie]

Anyone else having problem with post processing and FLV4 files? It gets really blurry as if there's an inloop filter + post processing enabled. I experienced the problem in rev3426 and rev3452 which I just downloaded.

[GrandeBoma]

Well, obviously it doesn't render the subtitle unreadable so it's not a critical bug. I've only seen this one title with subs like this and if it is too distracting MPC-HC will play it correctly.

Now I have a quick question. How does ffdshow handle forced subtitles in Blu-ray? I had someone ask me this and couldn't answer them and I don't have any samples to test it myself.

as far as I am concerned, mpc does not handle forced subtitles properly (yet).

there are 2 types of blu-rays: some have forced subtitles in a separate file (imagine 2 subtitle files, one for watching the entire movie with subtitles, one for watching only those forced subs). In this case it's simple: just select the right sub track and only forced subs will be displayed. An example of this would be the bourne or the lord of the rings trilogies

Some other movies have only one subtitles file for each language :mad: and those who are "forced" are flagged as such, but for the time being the mpc/ffdshow subtitle engine is not capable to tell which is which and will display everything. We would need an option such as those offered by vobsub filter to "only show forced subtitles". An example of this would be District 9 or Avatar

[STaRGaZeR]

Albain, what do you think about putting audio tracks, embedded subtitle tracks, subtitle files, etc. each in their own subnenu? i.e. ffdshow Video Decoder --> Subtitle files --> list of all subtitle files found. This would keep functionality while reducing the mess it is now with files with lots of subtitles or audio streams.

[leeperry]

reducing the mess
that sounds like a great idea! just like offering an option to switch all this off and allow the end-users to have it the exact same way it was before..no useless options clogging the screen, no "setup" button on the very top of the menus...a one clicker like before, ya know http://forum-images.hardware.fr/images/perso/ayuluna.gif

[STaRGaZeR]

that sounds like a great idea! just like offering an option to switch all this off and allow the end-users to have it the exact same way it was before..no useless options clogging the screen, no "setup" button on the very top of the menus...a one clicker like before, ya know http://forum-images.hardware.fr/images/perso/ayuluna.gif

The problem here is that the "Add tray icon menu to player's stream/laguages menu" doesn't work since the lastest changes. If you have it disabled nothing should appear, like it was before.

Try not to parrot the same things over and over please, it's annoying. We get it, ya know.

[leeperry]

We get it
Obviously some ppl don't, and english is not everyone's native tongue so I simply try to make it clear..you can always add me to your ignore list and call it a day. Those changes are uncalled for, and I understand that there's hardly any active ffdshow coders atm...so I don't mean to upset anyone, but if they all decide to make this big of a change on their own...who's in charge to judge what's good for the users community, and what isn't? Could there be some sort of end-users poll for major GUI changes? Just saying anyway http://forum-images.hardware.fr/images/perso/piksou.gif

Moving all that subs config to the tasktray icon would indeed be a great idea...as I don't use it :p

[Cloudstrifeff7]

I have a problem with ffdshow 64 and mpc-hc 64. If I use deband filter, I have desynchronisation of video and sound. O_o
With 32 bits version, I don't have problem.

I use win 7 64 bit.

[clsid]

Deband is mega slow on x64. We need a volunteer to fix it (or disable it).

[STaRGaZeR]

Deband is mega slow on x64. We need a volunteer to fix it (or disable it).

Deband has been ported to x64 by JoshyD, I asked for the source code but he's busy. I'd wait a bit.

[albain]

Obviously some ppl don't, and english is not everyone's native tongue so I simply try to make it clear..you can always add me to your ignore list and call it a day. Those changes are uncalled for, and I understand that there's hardly any active ffdshow coders atm...so I don't mean to upset anyone, but if they all decide to make this big of a change on their own...who's in charge to judge what's good for the users community, and what isn't? Could there be some sort of end-users poll for major GUI changes? Just saying anyway http://forum-images.hardware.fr/images/perso/piksou.gif

Moving all that subs config to the tasktray icon would indeed be a great idea...as I don't use it :p

What you don't understand is that nothing has changed in FFDShow on this part; I just improved the subtitle/menu extraction but this is just internal.

This change is visible to you because your player now supports this interface. You should complain to them so that they filter streams. Media Portal, MPC-HC do so to only retrieve the streams from the groups they are interested in

[STaRGaZeR]

What you don't understand is that nothing has changed in FFDShow on this part; I just improved the subtitle/menu extraction but this is just internal.

This change is visible to you because your player now supports this interface. You should complain to them so that they filter streams. Media Portal, MPC-HC do so to only retrieve the streams from the groups they are interested in

"Add tray icon menu to player's stream/laguages menu" should enable/disable everything though, like before. Now it makes no effect.

[leeperry]

What you don't understand is that nothing has changed in FFDShow on this part; I just improved the subtitle/menu extraction but this is just internal.

This change is visible to you because your player now supports this interface. You should complain to them so that they filter streams. Media Portal, MPC-HC do so to only retrieve the streams from the groups they are interested in
thanks for the reply! so to make it clear, KMP hasn't been updated in 9 months: http://www.kmplayer.com/forums/showthread.php?t=14938

older version of ffdshow: http://thumbnails2.imagebam.com/8237/7ef4e482367560.gif (http://www.imagebam.com/image/7ef4e482367560) newer version: http://thumbnails28.imagebam.com/8237/beb2c682367561.gif (http://www.imagebam.com/image/beb2c682367561)

and somehow, it's KMP's fault :( if you could please set an option to hide all this, I'm sure many other ppl would also enjoy it either way. I use subs in ffdshow from time to time(because mVR doesn't have a subs pin), but I'd rather make 3 clicks when I need to than scroll all the way up and need 2 clicks to reach "setup" on the very top of the screen when I don't..hope you will consider it :o
Asking questions always sounds like b*tching on public forums, I apologize for this. :thanks:

[Keiyakusha]

There was option to hide this "add tray icon's menu to player's menu" (or something like that). But as was said before now this option has no effect and ffdshow acts like this option always ON. This probably will be fixed. However even if this option will be ON, player's menu still has a lot of unneeded stuff which wasn't there before (or was, but In some better way than now, I don't remember...). Now in MPC-HC's case every possible subtitles in the folder + some random text files printed in this menu...
But my observations about that options valid only for RAW filter. I can't access directshow filter settings at all - immediate crash when I'm trying.

EDIT: if adding subtitles files into this menu is necessary for some reason, then it shouldn't print everything it can find in the folder, but follow common "tradition" which is only movie.srt or movie.any-words-here.srt, where movie - the name of the video file currently played. Of course .srt extension is only for example.

[teohhanhui]

Crash with DXVA on for certain problematic files (encoder settings blanked out).

Disabling DXVA for H.264 or using MPC-HC (r1976) built-in DXVA filter works fine. It was still working in r3456 (broken in r3460, no idea about the revisions in between).

[clsid]

@Albain

I think the majority wants to hide the filters from the player's menu. So I suggest to only show the subtitle filter in there. Full settings can still be accessed through "Setup", which should perhaps be renamed to "Settings"?

@teohhanhui
Could you please upload a sample file for me?
If it is large, you can cut a part with this tool:
http://neuron2.net/dgsplit/dgsplit11.zip

[leeperry]

Well, it'd be good to have an option to hide it all as it was before...the percentage of ppl using subs in ffdshow is prolly very very low, and why forcing the end-user to click twice("setup") to see ffdshow's config when one click has been enough for years? :o

[Sarasa]

Well, it'd be good to have an option to hide it as it was before...the percentage of ppl using subs in ffdshow is prolly very very low, and why forcing the end-user to click twice("setup") to see ffdshow's config when one click has been enough for years? :oThere are more people that use ffdshow to display sub that you think.
If you use the auto profiles to resize the video depending of the aspect (exemple: default > 4/3, HQ > 16/9,...) and it autoswitch to HQ, DirectVobsub won't display the sub even if you switch between the different sub.
Conclusion : ffdshow is needed to switch/activate subs ;)

Also some video lag when using DirectVobsub to display the sub, but play smoothly if you use ffdshow to display it ;)

[leeperry]

reason why an option would be nice, to either have it the way it is right now...or the way it was before, no "Setup" button, no filters, no subs. Best of both worlds :)

[STaRGaZeR]

reason why an option would be nice, to either have it the way it is right now...or the way it was before, no "Setup" button, no filters, no subs. Best of both worlds :)

Are you trolling or something? There IS an option, but it doesn't work after the lastest changes.

[hoborg]

Looks like MPC-HC DXVA decoder fixed problem with some MP4 file (http://hobring.ic.cz/samples/mp4.AAC+AVC1_(1280x720@50fps).rar)s produce only black screen in DXVA mode, but the problem is still there if FFDShow DXVA decoder is in use.
Can be FFDShow DXVA decoder updated with latest MPC-HC source?

I think it is related to this ticket: http://sourceforge.net/apps/trac/mpc-hc/ticket/24

BTW, there is glitch (like bad frame order) durning playback of this sample.

[leeperry]

Are you trolling or something?
Having a bad day, huh? :D

This will get rid of all the filters, but the "setup" button will most likely still be there.

I never had to check that option before...anyways, looking forward for a way to have it exactly the way it was before...if any possible. Thanks albain http://forum-images.hardware.fr/images/perso/sniperr.gif

[clsid]

Looks like MPC-HC DXVA decoder fixed problem with some MP4 file (http://hobring.ic.cz/samples/mp4.AAC+AVC1_(1280x720@50fps).rar)s produce only black screen in DXVA mode, but the problem is still there if FFDShow DXVA decoder is in use.
Can be FFDShow DXVA decoder updated with latest MPC-HC source?

I think it is related to this ticket: http://sourceforge.net/apps/trac/mpc-hc/ticket/24

BTW, there is glitch (like bad frame order) durning playback of this sample.
I shall update to the latest MPC code.

[STaRGaZeR]

Having a bad day, huh? :D

This will get rid of all the filters, but the "setup" button will most likely still be there.

I never had to check that option before...anyways, looking forward for a way to have it exactly the way it was before...if any possible. Thanks albain http://forum-images.hardware.fr/images/perso/sniperr.gif

That option is and was off by default, off means nothing appears and your beloved one click solution is there. If you never touched it, of course it was never activated. But now it doesn't work, it's always on, for the nth time. This is a bug and should be corrected. Do you get it now or will you post another 10+ posts asking for an option to remove the stuff? :rolleyes:

EDIT: r3445 is the cause.

[Midzuki]

And as my question remains unanswered... :rolleyes:

No pins will be removed.

Does that mean the end-user will be allowed to control their (mis)behavior { i.e., enable/disable them } ?

FWIW: I've just "upgraded" from r3176 to r3426. :p

[clsid]

Post a proper bug report.

[Atak_Snajpera]

How far are we from added VP8 decoder to ffdshow?

[clsid]

The official decoder works fine, so I am not investing time in adding it to ffdshow.

[Atak_Snajpera]

We have vp3,vp5,vp6,vp6f so why not vp8? Official decoder can not smooth picture in fullscreen (windows 7). Also DirectShowSource does not work without fps parametr.

http://img203.imageshack.us/img203/2003/new2mn.png

[clsid]

VP8 is open source. The rest is not.

If someone wants to add it to ffdshow, fine, but I am not going to do it. At least not until ffmpeg has its own decoder implementation.

[Atak_Snajpera]

At least not until ffmpeg has its own decoder implementation.
understood.

[Delerue]

I can't open this sample here (http://red.cachefly.net/video/milkgirls1080p.mov) that uses MJPEG-B codec. Anyone confirms that?

Thanks!

[hoborg]

I can't open this sample here (http://red.cachefly.net/video/milkgirls1080p.mov) that uses MJPEG-B codec. Anyone confirms that?

Thanks!

I doubt "mjpb" is supported.
You can play it in MPC-HC if you install QT frameworks (QT alternative).

[XhmikosR]

I'm getting some crashes with some mkv (h264+ac3) videos, r3363 ICL11 build. MSVC build does not have this crash. Here's the debugger's output:

Microsoft (R) Windows Debugger Version 6.12.0002.633 X86
Copyright (c) Microsoft Corporation. All rights reserved.


Loading Dump File [C:\Users\XhmikosR\AppData\Local\CrashDumps\mpc-hc.exe.1092.dmp]
User Mini Dump File: Only registers, stack and portions of memory are available

Symbol search path is: srv*H:\progs\Win7x86RTMSymbols2\websymbols*http://msdl.microsoft.com/download/symbols
Executable search path is:
Windows 7 Version 7600 MP (4 procs) Free x86 compatible
Product: WinNt, suite: SingleUserTS
Machine Name:
Debug session time: Mon May 31 01:16:18.000 2010 (UTC + 3:00)
System Uptime: not available
Process Uptime: 0 days 0:00:50.000
................................................................
........................
Loading unloaded module list
.............
This dump file has an exception of interest stored in it.
The stored exception information can be accessed via .ecxr.
(444.f3c): Access violation - code c0000005 (first/second chance not available)
eax=00000000 ebx=043cd92c ecx=00000400 edx=00000000 esi=00000002 edi=00000000
eip=775464f4 esp=043cd8dc ebp=043cd978 iopl=0 nv up ei pl zr na pe nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
ntdll!KiFastSystemCallRet:
775464f4 c3 ret
0:006> !analyze -v
*******************************************************************************
* *
* Exception Analysis *
* *
*******************************************************************************

*** ERROR: Symbol file could not be found. Defaulted to export symbols for splitter.ax -
*** ERROR: Symbol file could not be found. Defaulted to export symbols for mzvkbd3.dll -

FAULTING_IP:
ffdshow!get_ue_golomb+39 [h:\progs\compiling\ffdshow\src\ffmpeg\libavcodec\golomb.h @ 70]
06066f59 8b0402 mov eax,dword ptr [edx+eax]

EXCEPTION_RECORD: ffffffff -- (.exr 0xffffffffffffffff)
ExceptionAddress: 06066f59 (ffdshow!get_ue_golomb+0x00000039)
ExceptionCode: c0000005 (Access violation)
ExceptionFlags: 00000000
NumberParameters: 2
Parameter[0]: 00000000
Parameter[1]: 06fa5000
Attempt to read from address 06fa5000

DEFAULT_BUCKET_ID: INVALID_POINTER_READ

PROCESS_NAME: mpc-hc.exe

ERROR_CODE: (NTSTATUS) 0xc0000005 - The instruction at 0x%08lx referenced memory at 0x%08lx. The memory could not be %s.

EXCEPTION_CODE: (NTSTATUS) 0xc0000005 - The instruction at 0x%08lx referenced memory at 0x%08lx. The memory could not be %s.

EXCEPTION_PARAMETER1: 00000000

EXCEPTION_PARAMETER2: 06fa5000

READ_ADDRESS: 06fa5000

FOLLOWUP_IP:
ffdshow!get_ue_golomb+39 [h:\progs\compiling\ffdshow\src\ffmpeg\libavcodec\golomb.h @ 70]
06066f59 8b0402 mov eax,dword ptr [edx+eax]

MOD_LIST: <ANALYSIS/>

NTGLOBALFLAG: 0

APPLICATION_VERIFIER_FLAGS: 0

FAULTING_THREAD: 00000f3c

PRIMARY_PROBLEM_CLASS: INVALID_POINTER_READ

BUGCHECK_STR: APPLICATION_FAULT_INVALID_POINTER_READ

LAST_CONTROL_TRANSFER: from 0606726b to 06066f59

STACK_TEXT:
043ce024 0606726b 043ce324 043ce6f8 0053f250 ffdshow!get_ue_golomb+0x39 [h:\progs\compiling\ffdshow\src\ffmpeg\libavcodec\golomb.h @ 70]
043ce114 0606692a 043ce324 a393d1e2 027bd21c ffdshow!decode_hrd_parameters+0x7b [h:\progs\compiling\ffdshow\src\ffglobals.cpp @ 654]
043ce704 06133455 06fa4f08 00000026 027bd220 ffdshow!decodeH264SPS+0x87a [h:\progs\compiling\ffdshow\src\ffglobals.cpp @ 935]
043cea9c 06146843 00531c98 043cecd8 043ceb9c ffdshow!TffdshowVideoInputPin::init+0xb05 [h:\progs\compiling\ffdshow\src\tffdshowvideoinputpin.cpp @ 427]
043ceb90 06b5ddb4 00531c98 a393da02 043cef10 ffdshow!TinputPin::SetMediaType+0x63 [h:\progs\compiling\ffdshow\src\tinputpin.cpp @ 36]
043cece4 06132798 027bcb88 0497e444 00531c98 ffdshow!CBasePin::ReceiveConnection+0x2a4 [h:\progs\compiling\ffdshow\src\baseclasses\amfilter.cpp @ 2038]
043cef1c 1003fb7d 027bcb88 0497e444 00531c98 ffdshow!TffdshowVideoInputPin::ReceiveConnection+0x318 [h:\progs\compiling\ffdshow\src\tffdshowvideoinputpin.cpp @ 298]
WARNING: Stack unwind information not available. Following frames may be wrong.
043cef50 10041948 0497e438 027bcb88 00000000 splitter!DllGetClassObject+0x177e
043cef7c 100419f7 00000000 a3bc4ee5 0497e444 splitter!DllGetClassObject+0x3549
043cefa4 6c212cd7 04ae28f8 027bcb88 00000000 splitter!DllGetClassObject+0x35f8
043cf0ac 6c21ecfa 0497e444 027bcb88 00000000 quartz!CFilterGraph::ConnectDirectInternal+0xe9
043cf0cc 0137b5e7 0018e57c 0497e444 027bcb88 quartz!CFilterGraph::ConnectDirect+0x71
043cf154 0137c76e 0297bffc 0497e444 027bcb88 mpc_hc!CFGManager::ConnectDirect+0x1e7 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 506]
043cf1ac 0137c122 0297bffc 0497e444 027bcb88 mpc_hc!CFGManagerPlayer::ConnectDirect+0xfe [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 2615]
043cf1f0 01384b74 027bcb88 0497e444 027b8ac0 mpc_hc!CFGManager::ConnectFilterDirect+0x182 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 1088]
043cf330 01385128 0497e444 00000000 00000001 mpc_hc!CFGManager::Connect+0x954 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 701]
043cf340 01385651 0297bffc 0497e444 00000000 mpc_hc!CFGManager::Connect+0x18 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 541]
043cf3a8 01385a93 0297bffc 04ae28b4 00000000 mpc_hc!CFGManager::ConnectFilter+0x1a1 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 1019]
043cf42c 013ad9b6 04ae28b4 029636b8 00000000 mpc_hc!CFGManager::RenderFile+0x173 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\fgmanager.cpp @ 839]
043cf710 013c7673 0294b4f0 a362d4d4 0000c000 mpc_hc!CMainFrame::OpenFile+0x106 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\mainfrm.cpp @ 9861]
043cfa00 013c7fc5 0294b4f0 a362d968 00000000 mpc_hc!CMainFrame::OpenMediaPrivate+0x403 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\mainfrm.cpp @ 10966]
043cfa1c 01852e64 00000000 0294b4f0 0053c0d8 mpc_hc!CGraphThread::OnOpen+0x45 [h:\progs\compiling\mpc-hc_dev\src\apps\mplayerc\mainfrm.cpp @ 13570]
043cfa3c 01852f04 0053c0d8 0053c0d8 043cfa58 mpc_hc!CWinThread::DispatchThreadMessageEx+0x46 [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 771]
043cfa4c 01853104 0053c0d8 043cfa64 01852f6e mpc_hc!AfxInternalPreTranslateMessage+0x22 [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 227]
043cfa58 01852f6e 0053c0d8 043cfa94 01853151 mpc_hc!CWinThread::PreTranslateMessage+0xd [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 777]
043cfa64 01853151 0053c0d8 00000000 0295d5a0 mpc_hc!AfxPreTranslateMessage+0x19 [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 255]
043cfa74 01852ddc 002eb3a8 0295d5a0 0000021c mpc_hc!AfxInternalPumpMessage+0x2d [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 178]
043cfa94 0185353c a362d860 00000000 02962998 mpc_hc!CWinThread::Run+0x5b [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 629]
043cfb18 01886fc0 002eb3a8 a362d828 00000000 mpc_hc!_AfxThreadEntry+0xf4 [f:\dd\vctools\vc7libs\ship\atlmfc\src\mfc\thrdcore.cpp @ 126]
043cfb50 01887068 00000000 043cfb68 76be1194 mpc_hc!_callthreadstartex+0x1b [f:\dd\vctools\crt_bld\self_x86\crt\src\threadex.c @ 348]
043cfb5c 76be1194 02962998 043cfba8 7755b3f5 mpc_hc!_threadstartex+0x82 [f:\dd\vctools\crt_bld\self_x86\crt\src\threadex.c @ 326]
043cfb68 7755b3f5 02962998 71aeec2e 00000000 kernel32!BaseThreadInitThunk+0xe
043cfba8 7755b3c8 01886fe6 02962998 00000000 ntdll!__RtlUserThreadStart+0x70
043cfbc0 00000000 01886fe6 02962998 00000000 ntdll!_RtlUserThreadStart+0x1b


STACK_COMMAND: ~6s; .ecxr ; kb

FAULTING_SOURCE_CODE:
66: #if defined(__INTEL_COMPILER) || defined(DEBUG)
67: #ifdef ALT_BITSTREAM_READER_LE
68: re_cache= AV_RL32( ((const uint8_t *)(gb)->buffer)+(re_index>>3) ) >> (re_index&0x07);
69: #else
> 70: re_cache= AV_RB32( ((const uint8_t *)(gb)->buffer)+(re_index>>3) ) >> (re_index&0x07);
71: #endif
72: #else
73: // ICL9.1-Release and MSVC8-DEBUG build can't process this macro properly.
74: UPDATE_CACHE(re, gb);
75: #endif


SYMBOL_STACK_INDEX: 0

SYMBOL_NAME: ffdshow!get_ue_golomb+39

FOLLOWUP_NAME: MachineOwner

MODULE_NAME: ffdshow

IMAGE_NAME: ffdshow.ax

DEBUG_FLR_IMAGE_TIMESTAMP: 4c02e33e

FAILURE_BUCKET_ID: INVALID_POINTER_READ_c0000005_ffdshow.ax!get_ue_golomb

BUCKET_ID: APPLICATION_FAULT_INVALID_POINTER_READ_ffdshow!get_ue_golomb+39

Followup: MachineOwner
---------

BTW, I really think the /DEBUG should be enabled in linker even for release builds.

Another thing: is this (http://i46.tinypic.com/1h8fnq.png) normal?

EDIT: the weird thing is I cannot get the crash to happen with debug ICL11 builds. Anyway here (http://www.mediafire.com/?mo20wa0j01x) is a sample. The crash happens with ICL11 Release builds.

EDIT2: After some more digging, I don't think the above debug report is corrrect. I tried a few revs, and I found that r3368 works while r3369 doesn't. I really can't see how this can be related to ICL11 but I can reproduce it every time (not with the sample, but with the whole video).

[Delerue]

I doubt "mjpb" is supported.

But AFAIK FFMpeg already supports this format. For example, you can try the FFplay (http://ffmpeg.arrozcru.org/autobuilds/ffmpeg/mingw32/static/) or MPlayer; both can open MJPEG-B videos without any problem.

You can play it in MPC-HC if you install QT frameworks (QT alternative).

Yeah. I know, but would be great if we don't need this anymore. Furthermore, QT doesn't allow you to use any plugin (sharp, deinterlacing, denoise, subtitle, avisynth, etc), and doesn't have a x64 version.

[albain]

@Albain

I think the majority wants to hide the filters from the player's menu. So I suggest to only show the subtitle filter in there. Full settings can still be accessed through "Setup", which should perhaps be renamed to "Settings"?

As I said I have no objection to hide ffdshow internal filters based on that setting however I don't want to hide subtitles/audio streams because it is intended to be used and integrated by third party softwares this is not only a visual matter

This is the reason why I changed that option

I still think that this is a bad idea to disable an interface even partially based on a setting, we cannot implement special behaviours for some players.

So I propose the following solution : the hide/show streams menu (which will be renamed) will only disable ffdshow internal filters from IAMStreamSelect interface (so they won't be visible from external players).
But audio/subtitle streams menu will still be.

About FFDShow context menu, I will separate subtitle streams from subtitle files into 2 different submenus

Sounds okay to everyone ?

[Sebastiii]

Hi,

I think it's a good idea on my side :)
Seb.

[STaRGaZeR]

Sounds okay to everyone ?

Sounds good. Some thoughts:

- The submenus are OK in the tray icon, but in ffdshow context menu they're not.
- Expose video tracks too, if possible. Haali does this, and it'll be useful in the short future for 3D streams.
- Prefix the stream name with V: , A: and S: for video, audio and subtitle tracks respectively, like Haali. This will identify what tracks are what when no track name or no language is set, like in here: http://www.imagebam.com/image/550c2882667998
- Put the submenus in this order: video, audio, sub streams, sub files. Seems to be the standard.

[albain]

Sounds good. Some thoughts:

- The submenus are OK in the tray icon, but in ffdshow context menu they're not.
- Expose video tracks too, if possible. Haali does this, and it'll be useful in the short future for 3D streams.
- Prefix the stream name with V: , A: and S: for video, audio and subtitle tracks respectively, like Haali. This will identify what tracks are what when no track name or no language is set, like in here: http://www.imagebam.com/image/550c2882667998
- Put the submenus in this order: video, audio, sub streams, sub files. Seems to be the standard.

Context menu is the reflection of IAMStreamSelect interface, there is no code behind it and this is up to the implementation of player side.
So graphstudio for example will create a submenu for each group (audio/subs/sub files...)

Video track : not possible right now, ffdshow does not support multiple video input pins (so nogo for MPC splitter unless they plan to use one pin). About Haali's splitter, maybe it will work without changing anything but a reconnection is necessary (reload codecs, presets...). To be tested

Ok to improve the track name but not with prefix : what can be done is replace the track name by the language name when it is empty

[clsid]

I'm getting some crashes with some mkv (h264+ac3) videos, r3363 ICL11 build. MSVC build does not have this crash.
...
EDIT2: After some more digging, I don't think the above debug report is corrrect. I tried a few revs, and I found that r3368 works while r3369 doesn't. I really can't see how this can be related to ICL11 but I can reproduce it every time (not with the sample, but with the whole video).
Make a r3369 build and revert some of the changes made in r3369. That way you can figure out which change causes the issue.

I don't see anything wrong in r3369, so there must be some kind of deeper underlying issue.

[Keiyakusha]

Hi. I just tried ffdshow rev3466 ICL11 and I have a few questions.

1) I still can't access directshow filter settings. rundll32/ffdshow.ax crashing immediately. This is just me?
Now wmv3 and divx video crashing for me and I even can't check settings... EDIT: the same with msvc build
2) See here (http://imgur.com/IWMHE.jpg). Please tell me what and where I should enable or disable to get rid of that unneeded stuff (I mean not only the ones in blue box but everything). Note, that at the same time I do want to see ffdshow internal filters and presets in that menu.

[leeperry]

I still can't access directshow filter settings. rundll32/ffdshow.ax crashing immediately. This is just me?
works fine here.
Please tell me what and where I should enable or disable to get rid of that unneeded stuff
same here, it pick ups all the text files it can find: http://img717.imageshack.us/img717/2183/jmp2.png
the hide/show streams menu (which will be renamed) will only disable ffdshow internal filters from IAMStreamSelect interface (so they won't be visible from external players).
But audio/subtitle streams menu will still be.
so it'll always take two clicks and reaching "setup" to open ffdshow's config? this is not negotiable? :(

what's the point? http://img714.imageshack.us/img714/9153/ffdg.png

I understand you want ffdshow to look good in your Microsoft Media Center plugin, but many ppl don't use subs in ffdshow nor care about the audio stream specs...it'd be great to have an option to have it the exact same way it was before. Please be so kind as to allow hiding all that useless stuff to the ppl who don't care for subs in ffdshow or audio media info..if any possible. :thanks:

PS: and to the guy w/ the l33t nickname, you're on my ignore list so don't bOtHeR.

[albain]

I understand you want ffdshow to look good in your Microsoft Media Center plugin, but many ppl don't use subs in ffdshow nor care about the audio stream specs...it'd be great to have an option to have it the exact same way it was before. Please be so kind as to allow hiding all that useless stuff to the ppl who don't care for subs in ffdshow or audio media info..if any possible. :thanks:


I don't use IAMStreamSelect for media control and I don't understand your reaction, this is just a context menu, the splitters and some decoders have one too ?!! What is the big deal ??

@All : ffdshow is a codec with internal filters based on DirectShow which is a dying technology.
If one want to extend its lifetime, it has to adapt and move forward
I don't think that it will survive alone and it needs better integration with third party softwares.

This is done with media center (see this link (http://damienbt.free.fr)), I am actually doing it for Mediaportal (see this link (http://forum.team-mediaportal.com/submit-code-patches-mediaportal-tv-server-etc-325/integration-ffdshow-82613)), and I wish MPC team will one day drop its filters to integrate FFDShow too (but they are hesitating on this)

This is why I developed and improved FFDShowAPI .Net (which now works in direct access mode through interfaces against interprocess communication). This is for C# coding, and FFDShowAPI with existing C++ interfaces for C++ coding. This is a whole toolset to make integration easier

Also one will have to bring media foundation support as for divx 7

@Leeperry
I think that an - external - context menu is a negligible collateral damage. If you want to remove it, ask kmplayer team for an option to disable streams support. This is not a ffdshow matter because all the codecs are concerned (IAMStreamSelect is a MS interface)

[leeperry]

@Leeperry
I think that an - external - context menu is a negligible collateral damage. If you want to remove it, ask kmplayer team for an option to disable streams support. This is not a ffdshow matter because all the codecs are concerned (IAMStreamSelect is a MS interface)
so you make changes to ffdshow to make your Media Center plugin look good, and who cares if this "negligible collateral damage" fills the screen w/ useless text files in other players, right? End-users are supposed to bother all the media players coders in the world to fix your unique way of viewing things I presume? How about an end-users poll to make things even?

do you find this acceptable? http://img717.imageshack.us/img717/2183/jmp2.png

it was great the way it was before, no submenu bs, no "setup" button....but well, there's no project manager on ffdshow, anyone able to modify the SVN is allowed to do whatever fits his very own project plan.

[Sebastiii]

Hi,

IAMStreamSelect is a MS interface, so to use third party software Albain adapt it and i fully ok with that.
So now it's an KMplayer side to disable it if you don't want this, do you ask KMplayer team ?
Do you try with other codec who use IAMStreamSelect (MS interface) to see what happen ?

Kmplayer has to adapt it :)
Seb.

[leeperry]

Do you learn to speak english?

It's an uncalled change made for personal interests, this I understand...if everyone finds this useless bloat acceptable, fine.

[Sebastiii]

Yes i try :)
I don't think it's personal interests, and i understand that you are not agree with that, maybe you should ask KMplayer and see what is the answer ?
Bye,
Seb.

[albain]

so you make changes to ffdshow to make your Media Center plugin look good, and who cares if this "negligible collateral damage" fills the screen w/ useless text files in other players, right? End-users are supposed to bother all the media players coders in the world to fix your unique way of viewing things I presume? How about an end-users poll to make things even?

do you find this acceptable? http://img717.imageshack.us/img717/2183/jmp2.png

it was great the way it was before, no submenu bs, no "setup" button....but well, there's no project manager on ffdshow, anyone able to modify the SVN is allowed to do whatever fits his very own project plan.

You don't even read my posts : I told that I don't use this interface for media control

I am tired of non constructive criticisms, you don't represent the majority, and if you want to make a poll be my guest.

Last thing : the reason why you complain about the lack of FFDShow roadmap with project manager (although there is a manager, clsid) is the same reason why I am doing this

I have brought my reasons, please reread them carefully before posting again the same thing and especially this one : this is a player matter, other directshow players handle it fine (MPC, mediaportal) why not this one

Once again I don't like your tone, you may be upset but I don't appreciate this kind of talking at all

[zummi]

Serious trouble with rev 3466:

The 20th Century Fox intro how it should look:
http://i47.tinypic.com/2h88wso.jpg

And with rev 3466:
http://i49.tinypic.com/dyq8tv.jpg

This happens both with and without post-processing enabled. Integrated gfx: Intel GMA 4500MHD.

**** this is now fixed :)

[STaRGaZeR]

Context menu is the reflection of IAMStreamSelect interface, there is no code behind it and this is up to the implementation of player side.
So graphstudio for example will create a submenu for each group (audio/subs/sub files...)

Video track : not possible right now, ffdshow does not support multiple video input pins (so nogo for MPC splitter unless they plan to use one pin). About Haali's splitter, maybe it will work without changing anything but a reconnection is necessary (reload codecs, presets...). To be tested

Ok to improve the track name but not with prefix : what can be done is replace the track name by the language name when it is empty

Ah, I see. So I assume the tray icon takes the info from the IAMStreamSelect interface and then puts it on screen like ffdshow wants, and we can change that.

The MPC splitter should implement a stream switcher like Haali, and only expose one pin per stream type. But there are no volunteers for that unfortunately :(. Looking forward to test whatever you have in mind.

I'm fine with the language as name. However it'll look strange when both are not set, like Undefined [Undefined].

BTW, don't feed the troll.

[clsid]

it was great the way it was before, no submenu bs, no "setup" button....but well, there's no project manager on ffdshow, anyone able to modify the SVN is allowed to do whatever fits his very own project plan.Nonsense. Bad changes can easily be reverted if needed. But there is absolutely no need for that. Albain is doing a great job and invests a lot of time in ffdshow. I wish we would have more devs like him. If you don't like his changes then you are just out of luck. Criticism is always welcome if it comes with good argumentation. You are just whining like a little child.

[clsid]

@Albain
Could you make it so that only subtitles that match the filename are shown? As Keiyakusha's screenshot shows, it can be annoying if there are many unrelated sub files in the same folder as the current video. Imho, people should load wrongly named subtitles manually through ffdshow subtitle config.

VSFilter uses this scheme:
moviename[.subtitlename].extension


Do you have any idea what "heuristic search" does?

[clsid]

Hi. I just tried ffdshow rev3466 ICL11 and I have a few questions.

1) I still can't access directshow filter settings. rundll32/ffdshow.ax crashing immediately. This is just me?
Now wmv3 and divx video crashing for me and I even can't check settings... EDIT: the same with msvc buildNo problems here. Could you try my MSVC build?
http://sourceforge.net/projects/ffdshow-tryout/files/

[albain]

@Albain
Could you make it so that only subtitles that match the filename are shown? As Keiyakusha's screenshot shows, it can be annoying if there are many unrelated sub files in the same folder as the current video. Imho, people should load wrongly named subtitles manually through ffdshow subtitle config.

VSFilter uses this scheme:
moviename[.subtitlename].extension


Do you have any idea what "heuristic search" does?

Thanks buddy ;-)

Heuristic search is based on matching probabilities (Baylasian approach) : if your video file is named "Grand canyon.avi" and your subtitle file "GrUnd Canyon.srt" it will select it automatically (if of course no identical name can be found)


About the vsfilter scheme, I think that this is a good idea, Sebastii and I are thinking about it and I wondered it would be better to implement it on the player side or in ffdshow.

I think we can add one of the following 2 methods :
1/ retrieve all the heuristic searches (heuristic does autoselect for now, it does not make a list) based on a minimal matching percentage
2/ retrieve all the <file name>.suffix.[srt|ssa|ass...]

Which one ?

[Octo-puss]

Could someone explain the difference between libavcoded and ffmpeg please? I do not really know which and why to choose in video decoder settings.

[Atak_Snajpera]

ffmpeg-mt is multi-threaded, libavcodec not really .

[albain]

livavcodec = ffmpeg, always

For H264 you can choose libavcodec (ffmpeg) or ffmpeg-mt which is the multithreaded version of ffmpeg which may be less stable, but I have no problem using it considering the significant gain in CPU usage

[Keiyakusha]

Oh by the way I just remembered something I want to ask. Is it possible to implement queue output samples feature for EVR renderer? Or this won't be useful?

EDIT: also I resolved my problem with crash by rewriting all settings from scratch... But not sure what exactly was the cause of this.
EDIT2:
1/ retrieve all the heuristic searches (heuristic does autoselect for now, it does not make a list) based on a minimal matching percentage
2/ retrieve all the <file name>.suffix.[srt|ssa|ass...]
Personally I like 2nd, this is seems to be what I proposed earlier. But lets see what some other users think
Also in the way how it is now, this is not just annoying. For example I have some series with 70 episodes and each of them have 2 subtitles (forced and normal). Scrolling here and there to find stream I need takes minutes.

[clsid]

Thanks buddy ;-)

Heuristic search is based on matching probabilities (Baylasian approach) : if your video file is named "Grand canyon.avi" and your subtitle file "GrUnd Canyon.srt" it will select it automatically (if of course no identical name can be found)


About the vsfilter scheme, I think that this is a good idea, Sebastii and I are thinking about it and I wondered it would be better to implement it on the player side or in ffdshow.

I think we can add one of the following 2 methods :
1/ retrieve all the heuristic searches (heuristic does autoselect for now, it does not make a list) based on a minimal matching percentage
2/ retrieve all the <file name>.suffix.[srt|ssa|ass...]

Which one ?
How about doing #2 and fall back to #1 if no match is found?

[Delerue]

how about doing #2 and fall back to #1 if no match is found?

+1 :)

[Keiyakusha]

How about doing #2 and fall back to #1 if no match is found?

If implementing both variants possible, IMHO better to add an option and let the user decide. When there is no subs, falling back to heuristic can also pick up something that shouldn't be there.

[clsid]

Heuristic should not auto-select the subtitle. It should only show a list of available subs that closely match the filename of the video.

[leeperry]

But there is absolutely no need for that.
hehe, I guess you are the "boss" of ffdshow so you make the decisions for everybody. Glad to see that we don't agree, just like when it took you 2 years to update libsamplerate...I had been "whining" about this for years duh.

It was perfect the way it was before, no useless submenu(full of whatever text files it can find), no useless "setup" button requiring 2 buttons to open ffdshow's config...this is just very annoying, how good of an argumentation is that? I understand you need active coders so it's good that some of them are willing to make pointless changes that follow their very own agenda...oh my oh my, all that bloat must look awesomeness in his donationware Media Center plugin. Majority rules! Albain please keep adding bloat in ffdshow, how about some google ads?
You don't even read my posts
I can't seem to be able to decipher your approximative english http://forum-images.hardware.fr/images/perso/souk.gif

[clsid]

Your opinion has become totally irrelevant. Your attitude just causes people to ignore you.

I strongly recommend stopping the trolling or you might end up getting a strike from a mod.

[pankov]

How about doing #2 and fall back to #1 if no match is found?

If it's not too much work I support Keiyakusha's position - add an option and let the user decide if he wants heuristic enabled at all.
And only if it's enabled use it after unsuccessful attempt of finding subs with the other method. ... or to have a full functionality you can add an option for the user to select which method is used first

[Astrophizz]

Anyways, at least from my perspective, you don't have to click twice. You hover then click once, which isn't such a big deal in the grand scheme of things.

[avih]

leeperry, I second clsid about your attitude issue. It is not welcome here. Please keep that in mind.

[dimitrik]

@All : ffdshow is a codec with internal filters based on DirectShow which is a dying technology.
If one want to extend its lifetime, it has to adapt and move forward
I don't think that it will survive alone and it needs better integration with third party softwares.

.........


Also one will have to bring media foundation support as for divx 7



Albain - first of all thank you very much for your work on ffdshow.

I want to get into this question of whether ffdshow can survive alone or not.
I'm not sure if I misread you but are you saying that you don't believe in ffdshow's future as a standalone framework, or are you saying that you believe it should become the "decoding framework of choice" for any player?

From a technology strategy perspective, I would disagree with the first but agree with the second. But you don't want its future tied to the success or failure of other softwares...

[albain]

Sorry if I was not clear enough, I meant the second

The integration as the privileged decoding & postprocessing solution into 3rd party players will guarantee the future and why not may bring reinforcement to our team.

When you see what xmbc project has achieved, it is not very different from ffdshow (except that it does not use directshow) : they use ffmpeg for decoding and postprocessing (deinterlace, noise reduction, upscaling...) and bring a front end to configure all of this

The differences is that we support more codecs, more internal filters, we have HD audio bitstreaming, better subtitles support...

Also ffdshow GUI is too complex in my opinion, it should be an "advanced user mode" that is not enabled by default against a beginner mode much lighter.

Last but not least, the current interfaces that FFDShow implement are able to get/set any setting in any preset or any live setting so all the APIs are there to facilitate this integration (in C++ or C# code)

[madshi]

ffdshow is a codec with internal filters based on DirectShow which is a dying technology.
Once upon a time MS planned to replace the native win32 APIs completely with DotNet. At that time some people used to talk about native win32 programming as a dying breed, too. Now look what the Media Foundation introduction says:

Media Foundation [...] requires C/C++. Microsoft does not provide a managed API for Media Foundation.
Ironic?

Probably in 5-10 years from now MS will introduce a new media related software and in its introduction will say:

blabla requires DirectShow. Microsoft does not provide a Media Foundation API for blabla.
History tends to repeat itself.

[albain]

You may be right, maybe directshow support won't be dropped very soon but MS is migrating all its codecs and new ones to media foundation.
This is annoying but we have to add support for Media Foundation. That does not mean dropping directshow, but simply add the missing interfaces to support both.

The tough part will be to make it not too hard to maintain, but we did it with DXVA, although it is more complicated here



About subtitle files : I committed some changes that should suit all needs
There is a dropdown box in the tray/icons section to select the way the subtitle files are retrieved. This affects both the tray icon menu and the IAMStreamSelect interface (external menus).
4 options : all subtitles files that can be found, subtitle files with exact matching on the video filename (<video>.ext or <video>.<suffix>.ext), heuristic matching (all the subtitle files with similarities in the file name based on a default tolerance), and lastly a mix of the 2 previous methods : list with just the exact matching names or if nothing found an heuristic list is returned

[madshi]

You may be right, maybe directshow support won't be dropped very soon but MS is migrating all its codecs and new ones to media foundation.
Their codecs will still be accessible from within DirectShow, or won't they? At least the VC-1 decoder is on win7, and that's the only MS DirectShow filter I really care about. All other DirectShow filters I'm using are non-MS, anyway.

[SamuriHL]

DirectShow isn't going away any time soon. Yea, media foundation classes are (currently) the "next great thing" but I only know of a handful of commercial players that support it so far. The one that springs to mind is WinDVD. I'm sure there could be others. Dropping DirectShow for anyone isn't really an option, so, it's good to see you say that's not the intent. Bolting on the MF API support will definitely pose some unique challenges. Fortunately I personally believe you have plenty of time to jump on that bandwagon and do it correctly. Don't hack it in just to get it done, IOW. :) Plan it out and do it the right way so it's easy to maintain.

[Sebastiii]

You may be right, maybe directshow support won't be dropped very soon but MS is migrating all its codecs and new ones to media foundation.
This is annoying but we have to add support for Media Foundation. That does not mean dropping directshow, but simply add the missing interfaces to support both.

The tough part will be to make it not too hard to maintain, but we did it with DXVA, although it is more complicated here



About subtitle files : I committed some changes that should suit all needs
There is a dropdown box in the tray/icons section to select the way the subtitle files are retrieved. This affects both the tray icon menu and the IAMStreamSelect interface (external menus).
4 options : all subtitles files that can be found, subtitle files with exact matching on the video filename (<video>.ext or <video>.<suffix>.ext), heuristic matching (all the subtitle files with similarities in the file name based on a default tolerance), and lastly a mix of the 2 previous methods : list with just the exact matching names or if nothing found an heuristic list is returned

Hi,

Very cool :)
Thank you so much :)
Seb.

[mrcorbo]

Are there any advantages to the new structure/interfaces offered by Media Foundation that would allow for better/cleaner implementations of existing ffdshow functions or would this support be added purely for added compatibility and future-proofing?

[SamuriHL]

I'm wondering if media foundation classes add any stability over DirectShow? I doubt it, but, if anyone has any insight to that question I'd appreciate it. What I mean is, does using media foundation classes offer better compatibility with other codecs that use it, etc...I can see the benefit of adding it to ffdshow since MS seems hellbent on making it the new standard.

[Eliminateur]

are there any plans to fix the totally broken deband filter in x64 that has been reported several times?

[albain]

Media foundation is supposed to be more stable on the audio/video synchronization part

Also it is announced as easier to develop against directshow

[Mike5]

Media foundation is supposed to be more stable on the audio/video synchronization part

What does this means, do they changed the reference clock mechanism ?

An improvement in this direction IMHO would be worth a conversion, because I think the biggest problem in PC players vs stand-alone players is still the fight between the sound card clock and the video card clock (or others clocks in the systems).

[Qaq]

...I think the biggest problem in PC players vs stand-alone players is still the fight between the sound card clock and the video card clock (or others clocks in the systems).Hmm...if i use onboard video with HDMI audio, i got video and audio perfectly matched?

[Mike5]

It's the same thing. Modern video cards with HDMI or motherboards with HDMI have a built-in audio card (actually a chip, Realtek generally) to feed the audio part of HDMI. So you still have (at least) two clocks: the video card clock and the Realtek chip audio clock.

[ranpha]

What does this means, do they changed the reference clock mechanism ?


Media Foundation players can use MMCSS (http://msdn.microsoft.com/en-us/library/ms684247.aspx) to ensure stable, jitter-free playback.

And of-course, the MFT-initiated DXVA playback is another obvious Media Foundation advantage that DirectShow doesn't have.

[MatMaul]

Media Foundation players can use MMCSS (http://msdn.microsoft.com/en-us/library/ms684247.aspx) to ensure stable, jitter-free playback.
like any player, there is an API for that.

[nevcairiel]

Writing filters and players for Media Foundation is alot easier then writing for DirectShow. You really do feel that it was designed for proper media playback, and alot was learned from the days of DirectShow. Someone just has to start, and provide a splitter for MediaFoundation, so we can at least play our MKVs using the MS codecs, and then maybe more people will join and provide codecs and players.

I wrote a *very* basic proof of concept MF player some weeks back. It basically can open a file and play it, just to get a feel for the APIs. Maybe once this splitter project i'm working on gets into a better shape, i can add a MF layer on top of it .. abstraction ftw.

[_xxl]

I wrote a *very* basic proof of concept MF player some weeks back. It basically can open a file and play it, just to get a feel for the APIs. Maybe once this splitter project i'm working on gets into a better shape, i can add a MF layer on top of it .. abstraction ftw.
You could share your code...

[Mixer73]

DirectShow was essentially a hack, I have some insight into its development and it wasn't so much planned as it just rose from the very early capture cards...

[Delerue]

Sorry to interrupt this good discussion, but about the last subtitles changes, I think that there's a bug in the automatic selection of the subtitle (i.e. external or embedded). See:

http://img.photobucket.com/albums/v256/Delerue/Contextmenu.png

Both embedded and external subtitles are selected, but by default you can actually only see the internal in the screen; if you select the external, than you'll see it in the screen. I think that external subtitles should receive higher priority over embedded. What you think?

[albain]

Right there is a bug : both embedded and external should not be selected

However I disagree with you : embedded should be taken by at first over external subtitles.
This is the adopted behaviour on all players

Fixed in rev 3471

[pankov]

albain,
I kind of disagree with you.
To have external subtitles file means that someone needed them, made the effort of placing them in the same folder and naming them in the same way as the movie.
On the other hand embedded subs are present without any work.
Usually having external subs meant that the internal ones didn't work (not the correct languages) for the user and he needed an external one so why make it even harder for him and make him select them manually instead of doing it automatically for him?
at least that's what usually happens to me.
And one more thing - I don't know which players do you have in mind but I'm using Zoomplayer with VobSub (VSFilter) and it plays the external ones with priority.

[clsid]

I have to agree with pankov. If a user has added an external sub to a file that already has embedded subs, then it is likely that the user prefers the external one.

[albain]

You're right, I didn't see that way.

This behaviour is historic though but it should not be too difficult to change

*EDIT* :Fixed in revision 3472

[Sebastiii]

hi all :)

It works really great :)
Thank you man :)

[SamuriHL]

Right. They can always rename or remove the external sub if they want the internal by default. I agree, as well, that externals should take priority.

[fastplayer]

When you mux a sub into a file, you expect that sub to be loaded when starting playback and not some other external one. Why put it there in the first place then?! Having an identical/similar-named external file essentially negating all the reasons for an embedded one, is illogical and anything but practical.

[SamuriHL]

It's easier to remove the external sub from the directory than it is to eliminate an internal sub, yea?

[Keiyakusha]

I also agree, external should be preferred.
By the way MPC-HC has similar problem with internal renderer and if container has subtitles but they not enabled by default then it shows nothing and you have to load subtitles manually.

[albain]

I fixed that priority in revision 3742 : now subtitle file if any is loaded first

@Stargazer

Recently you changed the subtitles font size management in order to be constant across different video resolutions.

Someone reported me the following :
"The change made to get a fix font size for subtitles doesn't work if "letterbox" is checked, I'm using it in order to display subtitles below the movie when the video height allow it !
It's worse using ffdshow DXVA with letterbox enabled, subtitles are not displayed at all "

Do you have an idea of the problem or could you have a look at it ? Thanks :thanks:

[STaRGaZeR]

I just tested it and the font size is constant if you change resolution while maintaining letterbox checked, but I guess he is talking about between letterbox on and off, the font size is indeed smaller with letterbox enabled. Can you ask him about this? The font size is calculated based on original video dimensions, not video dimensions after the resizing done by letterboxing. Should be easily fixable if this is the case.

About DXVA, it's normal. ffdshow can't resize the video here, I think the problem is that the subtitle filter thinks it has been resized and puts the subs below the actual video. Wouldn't it be better to ignore letterbox if DXVA is on since we can't touch the decoded video?

BTW, now that you mention it, do you know how to fix this (http://img375.imageshack.us/img375/1369/83497053.png)? It got broken some time ago and I don't know how to do it.

[Delerue]

Great. Thanks, guys. Albain, as soon as a compiled build is released, I'll test it. ;)

Another silly question: why 'Resize & aspect' option changes the colorspace result, even if you don't use any Luma method? I know this can be fixed by selecting RGB32 HQ conversion. I just want to know why this thing happens.

[Keiyakusha]

Delerue
Do you mean this?

1) you have YV12 SD video, it is resized to HD and output in ffdshow set to RGB
ffdshow knows that source is SD and uses appropriate coefficients for YUV-RGB conversion
everything looks as it should
2) you have YV12 SD video, it is resized to HD and output in ffdshow set to YV12 (or other YUV...)
after resizing SD to HD, colorimetry should be corrected which is NOT done by ffdshow.
video renderer doesn't knows that video uses SD coefficients and applies HD ones, which is wrong ==> we have wrong colors.

If you mean this, then to fix that, colormatrix avisynth plugin should be ported to ffdshow and applied automatically based on what resizing is done. Alternatively you can use plugin by yourself: for SD->HD resize this should be applied to the video: ColorMatrix(mode="Rec.601->Rec.709")

[DrKnowLittle]

I fixed that priority in revision 3742 : now subtitle file if any is loaded first



:thanks:
I been waiting for this ...
http://forum.doom9.org/showthread.php?p=1208815#post1208815

:eek:

DrKnowlittle

[mingleberry]

Hi

Great forum - but can anyone help me?

I'm trying to use FFDShow to decode my MPEG2 MKVs ripped from my DVD collection in order to utliise its upscaling/post-processing capabilities.

I'm using Win7 Filter Tweaker 3.1 to change the codec from Microsoft's to FFDShow 3316 but when I try to play an MPEG2 MKV in 7MC or WMP, MPC-HC or Graphedit,, the application crashes. Other media types all work fine with FFDShow - H264 and VC1 blu ray rips, XVID DVD rips. It's just the MPEG2 rips that are causing this problem.

I'm using Haali 1.9.42.1 if that helps.

I've checked FFDShow is set to decode MPEG2 etc etc. Is there something else I could check to make this work?

The rips work fine with MS or MPC decoders incidentally.

Thanks

Dom

[mremulator]

Hi

Great forum - but can anyone help me?

I'm trying to use FFDShow to decode my MPEG2 MKVs ripped from my DVD collection in order to utliise its upscaling/post-processing capabilities.

I'm using Win7 Filter Tweaker 3.1 to change the codec from Microsoft's to FFDShow 3316 but when I try to play an MPEG2 MKV in 7MC or WMP, MPC-HC or Graphedit,, the application crashes. Other media types all work fine with FFDShow - H264 and VC1 blu ray rips, XVID DVD rips. It's just the MPEG2 rips that are causing this problem.

I'm using Haali 1.9.42.1 if that helps.

I've checked FFDShow is set to decode MPEG2 etc etc. Is there something else I could check to make this work?

The rips work fine with MS or MPC decoders incidentally.

Thanks

Dom
You need to disable the Microsoft DTV-DVD Audio decoder and Media Foundation using the Win7 Filter Tweaker. Also if you have a 64-bit OS, you need to do the same for the 64-bit codecs.

[clsid]

His problem also occurs with MPC and GraphEdit, both of which do not use Media Foundation.

[mremulator]

His problem also occurs with MPC and GraphEdit, both of which do not use Media Foundation.

True.

However, 7MC and WMP do... which will crash if the aformentioned are not disabled. No idea why MPC is crashing out though?

Perhaps he should enable the internal MPEG2 decoder in MPC and see if that helps?

[Peuj]

Hi,

When I load subtitles with ffdshow my CPU is 100% used after a seek.
I use ffdshow v.3467 and MPC-HC v.1976 from K-Lite Codec Pack.

When I first play the video everything is ok but if I seek forward the CPU go to 100% because of MPC-HC.
Even if I pause the video, the CPU is still used at 100%.

If I start the video without activate the subtitle from ffdshow I can seek and everything is normal but I activate then the subtitle the CPU go to 100%.

Should I configure/disable something?

[Delerue]

2) you have YV12 SD video, it is resized to HD and output in ffdshow set to YV12 (or other YUV...)
after resizing SD to HD, colorimetry should be corrected which is NOT done by ffdshow.
video renderer doesn't knows that video uses SD coefficients and applies HD ones, which is wrong ==> we have wrong colors.

I see. Maybe my question is: why HD has a different colorspace in comparison with SD? What's the specific resolution that changes the colospace and why?

Thanks

[DeathTheSheep]

Bug (either in ffdshow or ffmpeg):
Changing the I-frame multiplier or offset in the MPEG-4 VfW encoder has no effect in 1-pass mode (Quality or Quant). Note that masking + normalization are enabled.

Maybe check to see if it sets the right parameter according to lavc's api--if everything checks out, it's an ffmpeg problem and you shouldn't worry about it. If the parameter was only intended for 2-pass mode, perhaps consider graying out the box if 1-pass is selected.

Cheers!

[mingleberry]

Hi

Thanks for all the advice. Disabling WMF etc does indeed prevent the crash for some movies but others (such as Bruno) still have the problem, particularly with FFD in 7MC (I accept the FFD bug could be something different though).

(By the way, I only mentioned that the prob occurs in MPC to demonstrate it's not a bug peculiar to MS players. It may be sacrilege in these parts but I actually want to use 7MC as my main player.)

[albain]

Hi Stargazer, thanks for your reply and sorry for the delay

Here are the last results that were reported to me :

- Font autosize off + letterbox on : the font size varies accross video dimensions (normal)
- Font autosize on+ letterbox on/off : as you said the font size varies accross video dimensions

I'll try to disable letterbox in DXVA mode because it is useless

[Peuj]

Hi,

When I load subtitles with ffdshow my CPU is 100% used after a seek.
I use ffdshow v.3467 and MPC-HC v.1976 from K-Lite Codec Pack.

When I first play the video everything is ok but if I seek forward the CPU go to 100% because of MPC-HC.
Even if I pause the video, the CPU is still used at 100%.

If I start the video without activate the subtitle from ffdshow I can seek and everything is normal but I activate then the subtitle the CPU go to 100%.

Should I configure/disable something?

I've found that it is related with the option Subtitle -> Font -> Autosize.
If I uncheck it, I don't have the problem anymore.
So I use the "Scale" parameters instead.

Am I the only one with this problem? Is it known ?

Thanks

[STaRGaZeR]

OK albain, will look into it next week.

I've found that it is related with the option Subtitle -> Font -> Autosize.
If I uncheck it, I don't have the problem anymore.
So I use the "Scale" parameters instead.

Am I the only one with this problem? Is it known ?

Thanks

fastplayer (http://forum.doom9.org/showthread.php?p=1398061#post1398061) reported a similar one earlier, but he didn't test autosize. Do you have the issue with r3426 or earlier? Please test before and after r3427.

[fastplayer]

fastplayer (http://forum.doom9.org/showthread.php?p=1398061#post1398061) reported a similar one earlier, but he didn't test autosize.
My particular 100%-CPU issue appeared way before any changes were made to the "Autosize" feature. Peuj's issue has to be a different one with the same symptom. Although, I noticed a pattern: When pausing a lot during playback, it is more likely to happen than when playing from start to finish without any pauses.

[Peuj]

Hi,

I've made some tests and I don't have the issue with 3422 and 3427.
With 3432, after some seek forward the CPU go to 100%. But if I pause the video, the CPU usage go down.
The symptoms are a little different because with 3467, the CPU go to 100% as soon as I seek forward and if I pause the CPU is still used at 100%.

[STaRGaZeR]

My particular 100%-CPU issue appeared way before any changes were made to the "Autosize" feature. Peuj's issue has to be a different one with the same symptom. Although, I noticed a pattern: When pausing a lot during playback, it is more likely to happen than when playing from start to finish without any pauses.

Any progress with that issue? I can't reproduce, and if I could I probably wouldn't know how to fix it :(

Hi,

I've made some tests and I don't have the issue with 3422 and 3427.
With 3432, after some seek forward the CPU go to 100%. But if I pause the video, the CPU usage go down.
The symptoms are a little different because with 3467, the CPU go to 100% as soon as I seek forward and if I pause the CPU is still used at 100%.

Try with this build: http://www.megaupload.com/?d=PBY9QGIG

Sorry for the megaupload link, mediafire is down.

[kieranrk]

Can we have ffmpeg's internal aac decoder instead of faad now? It's significantly faster than faad is feature complete for the most common aac profiles. It also supports channel switches better than faad does

Thanks.

[fastplayer]

Any progress with that issue?
If by progress you mean using MPC's subtitle renderer, then yes :D
The day before yesterday, I switched to ffdshow for subs and everything was fine for the entire day. Yesterday, on the other hand, not so much... The bug appeared twice. Steps to reproduce failed, of course. :rolleyes:
I can rule out the parser, though. First, I thought ffdshow was stumbling over some Unicode characters or some HTML tags but after reformatting everything to ANSI and stripping out any tags, this issue still keeps popping up.
My guess is that this is some threading issue: A deadlock, race condition or something similar.
I can't reproduce, and if I could I probably wouldn't know how to fix it :(
At least, you're listening :D

[STaRGaZeR]

My guess is that this is some threading issue: A deadlock, race condition or something similar.

*Runs and hides* :D

Have you tested old revs? If not, it's gonna be a pain in the ass, haruhiko made his changes like what, 2000 revisions ago?

[fps]

I think there is a bug in the VC-1 decoder:
With DXVA enabled everything works fine. The internal non DXVA VC-1(FFmpeg) decoder of MPCHC (build 2018) also works fine.
But whenever I try to play a VC-1 video (1080p Blu-Ray) with ffdshow(libavcodec) and DXVA disabled, playback becomes very stuttering.
I've tried various ffdshow revisions the last one is 3472. Changing video renderers doesn't help either (EVR-CP,Haali,madVR).
I'm sure my PC is fast enough to play 1080p VC-1 Blu-Rays since with the MPCHC decoder works smooth with CPU usage at about 25%.

My specs:
Win7/32, Aero on
Core2Quad9400s
ATI HD 4830
4GB DDR3
ffdshow rev 3472
MPCH svn 2018

Anybody else got the same problem?

[fastplayer]

*Runs and hides* :D

Have you tested old revs? If not, it's gonna be a pain in the ass, haruhiko made his changes like what, 2000 revisions ago?
I think my chances of making passionate love with one of Victoria's Secret Angels (preferably Doutzen Kroes) are a gazillion times higher than me finding the very first buggy rev :D

What we need is someone who reads process/memory dump files for a living. This would be the fastest and most effective way of dealing with this issue. Until then, I have to stick with the subpar alternative... :)

[Sebastiii]

I think there is a bug in the VC-1 decoder:
With DXVA enabled everything works fine. The internal non DXVA VC-1(FFmpeg) decoder of MPCHC (build 2018) also works fine.
But whenever I try to play a VC-1 video (1080p Blu-Ray) with ffdshow(libavcodec) and DXVA disabled, playback becomes very stuttering.
I've tried various ffdshow revisions the last one is 3472. Changing video renderers doesn't help either (EVR-CP,Haali,madVR).
I'm sure my PC is fast enough to play 1080p VC-1 Blu-Rays since with the MPCHC decoder works smooth with CPU usage at about 25%.

My specs:
Win7/32, Aero on
Core2Quad9400s
ATI HD 4830
4GB DDR3
ffdshow rev 3472
MPCH svn 2018

Anybody else got the same problem?

I can confirm the same, i have remark this too. And i don't investigate more on this.
Maybe we can nail it but make log but don't sure.
Seb.

[nevcairiel]

CPU Usage of 25% on a Quad-Core means that its at 100% of one core. I would guess the VC1 decoder is not multi-threaded, so its basically capped already. A ever so slight performance decrease can then cause stuttering.

[Sebastiii]

Good point :)
Cheers,
Seb.

[fps]

CPU Usage of 25% on a Quad-Core means that its at 100% of one core. I would guess the VC1 decoder is not multi-threaded, so its basically capped already. A ever so slight performance decrease can then cause stuttering.
Unfortunately I don't think that's the cause of the problem at all.
The maximum usage of one of the 4 cores is about 50%, so there should be plenty room left. Also the same processor can handle every H.264 Blu-Ray in software decoding mode and even the internal MPCHC VC-1 decoder gives me perfect stutterfree playback.
In addition to that the stuttering is kind of constant and doesn't change depending on the actual bitrate of the scene.
So I really think it's a bug in the ffdshow implementation.
But as far as I know the MPCHC decoder and the ffdshow one shouldn't be too different, should they?

[clsid]

Change the renderer in MPC-HC. EVR Custom has some bugs with VC-1 when decoded by the Microsoft decoder or ffdshow.

[_xxl]

I have reported some time ago when I seek in some samples (xvid + mp3) I get this:
http://i49.tinypic.com/servjq.jpg
No subtitles.
http://i46.tinypic.com/f0c8hx.jpg
With external srt subtitle. (ffdshow's internal subtitle renderer is used)
Is this your case?

[fps]

Change the renderer in MPC-HC. EVR Custom has some bugs with VC-1 when decoded by the Microsoft decoder or ffdshow.

Hmmm, thanks for the info.
I've now tried all remaining renderers and only Overlay, EVR (non CP)/EVR Sync seem to work.
Do I understand you correctly that the error is just on the renderer's side?
If that's the case I'd let madshi know, cause I really would love to be able to render everything with madVR.

[Killerattacks]

Using Win 7 (64 Bit) and the latest ffdshow no player will load ffdshow's audio decoder when I select the encode only 5.1 to AC 3 feature, when playing a file with 5.1 flac/aac/etc. audio. SPDIF output for AC3 and DTS is selected.
When I disable the encode to AC3 option MPC-HC uses ffdsho for the FLAC decoding (instead of some low merit FLAC decoder). Tried everything to figure this out to no avail.

[Matching_Mole]

Hmmm, thanks for the info.
I've now tried all remaining renderers and only Overlay, EVR (non CP)/EVR Sync seem to work.
Do I understand you correctly that the error is just on the renderer's side?
If that's the case I'd let madshi know, cause I really would love to be able to render everything with madVR.

The bug also occurs for EVR Sync and not only for FFdshow and Microsoft decoders but also for Cyberlink and Arcsoft ones (so all the external VC1 decoders).

It's not a renderer error but a splitter error. I discussed with Testuo55 and unfortunately they don't plan to solve it with the current splitter. They want to use the ffplay m2ts splitter instead but this need time to make the needed adaptations to use it in MPC and it seems that they are missing developers to make the work.

IMHO, I can understand the view of Tetsuo55 but it's sad that this error is present in MPC "HC" (so a player dedicated to the HC experience) for one year or two will be not solve until one year again at least.

[clsid]

It is not really a splitter error. If it works with other renderers, then I see no reason why it shouldn't be possible to make it work with the modified EVR renderers. In fact, the pre-beliyaal EVR CP worked ok with VC-1.

[Matching_Mole]

It is not really a splitter error. If it works with other renderers, then I see no reason why it shouldn't be possible to make it work with the modified EVR renderers. In fact, the pre-beliyaal EVR CP worked ok with VC-1.

Using Arcsoft splitter or Haali splitter, VC1 is working correctly with the EVR Sync in MPC as far as I know. So the splitter seems to be a part of the issue.

[Matching_Mole]

I quickly made a test and I confirm that the bug is not present (for both EVR cust and EVR Sync) using Arcsoft splitter ("Arcsoft MPEG Demux").

[lordadmira]

Greets all.

I just upgraded to the rev 3466 ffdshow and now the vdf plugin does not work with VirtualDub or VirtualDubMod. When selecting it and setting postprocessing, the video is not affected. Clicking About shows that it is the new ffshow that is putting up the window. Downgrading to my old ffdshow of 2546 makes the VDM plugin work again. Any ideas?

edit: Some new info. I located and installed rev 1082 of tryouts and it has the same problem. I did some experimenting and it seems the other filters work except for Postprocessing. E.g. when I play with the levels I see the image change. No matter what settings I use under PP nothing happens. The build 2546 I've been using seems not to be tryouts. It doesn't really say who made it but the file name is ffdshow-2546-gcc4.0.3-sse.exe. So why doesn't PP work with tryouts builds?


LA

[Peuj]

Any progress with that issue? I can't reproduce, and if I could I probably wouldn't know how to fix it :(



Try with this build: http://www.megaupload.com/?d=PBY9QGIG

Sorry for the megaupload link, mediafire is down.

Hi STaRGaZeR,

This build has fixed my problem. Thanks

[SamuriHL]

So, question about ffdshow and subtitle support. I've FINALLY gone to the dark side and started converting my stuff to MKV. I'm using the beta of MakeMKV right now which rocks since it embeds the PGS subtitles into the MKV. However, I can't seem to get ffdshow to notice that they're in there. Is this an issue with the splitter that comes with MPC-HC? If so I'm screwed on that one then. ffdshow seems to recognize the sup files just fine if I throw them in a subtitle sub-folder and point it there. HAALI's not an option for me as it breaks bitstreaming. I want so badly to get MKV's working as well as my movie only ISO's and this is the only remaining problem. The MKV's absolutely rock and MPC-HC+ffdshow makes it a breeze to play back on any of my machines even over the network. Almost in heaven. :)

[SeeMoreDigital]

I'm using the beta of MakeMKV right now which rocks since it embeds the PGS subtitles into the MKV. Really...

Are you able to provide a short sample please? I'd like to test in my hardware players.


Many thanks

[SamuriHL]

Really...

Are you able to provide a short sample please? I'd like to test in my hardware players.


Many thanks

At some point I might be able to. At the moment I'm trying to get all my movie only ISO's converted which is extremely time consuming. But since MKV is a more efficient container, I'm actually reclaiming a bunch of space. Gotta love that. :) In any case, I have verified that the PGS subs are in there using media info. You can try MakeMKV out easy enough just by downloading it and installing it. It's very easy to use. They offer a 30 day trial. I'll probably end up buying it.

[Inspector.Gadget]

SamuriHL: ffdshow works with external, unmuxed BD-Sups?

[SamuriHL]

Why yes, yes it does. :) Albain rocks! I just throw them all in a subfolder called subtitles and in ffdshow's subtitle settings I set the directory to .\subtitles. I'm then able to pick the one I want and I'm good to go. So, I may have to extract them after the MKV is created and drop em in the subtitle folder for now until I can find a solution to the splitter issue.

[SamuriHL]

Ok, so, I put Haali on my laptop and ran one of my new MKV's with the PGS subs embedded through it. Haali sees the sub and splits it for ffdshow just fine. So, MPC-HC's internal splitter is what's not working for PGS subs. This sucks because to date bitstreaming with Haali isn't so good. I'm open to advice on what to do. Also remember that I'm doing everything 64 bit, so, that complicates things slightly.

[foxyshadis]

Unfortunately I don't think that's the cause of the problem at all.
The maximum usage of one of the 4 cores is about 50%, so there should be plenty room left. Also the same processor can handle every H.264 Blu-Ray in software decoding mode and even the internal MPCHC VC-1 decoder gives me perfect stutterfree playback.
In addition to that the stuttering is kind of constant and doesn't change depending on the actual bitrate of the scene.
So I really think it's a bug in the ffdshow implementation.
But as far as I know the MPCHC decoder and the ffdshow one shouldn't be too different, should they?

I should really make this a faq somewhere. The cpu graphs in task manager are in no way representative of thread performance, because execution jumps around the cores! Go to process explorer's thread view and you'll see that it's only one single thread taking up all 25%.

VC1 decoder is single threaded and that's not going to change anytime soon, here's the current status (https://roundup.ffmpeg.org/issue1280). In other words, the only solution for the foreseeable future is for you to buy a processor that's faster at single-threaded tasks, like an i7 with boost or a dual-core i5 (which are all clocked faster than their quad-core brethren). Of course you'll lose all quad-core encoding perks doing that. Or use DXVA.

[albain]

Ok, so, I put Haali on my laptop and ran one of my new MKV's with the PGS subs embedded through it. Haali sees the sub and splits it for ffdshow just fine. So, MPC-HC's internal splitter is what's not working for PGS subs. This sucks because to date bitstreaming with Haali isn't so good. I'm open to advice on what to do. Also remember that I'm doing everything 64 bit, so, that complicates things slightly.

Hi SamuriHL, this is a known bug of MPC splitter : it does not detect embedded subs until the first one is parsed.
So you have to wait for it and then query again for the embedded streams from the context menu

[cge]

Hi,
I'm trying to display embedded Vobsub subtitles from MKVs through ffdshow but only 2 or 3 texts are shown then they disappear. Is this a known issue ?
Note: Using DirectVobSub, they work fine
Christophe

[SamuriHL]

Hi SamuriHL, this is a known bug of MPC splitter : it does not detect embedded subs until the first one is parsed.
So you have to wait for it and then query again for the embedded streams from the context menu

Hmmm. Fascinating. I'll have to give that a try in a little bit. I'll get back to you on this. :) The reason I question it is that if I convert to idx/subs it seems to work fine. But if I use MakeMKV and allow it to put the native PGS subs in, it never showed up for me. It does with haali. But I'll give it a try again and see if it eventually shows up.

[fps]

I should really make this a faq somewhere. The cpu graphs in task manager are in no way representative of thread performance, because execution jumps around the cores! Go to process explorer's thread view and you'll see that it's only one single thread taking up all 25%.

VC1 decoder is single threaded and that's not going to change anytime soon, here's the current status (https://roundup.ffmpeg.org/issue1280). In other words, the only solution for the foreseeable future is for you to buy a processor that's faster at single-threaded tasks, like an i7 with boost or a dual-core i5 (which are all clocked faster than their quad-core brethren). Of course you'll lose all quad-core encoding perks doing that. Or use DXVA.
Thanks for the info on that, didn't know this :thanks:.
However I looked at the thread view of process explorer and altough it shows only "1 big thread" while decoding a VC-1 video with ffdshow (libavcodec) the usage never reaches 25%. In fact it's somewhere around 13-20%.
With the internal VC-1 decoder of MPCHC (single threaded as well), which produces stutterfree playback, the cpu usage is about the same. I've attached 2 screenshots to make sure that I looked at the right statistics :).
So if my observations are right I think the stutterness isn't entirely caused by the CPU at least because the single threaded MPCHC decoder should have the same problems in theory but actually doesn't :confused:.
ffdshow (libavcodec) VC-1 decoder (http://img31.imageshack.us/i/ffdshowlibavcodec.png/)
MPCHC internal VC-1 decoder (http://img526.imageshack.us/i/mpchcvc1.png/)

[SamuriHL]

So....I kinda like how Haali controls the streams in combination with ffdshow, so, I installed it on my main HTPC figuring I would be screwed on bitstreaming, but, at least it'd give me subtitles like I want. MUCH to my surprise, bitstreaming is still working! I'm using the latest Haali from I believe 5/20. DTS-HD MA working fine from my MKV's. TrueHD working fine from my MKV's! Absolutely sweet. I'm SOOO happy now!!!

[SamuriHL]

@albain

As you've seen in the HD audio thread most likely, it's not as simple as what I just posted. In any case, back to subtitles for the moment. Haali does see the embedded PGS subtitles. MPC-HC's splitter does not. Ever. I let it play for several minutes and ffdshow never showed any internal subtitles for that MKV. It should be noted that if I convert the sup to idx/sub and use that as the internal sub that ffdshow sees that just fine. My guess is that mpc-hc's splitter doesn't see PGS subs.

[SamuriHL]

I have a probably stupid question but I honestly don't know the answer to this. Is there any problem/conflict with installing both x86 and x64 versions of ffdshow simultaneously? I would assume that 32 bit apps would pick ffdshow x86 and 64 bit apps would pick ffdshow x64, but, is there an issue having them both installed at the same time? Thanks!

[Inspector.Gadget]

SamuriHL, so far I've had no issues with having both versions installed - each is loaded automatically and painlessly by the appropriate platform version of MPC-HC, Avisynth, and other DirectShow and VFW apps. The only caveat is that you will have to configure everything twice - but you may even be able to work around this by figuring out how to export settings from your current version and importing it into the other.

[SamuriHL]

Awesome. Thanks for that. I don't mind about the configuration. I used to have the 32 bit version installed before I upgraded to the 64 bit version. I apparently needed the 32 bit version for RipBot264. Trying to fix the mistake of an idiot studio that thought mpeg2 was a useful format for blu-ray. :D

[STaRGaZeR]

I think my chances of making passionate love with one of Victoria's Secret Angels (preferably Doutzen Kroes) are a gazillion times higher than me finding the very first buggy rev :D

What we need is someone who reads process/memory dump files for a living. This would be the fastest and most effective way of dealing with this issue. Until then, I have to stick with the subpar alternative... :)

:D

You should try to find it though, that someone will probably never appear so we have higher chances of fixing it if you do it :p

Hi STaRGaZeR,

This build has fixed my problem. Thanks

Good, I've commited it in r3475.

[nevcairiel]

There seems to be some sort of heap memory corruption in ffdshow, or at least the audio decoder part of it. Whenever i use it in MPC-HC (or WMP, too), on exit or track-change, the app crashes.

I tracked the crash down to the destructors, happening when the filter-graph is being destructed, but knowing heap corruption, thats likely not where its being corrupted, just where it crashes.
Anyway, i know without specifics this report isn't that helpful, i just wanted to check if anyone else is seeing those crashes on exit or track change, or if something else on my system might cause it.

I tested different players and different splitters, both with reclock on and off. Tracking down the source of the corruption is a PITA, so before i spend more time on it.. anyone else seeing it, or is it just me?

[STaRGaZeR]

anyone else seeing it, or is it just me?

Nothing here.

[Momber]

http://img.pixs.ru/storage/5/0/5/ffdshowjpg_1810279_744505.jpg

This happens every time I run the video decoder configuration.
XP SP3, ffdshow_rev3466_20100531_clsid

[clsid]

Re-install ffdshow.

[Momber]

I just did. Same difference.

[helix89]

Can anyone confirm wether this is a bug or not...

I'm trying to play mkv with multiple editions of same movie - like extended and normal cut. It's in mkv with DTS and x264 and has subtitles. Haali splitter is known to handle this well, but it's not working with ffdshow dvax. The setup is tested in MPC-HC and graphstudio.
MPC-HC(every internal filter disabled): Changing edition in haali makes mpc-hc crash. Using ffdshow audio and ffdshow dvax
Graphstudio: Changing edition in haali makes mpc-hc crash
http://thumbnails33.imagebam.com/8463/5ac67f84625578.jpg (http://www.imagebam.com/image/5ac67f84625578)

MPC-HC(with own codec for dvax enabled - ffdshow dvax blocked) + haali: changing works well!

Newest haali splitter from may and ffdshow svn from a week ago.

[fps]

I think there is a bug in the VC-1 decoder:
With DXVA enabled everything works fine. The internal non DXVA VC-1(FFmpeg) decoder of MPCHC (build 2018) also works fine.
But whenever I try to play a VC-1 video (1080p Blu-Ray) with ffdshow(libavcodec) and DXVA disabled, playback becomes very stuttering.
Change the renderer in MPC-HC. EVR Custom has some bugs with VC-1 when decoded by the Microsoft decoder or ffdshow.
It is not really a splitter error. If it works with other renderers, then I see no reason why it shouldn't be possible to make it work with the modified EVR renderers. In fact, the pre-beliyaal EVR CP worked ok with VC-1.
Using Arcsoft splitter or Haali splitter, VC1 is working correctly with the EVR Sync in MPC as far as I know. So the splitter seems to be a part of the issue.
I quickly made a test and I confirm that the bug is not present (for both EVR cust and EVR Sync) using Arcsoft splitter ("Arcsoft MPEG Demux").
I just noticed the same. If I use Haali splitter then all the problems are gone no matter which decoder or renderer (EVR-CP, madVR) I use. So it really seems to be an error in the internal MPCHC splitter.

[albain]

@SamuriHL : I thought that both mkv/mpeg splitters acted the same way.
So MPC M2TS splitter + PGS => ok but one should wait for the first parsed gub
MPC MKV splitter + PGS => not detected

But Haali still suffer from HD audio splitting, no ?

And do you have any news concerning audio dropouts with radeons 5xxx series ?
Thanks

[SamuriHL]

The MPC m2ts splitter is all good as far as I can see.
The MPC MKV splitter PGS subs aren't seen by FFDSHOW and I can't select them.

Haali all the subs work perfectly, but, if I make the MKV with mkvmerge, the DTS-HD MA track breaks. If I make it with MakeMKV, *EVERYTHING* works flawlessly. I'm watching one right now in WMP with Haali+ffdshow. Bitstreaming perfectly, video rocks, and subs are selectable.

Audio dropouts will happen if you've got DXVA disabled and PowerPlay is running. This is a perpetual annoying bug that ATI has so far failed to fix. You can get around it by running the ATI GPU tool and lock the clock speed.

[albain]

The MPC m2ts splitter is all good as far as I can see.
The MPC MKV splitter PGS subs aren't seen by FFDSHOW and I can't select them.

Haali all the subs work perfectly, but, if I make the MKV with mkvmerge, the DTS-HD MA track breaks. If I make it with MakeMKV, *EVERYTHING* works flawlessly. I'm watching one right now in WMP with Haali+ffdshow. Bitstreaming perfectly, video rocks, and subs are selectable.

Audio dropouts will happen if you've got DXVA disabled and PowerPlay is running. This is a perpetual annoying bug that ATI has so far failed to fix. You can get around it by running the ATI GPU tool and lock the clock speed.

Okay, thanks, already did that

[SamuriHL]

Okay, thanks, already did that

Supposedly 10.6 is out now but I have no idea what they've fixed. They broke their website so downloading the driver is beyond painful right now. I've still not gotten it yet. Also in 10.5, trying to fix it in the profile xml did NOT work. You had to use the GPU tool thing. I just keep dxva on and call it good. :D

[STaRGaZeR]

I have a problem with another ass subtitle

Spec of Clip : MKV > AVC1 + Vorbis + ASS sub
FFDSHOW > ffmpeg-mt h264 + Tremor
Tested Output > YUY2 & RGB32
Using ffdshow (3448) / MPCH 1922 (EVR)

The sub should appear like this
> http://i50.tinypic.com/nxvyu.jpg

But FFdshow subtitle make it appear like this
> http://i47.tinypic.com/25uqjrk.jpg


Here is the ass files ;)

This is working for me in r3478, can you test it? :)

[SamuriHL]

@albain

James loves us. :D

http://forum.slysoft.com/showthread.php?t=41146

EDIT: PowerPlay bug is fixed in 10.6! WOO HOO! This is for the silent stream bug. YEY!

[STaRGaZeR]

This is working for me in r3478, can you test it? :)

I'm reverting this, it's wrong. This subtitle is using /t + /alpha for fades, and /t is not supported in ffdshow :(

[lordadmira]

Greets all.

I just upgraded to the rev 3466 ffdshow and now the vdf plugin does not work with VirtualDub or VirtualDubMod. When selecting it and setting postprocessing, the video is not affected. Clicking About shows that it is the new ffshow that is putting up the window. Downgrading to my old ffdshow of 2546 makes the VDM plugin work again. Any ideas?

edit: Some new info. I located and installed rev 1082 of tryouts and it has the same problem. I did some experimenting and it seems the other filters work except for Postprocessing. E.g. when I play with the levels I see the image change. No matter what settings I use under PP nothing happens. The build 2546 I've been using seems not to be tryouts. It doesn't really say who made it but the file name is ffdshow-2546-gcc4.0.3-sse.exe. So why doesn't PP work with tryouts builds?


LA


So like who maintains the VirtualDub plugin?


LA

[Sebastiii]

@albain

James loves us. :D

http://forum.slysoft.com/showthread.php?t=41146

EDIT: PowerPlay bug is fixed in 10.6! WOO HOO! This is for the silent stream bug. YEY!

Cool :) great news and thanks James :)

[clsid]

So like who maintains the VirtualDub plugin?If you can find the latest working revision and the first broken one, then maybe we can fix it.
You can find old build here:
http://sourceforge.net/projects/ffdshow-tryout/files/SVN%20builds%20by%20clsid/old%20builds/
http://sourceforge.net/projects/ffdshow-tryout/files/SVN%20builds%20by%20clsid/very%20old%20builds/

[clsid]

@albain
Closed captions are decoded by ffdshow even when the subtitle filter is disabled. This sometimes confuses people, making them unable to figure out how to disable the subtitles. Perhaps an option called "Closed Captions" could be added to the tray context menu (visible only if CC are present) where users can toggle the display of CC on/off. The option could be placed where "Flip" and "Keyboard & Remote" are also located.

[lordadmira]

If you can find the latest working revision and the first broken one, then maybe we can fix it.
You can find old build here:
http://sourceforge.net/projects/ffdshow-tryout/files/SVN%20builds%20by%20clsid/old%20builds/
http://sourceforge.net/projects/ffdshow-tryout/files/SVN%20builds%20by%20clsid/very%20old%20builds/
I got the oldest one there, rev164, and it still doesn't work.

Unfortunately I need the newer builds to watch new AVC encodes but I still need ffvdub for VDM so that means installing one or the other as I change tasks. o_O


LA

[clsid]

I have just tested with VirtualDub 1.9.9 and it works fine with the latest version of ffdshow.

VirtualDub -> Video -> Filters -> add ffvdub

[DigitalDeviant]

@albain
Closed captions are decoded by ffdshow even when the subtitle filter is disabled. This sometimes confuses people, making them unable to figure out how to disable the subtitles. Perhaps an option called "Closed Captions" could be added to the tray context menu (visible only if CC are present) where users can toggle the display of CC on/off. The option could be placed where "Flip" and "Keyboard & Remote" are also located.

I second this. The CC decoding confused the hell out of me for a while.

[SamuriHL]

@albain

Is there a way in ffdshow subtitle configuration to set it so that it only shows subtitle files with the same base name as the file being played? I have a subtitle sub dir that I drop PGS files into (since mkvmerge can't currently embed them) and ffdshow shows every file in there. Am I missing an option somewhere? I name them "Movie(ENG)" as the main sub and then "Movie(ENG)(2)" "Movie(ENG)(3)" etc for the rest of the subs. I'd like to, by default, set ffdshow to only show "Movie*" subs. Possible?

[pankov]

SamuriHL,
the "default" naming convention for subtitles is
"movie name.lang.ext"
in your case I'd suggest naming your subs in the following way
"Movie name.eng.sub"
"Movie name.eng2.sub"
"Movie name.eng3.sub"
and I believe that the latest ffdshow build will show only these three files, not everything in the folder.

[SamuriHL]

Ah! Thank you for that. I'm still very new to this MKV stuff, so, learning something every day. :) I'll try that out!

[STaRGaZeR]

@albain
Closed captions are decoded by ffdshow even when the subtitle filter is disabled. This sometimes confuses people, making them unable to figure out how to disable the subtitles. Perhaps an option called "Closed Captions" could be added to the tray context menu (visible only if CC are present) where users can toggle the display of CC on/off. The option could be placed where "Flip" and "Keyboard & Remote" are also located.

There's already a CC on/off checkbox in the subtitle section, I guess it doesn't work as it should?

[SamuriHL]

I renamed my files to the convention pointed out by pankov, but, ffdshow is still picking up ALL the sup files in the directory. It doesn't seem like ffdshow is following the naming convention at all. Thanks!

EDIT: Sorry, sorry, my bad. I didn't have the latest ffdshow installed on my laptop where I was testing. Sorry!

[clsid]

There's already a CC on/off checkbox in the subtitle section, I guess it doesn't work as it should?I know, but the problem is that CC functions independently from the subtitle filter itself. That is extremely confusing. Hence the idea of a more obvious option.

[Midzuki]

Originally Posted by clsid

@albain
Closed captions are decoded by ffdshow even when the subtitle filter is disabled.

Besides:

DVD subtitles are (wrongly) decoded by ffdshow "raw video filter" even when the subtitle filter is disabled.

Hopefully this time I have been clear enough. ;)

[SamuriHL]

I have what is probably a stupid question. When I've got an MKV, and I use haali to split it, and it's H264 or VC1, ffdshow dxva allows me to nicely control the subs. What happens when it's mpeg2? The ffdshow dxva icon isn't showing so....um, how do I control the subs in that case? :confused:

[Midzuki]

Yet Another Not-Very-Smart ffdshow's "feature":

Dolby TrueHD decoder. Works only if AC3 decoding is enabled!

MLP decoder. Works only if AC3 decoding is enabled!

It would make some sense if the TrueHD decoder
depended on the MLP decoder only.

Reminds of Netscape 4.xx, in which the CSS interpreter depended on the Javascript engine. :rolleyes:

[clsid]

I have what is probably a stupid question. When I've got an MKV, and I use haali to split it, and it's H264 or VC1, ffdshow dxva allows me to nicely control the subs. What happens when it's mpeg2? The ffdshow dxva icon isn't showing so....um, how do I control the subs in that case? :confused:For mpeg-2 you need to use the regular ffdshow video decoder, either for decoding the video or for processing the raw output of another decoder.

Yet Another Not-Very-Smart ffdshow's "feature":

Dolby TrueHD decoder. Works only if AC3 decoding is enabled!

MLP decoder. Works only if AC3 decoding is enabled!

It would make some sense if the TrueHD decoder
depended on the MLP decoder only.
It is actually a very smart feature because TrueHD and MLP are sometimes labeled as AC3. That is why ffdshow got a special audio parser to determine the real audio format and handle it accordingly. This is also why there is a dependence on AC3. I am not sure if the dependence also exist for streams that are properly labeled, Albain should be able to answer that.

[lordadmira]

Reminds of Netscape 4.xx, in which the CSS interpreter depended on the Javascript engine. :rolleyes:
Plus 10 points for the old school reference!


LA

[SamuriHL]

For mpeg-2 you need to use the regular ffdshow video decoder, either for decoding the video or for processing the raw output of another decoder.


Sorry, I know I'm starting to get annoying on this but I'm still trying to learn here. How would I go about configuring the ffdshow video decoder to process another decoder? That would be absolutely ideal I think. Thank you for your patience and I apologize for all the questions.

[clsid]

Enable "raw video" at the bottom of the Codecs page in the config.

[SamuriHL]

Thank you! I very much appreciate the help. I'm ALMOST where I want to be with this MKV experiment. :)

[JarrettH]

What does "updated ICL11 project files" mean? Is there an ICL11 compile somewhere?

[esoteradactyl]

Is anybody else experiencing any problems with the VC-1 (libavcodec) decoder? Since ATI seems to have fixed the powerplay issues, i have switched back to software decoding. I upgraded my version of ffdshow to a newer build and seem to be having problems. Im getting stutter and if i seek it stutters really bad. Both my machines are more than capable of software decoding. Strange.

[fps]

Is anybody else experiencing any problems with the VC-1 (libavcodec) decoder? Since ATI seems to have fixed the powerplay issues, i have switched back to software decoding. I upgraded my version of ffdshow to a newer build and seem to be having problems. Im getting stutter and if i seek it stutters really bad. Both my machines are more than capable of software decoding. Strange.
Some posts above (http://forum.doom9.org/showpost.php?p=1408934&postcount=11966) I noticed the same.
Which splitter and player are you using? With MPCHC either selecting the internal VC-1 decoder or using Haali as splitter + ffdshow should fix the issue.

[MatLz]

Hi !What does "updated ICL11 project files" mean?That's icl10+1...:D
Normaly we can expect a small speedup with this kind of build...1% maximum if I'm right.Is there an ICL11 compile somewhere?You can find these builds at www.xvidvideo.ru

[XhmikosR]

Unless you mean the difference between ICL10 and ICL11 builds, then the difference between MSVC2008 and ICL11 with resizing+deband+HQRGB32 is ~=15%. Which reminds me I have to finish that benchmark at some point. :) But you can find some previous results by searching this thread.

[esoteradactyl]

Some posts above (http://forum.doom9.org/showpost.php?p=1408934&postcount=11966) I noticed the same.
Which splitter and player are you using? With MPCHC either selecting the internal VC-1 decoder or using Haali as splitter + ffdshow should fix the issue.

I'm using the mpc splitter. Ive tried both mpc and haali with the same results. I'm playing everything through mediaportal and wmp suffers from the exact same problem.