ffdshow tryouts project: Discussion & Development

[clsid]

About ffdshow tryouts

The ffdshow tryouts project (http://sourceforge.net/projects/ffdshow-tryout/) is a fork of the ffdshow project (http://sourceforge.net/projects/ffdshow/). This fork was created by a group of members here at Doom9. The last modification to the source code in the SVN repository of the original project was done in may 2006. The ffdshow tryouts project has continued where the original project stopped. Several bugs have been fixed. Lots of code (that ffdshow borrows from the FFmpeg project (http://ffmpeg.org/)) has been updated, and also support for a few new formats has been added.

Important Notice

Development of ffdshow has been discontinued!

All former developers of ffdshow have lost interest in the project. There have not been any significant changes to the code since 2012, only some minor trivial changes have been made in the past year.

A superior audio and video decoder is available in the form of LAV Filters (http://forum.doom9.org/showthread.php?t=156191).

Downloads

ffdshow builds (http://sourceforge.net/project/showfiles.php?group_id=173941)

Links


ffdshow tryouts project page on SourceForge.net (http://sourceforge.net/projects/ffdshow-tryout/)
ffdshow tryouts SVN repository (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/) (SVN changelog (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/?view=log))
ffdshow tryouts bugtracker/feature request/patches (http://sourceforge.net/tracker/?group_id=173941)
ffdshow tryouts changelog (incomplete!) (http://ffdshow-tryout.sourceforge.net/html/en/)
ffdshow tryouts FAQ (http://ffdshow-tryout.sourceforge.net/wiki/faq)
old ffdshow SVN repository (http://ffdshow.svn.sourceforge.net/viewvc/ffdshow/trunk/?sortby=date)
previous ffdshow discussion topic (http://forum.doom9.org/showthread.php?t=98600)


Known issues & bugs:

Known bugs in latest revision:
bugtracker (https://sourceforge.net/tracker/?group_id=173941&atid=867360&status=1)

Limitations:

B-frame decoding delay in VFW when using libavcodec for Xvid/DivX.
In some cases VobSub subtitle files (.idx/.sub) are loaded, but just some or even none of the subs are displayed. VobSub support is still incomplete.


What is the difference between the ICL and Generic builds?

The ICL builds require at least a SSE capable CPU. The generic builds only require MMX. However, all builds contain the same hand-written assembly code with MMX/MMXext/SSE/SSE2/3dnow/3dnow2/SSSE3 instructions, which will get used if and only if your CPU supports it.

The difference in builds only applies to the ffdshow.ax file. All builds use GCC for compiling libavcodec. So the decoding performance is equal for all builds. The only situation where the ICL build might give a bit better performance is when you make use of the internal filters in ffdshow. And even then, there is only a gain for some of the filters.

So to summarize, it does not matter much which build you use.

[ffdshow_tryouts]

About ffdshow tryouts:
ffdshow is a DS filter and VFW codec for many audio and video formats, such as DivX, Xvid and H.264. Over 60 bugs have been fixed, codecs have been updated, and support for a few new formats has been added in the tryouts. Windows Vista is now supported.
History:

The first versions of ffdshow were published in April 2002.The main developer was Milan Cutka. When he stopped updating the project in 2006, new maintainers opened the ffdshow-tryouts as a fork, where bugfixes, stability fixes, new features and codec updates continue.

Notes:
A common misconception is that SSE/SSE2 builds will decode video better than "generic" builds. In fact, the libavcodec video decoders are always compiled in GCC and are usually hand-optimized; it's the ffdshow filters that benefit from SSE/SSE2.

Please remember:
All the developer working on FFdshow are volunteers.
All used sources and ffdshow itself are distributed under GPL.
Bug reporting:

1).operating system (OS)
2).ffdshow version, rev, compiler, ansi or unicode
3).ffdshow settings and used filters, ex:queue, resize, sharpening, noise, levels...
4).media player used and settings ex: mpc, media player2, wmplayer...
5).output colorspaces "YUY2", "YV12", "RGB32", "RGB24"...
6).internal/external splitters ex: Haali Media Splitter, Gabest's Splitters...
7).video renderers ex:Haali renderer, Overlay Mixer, VMR9, VMR7, EVR...
8).video/audio samples <10MB
9).what you were trying to accomplish, the problem you encountered...anything else that could be useful in reproducing that bug.

[Inventive Software]

Having downloaded a newer revision, I'm pleased to see the "multithreaded" encoders bug I posted a while back is fixed, although brutally! ;) I don't know whether it's a ffmpeg bug or not, other people would have to provide me a binary with AVS input so I can test! :D

[Henrikx]

German translation (About ffdshow tryouts) :)
http://forum.gleitz.info/showthread.php?p=314937#post314937

[LoRd_MuldeR]

Anybody able to download:
http://downloads.sourceforge.net/ffdshow-tryout/ffdshow_rev756_20070109_clsid_icl9.exe?
???

//EDIT

strange, before all mirrors were either down or redirected back to download page.
just tried again, and it worked...

[Eragon4ever]

Not from the default mirror.

Edit: Seems the server Dublin, Ireland is not working.

[SeeMoreDigital]

Great news... It's about time we had a dedicated thread for this sort of thing ;)

[zambelli]

The older ffdshow thread brought up the idea of a more unified versioning scheme for ffdshow. Will this project perhaps deliver on that?

[haruhiko_yamagata]

Queue is off in Overlay Mixer. (reported by clsid)
It's the spec. For most cases Overlay Mixer connects to old video renderer.
In that case, multithreading can't be used.
Queue is on only if Overlay Mixer is connected to VMR or something.

[kurt]

hm, but when using either old and system default renderer in mpc, ffdshow's OSD tells queued output samples are in use.

Therefore shouldn't it be possible to get it work with overlay mixer too?

[bob0r]

Anything worth updating for H.264 decoding since revision 684 beta1?

[Leak]

Hi guys - long time no see... :D

Anyway, first let me thank you all for keeping ffdshow's development going - it's continually getting better... :)

Now, I really didn't have much use for ffdshow in the last few months, as I preferred watching my DVDs on our TV, but since I bought a projector last week I've been back to using my PC for my DVD watching needs.

As DVD decoding isn't working reliably in ffdshow, I've been using Media Player Classic's MPEG2 decoder and postprocessing that via ffdshow's "Raw video" set to "YUV modes".

Using a combination of PowerStrip, Reclock, ffdshow and MPC, I've got a great picture out of my Samsung Pocket Imager, at least for stuff that doesn't need deinterlacing. But since most anime is of course interlaced with 3:2 pulldown, I wanted to do the following:

Set the refreshrate for the image sent to my projector to 59.9xx Hz
Have MPC run in fullscreen on display 2
Play the DVD in MPC
Using ffdshow, use Kernel Bob to get the 29.9xx FPS video to 59.9xx FPS
Using ffdshow, resize the image to 856x480 for 16:9 DVDs or to 800x600 for 4:3 DVDs using Lanczos resizing
Using ffdshow, crop 28 pixels left and right in case of a 16:9 DVD to get back to 800x480

At least that's the plan.

The problem is that resizing only works if I do it before deinterlacing, but I guess I needn't tell you that resizing interlaced material is a big no-no. ;)

If I deinterlace first and resize afterwards, ffdshow either crashes or at least produces a distorted image as if the line pitch was incorrect, plus MPC's interface becomes rather sluggish and takes seconds to react. I've experienced crashes even if I turned off deinterlacing when it was above the crop filter in the list, though... :(

So for the time being, I'm resizing 16:9 DVDs to 856x480 before deinterlacing, as that doesn't resize vertically, but I have to let my graphics card (ATI Radeon X1900 XT) stretch kernel-bobbed 4:3 DVDs to 800x600 which of course isn't great quality-wise. On the other hand, I've found that at least the card's 3:2 pulldown removal kicks in if I change my refresh rate to 48Hz, which is nice - but which only looks good if there aren't any 30 FPS pans/zooms/etc. in the video... :)

So - could you please add resizing after deinterlacing to the bug list? It's possible that it only happens with the kernel deinterlacer, but I haven't checked it that thoroughly yet.

I might take a look at the code myself if I can find some time for it, but the last time I looked at the code I was rather lost - is there some documentation somewhere for ffdshow's internals?

[haruhiko_yamagata]

@Leak
Thank you, I confirmed your report.

[clsid]

Anything worth updating for H.264 decoding since revision 684 beta1?
No performance increase or fixes for H.264 decoding. But there have been a couple of other bugfixes, so you could post an update.

[chros]

I'm using Winamp's DFX plugin for a long time, to increase the audio quality.
There's a new version of it (8.0) which can deal with multichannel source ! But ffdshow doesn't allow to connect them ...
There's a source which shows the splitter that it has only 2ch audio but it has true 5.1: it works well with DFX.

Can you do anything about this?
dk75 asked me for sources:
2ch: http://images.apple.com/movies/fox_atomic/thehillshaveeyes2/thehillshaveeyes2_h480p.mov
6ch: http://images.apple.com/movies/newline/the_last_mimzy/the_last_mimzy-tlr1_h480p.mov

It's easy to recognize if DFX kicked in: it stoles the focus from MPC ...

Thanks

[Didée]

About Deinterlacing(Bob) + Resize:

It's not only this combination being faulty ... also some other filters are affected, if they are placed downstream of "deinterlacing". Effect is that the framerate is doubled by bobbing, but the upfollowing filters take only place every other frame.
From memory: "Picture properties", "Levels" and "DeBand" show that misbehaviour.
Oh, and it's not exactly a "new" bug ...

[foxyshadis]

The avisynth filter has some special handling for internal framerate doubling, but it's quite possible that the deinterlacing filter doesn't. Hmm. Or even if it does, it looked a little shaky to me anyway. I guess it'll have to get cleaned up.

[dk75]

dk75 asked me for sources:
2ch: http://images.apple.com/movies/fox_atomic/thehillshaveeyes2/thehillshaveeyes2_h480p.mov
6ch: http://images.apple.com/movies/newline/the_last_mimzy/the_last_mimzy-tlr1_h480p.mov

It's easy to recognize if DFX kicked in: it stoles the focus from MPC ...

Thanks
First file, this 6ch won't connect to DFX but is report by ffdshow as 6ch.
Second file is reported as 2ch and connects to DFX.
I've tryed other files that i have and DFX won't connects to 6ch file decoded by ffdshow.
Maybe it's restriction with free unregistered version...

[SeeMoreDigital]

First file, this 6ch won't connect to DFX but is report by ffdshow as 6ch.
Second file is reported as 2ch and connects to DFX.
I've tryed other files that i have and DFX won't connects to 6ch file decoded by ffdshow.
Maybe it's restriction with free unregistered version...Sounds like the MOV splitter isn't able to correctly parse QT sources containing 6Ch AAC streams....

This issue is not new, there was a time when MediaPlayer Classic's MOV/MP4 splitter could not do this either.... Until it was tweaked ;)

For some reason 6Ch AAC streams are stored differently in the MOV container, compared to the .MP4 container....


Cheers

[chros]

I've tryed other files that i have and DFX won't connects to 6ch file decoded by ffdshow.
Maybe it's restriction with free unregistered version...
No, I have the full version and in winamp it can connect to 6ch mp4 files (where previous versions don't) ...

@SeeMoreDigital: you are very kind (as always :) ), but the problem is not with mov files (try it some avi files) but with ffdshow :)

Of course, there are lots of dsp plugins which don't support 6ch audio, so that's why I have asked this question...
So again: is there any way (of course a patch) to support only DFX 6ch audio?

Thanks

[Yong]

latest snow decoder patch for ffdshow:
http://www.geocities.com/y0ngc/ffsnow.txt
untested;)

[Peuj]

Originally Posted by CoRoNe View Post
The fact that the following MOV[SVQ3+AAC] crashes...is this FFDShow to blame or the MP4 Splitter?

link: http://www.a-film.nl/film/trailer/00...ktime_HOOG.mov

Whith the latest MPC rev611-3.2, ffdshow_rev756_20070109_clsid_icl9, I can reproduce this issue too. Can someone tell if it's a ffdshow issue ? (because clsid doesn't add it in the known bug list... :p)

thanks

[fastplayer]

Here the proper link:
http://www.a-film.nl/film/trailer/00000397_quicktime_HOOG.mov
MPC crashes, MPUI plays the file back though.

[Henrikx]

WMP11 (XP) ffdshow crashes too !

[_xxl]

SVQ3 is not fully suported by ffdshow.Maybe it is a splitter issue...

[akupenguin]

SVQ3 is fully supported by libavcodec, and http://www.a-film.nl/film/trailer/00000397_quicktime_HOOG.mov plays fine in MPlayer.

[_xxl]

True.But ffdshow uses an external mp4/mov splitter.

[Inventive Software]

So all that's needed, essentially, is to make sure you have a good splitter installed, correct?

[fastplayer]

Has anybody checked if the MOV file works with Haali's splitter?

[Peuj]

Has anybody checked if the MOV file works with Haali's splitter?

yep it crashed too, (i have made the test from graphedit)

[cc979]

yep it crashed too, (i have made the test from graphedit)

hali splitter worked for me, but complained and the audio did not work

[haruhiko_yamagata]

@Leak
I patched a bit. Please see if it is fixed.
Binary is also updated.

[Leak]

@Leak
I patched a bit. Please see if it is fixed.
Binary is also updated.
Yay! :D

It works great if I set up ffdshow before starting DVD playback in MPC.

I still get a half-height picture during playback if I activate/deactivate the crop filter if it's put after resizing and deinterlacing, but at least it works right after restarting playback - and I really don't want to fiddle with ffdshow's settings while watching anyway... :)

Thanks!

(Of course, this was just a quick check - I'll watch a DVD or two this evening and post any findings... ;))

np: autechre - part 02 (xltronic radio)

[dk75]

hali splitter worked for me, but complained and the audio did not work
Haali Media Splitter 1.6.400.11 20061228 + ffdshow 752 sse2 XXL + ZoomPlayer 5.0 std and i've got unresponsible system (98% CPU load).

[haruhiko_yamagata]

English version of change log (http://ffdshow-tryout.sourceforge.net/html/en/)
Japanese version of change log (http://ffdshow-tryout.sourceforge.net/html/jp/)

Please fix my English. As you all know, my English is very broken.
Please fix my html. This is my first html.
Please add anything missing, fix anything incorrect.

TODO: add style sheet. It is appreciated if you write it.

It's also in the svn.

@developers
Please feel free to update svn and web.
Please update svn when you update web.

[BinGix]

Please add support for vobsub embedded subtitles like those found in nero digital files.

[fastplayer]

English version of change log (http://ffdshow-tryout.sourceforge.net/html/en/)
Please fix my English. As you all know, my English is very broken.
I'm on it and it's not that bad :)

Here's the file. Please do a quick diff with the original file to make sure I didn't change the meaning.
http://d.turboupload.com/d/1426823/ffdshow_tryouts_changelog.zip.html

[fastplayer]

TODO: add style sheet. It is appreciated if you write it.
Just for basic formatting of the page? If it's not something fancy I can do that too.

[haruhiko_yamagata]

I'm on it and it's not that bad :)

Here's the file. Please do a quick diff with the original file to make sure I didn't change the meaning.
http://d.turboupload.com/d/1426823/ffdshow_tryouts_changelog.zip.html
Thank you very much. Updated.

Just for basic formatting of the page? If it's not something fancy I can do that too.
I'm not sure if it is necessary. I don't know how it looks on other browser.
Current html depends on browser's default style.
Probably not too bad.

[fastplayer]

I'm not sure if it is necessary. I don't know how it looks on other browser.
Current html depends on browser's default style.
Probably not too bad.
It's not necessary but it allows to change quickly the layout and format of the page in case it's need in the future. Right now, it looks identical on all 3 big browsers. Adding style support won't break that. I'll "convert" it to XHTML too - for fun's sake :)

[sillKotscha]

what is the difference between:

- Nightly generic builds by clsid - ffdshow_rev770_20070113_clsid.exe and

- Nightly generic builds by xxl - ffdshow_rev770_20070113_xxl.exe

except the name/ person who compiled it ??

thank you

[Spaceman-Spiff]

I was trying to find the same information, but doesn't seem to be shown anywhere on the ffdshow-tryout website or its build forums.

So far the only thing I notice is clsid's builds support MMX, some of xxl's builds support SSE, and that they use different compilers. Is Yamagata's build in Japanese? *shrug*

Would be nice if there's an explanation on what each build is best at doing compared to each others.

[clsid]

You can get compiler information for my build by clicking on the clipboard icon on the SourceForge website.

[sillKotscha]

Would be nice if there's an explanation on what each build is best at doing compared to each others.

my concern was not to ask which one is better compared to each one but to find out (because of testing) if different patches are used or what_not...

I ask because it is the same revision in both cases and they should be compiled with the same compilers (for generic use)...

[cc979]

Here the proper link:
http://www.a-film.nl/film/trailer/00000397_quicktime_HOOG.mov
MPC crashes, MPUI plays the file back though.

mpc using built mp4/mov splitter with ffdshow-rev775 plays fine now, but hali's has problems - no audio, vid is slow

[fastplayer]

Changelog with basic CSS support. Feel free to change... everything :D
http://d.turboupload.com/d/1427404/ffdshow.zip.html

[fastplayer]

my concern was not to ask which one is better compared to each one but to find out (because of testing) if different patches are used or what_not...
Same source code, different compiler versions and parameters, and a slightly different InnoSetup script. That's it. Just close your eyes and click on either one of the download links ;)

[shon3i]

Why is VSSH decoding different than avc?

[Peuj]

mpc using built mp4/mov splitter with ffdshow-rev775 plays fine now, but hali's has problems - no audio, vid is slow

No problem too with the latest ffdshow version, thanks for the fix.

...but I have another issue with another mov file. :D

I use the version ffdshow_rev776_20070114_xxl and I try to play the mov file from http://www.jeuxvideo.fr/telecharger-video-1-trailer-38661.html

For the video all is ok but for the sound I have to enable the "Uncompressed=(all supported)" and the problem is that with the "Uncompressed" enabled the mov lags during the playback (and CPU usage 70%).

Without the "Uncompressed" enable I have no sound but the mov doesn't lag at all.

Thanks

[foxyshadis]

Well, it's PCM audio, so the uncompressed makes sense. I'm not seeing any cpu spike though. Things to check: Are any audio filters enabled? Is ffdshow being inserted more than once? (Info - multiple instances - should be set to only one, check previous.) Is audio being output in a format your drivers are having trouble with? (16bit integer is the most compatible.)

You can try building a graph in graphedit to see if anything looks out of place, as well.

[_xxl]

For the video all is ok but for the sound I have to enable the "Uncompressed=(all supported)" and the problem is that with the "Uncompressed" enabled the mov lags during the playback (and CPU usage 70%).
Thanks
MPC works ok for me with nero digital parser.
MPC is crashing with internal mp4 splitter.

[Peuj]

Well, it's PCM audio, so the uncompressed makes sense. I'm not seeing any cpu spike though. Things to check: Are any audio filters enabled? Is ffdshow being inserted more than once? (Info - multiple instances - should be set to only one, check previous.) Is audio being output in a format your drivers are having trouble with? (16bit integer is the most compatible.)

You can try building a graph in graphedit to see if anything looks out of place, as well.

Thanks for the reply. :)

I was not surprise by the use if the Uncompressed it was just to explain why I use it.
There are no other filters selected in ffdshow audio.
It is set to "only one, check previous only".
Only the 16bit integer is checked.

About the graphedit I don't know what to do but when I open the mov inside I get this:
http://img77.imageshack.us/img77/4262/graphediteo7.jpg

And I use the latest version of MPC.

thanks

[haruhiko_yamagata]

Changelog with basic CSS support. Feel free to change... everything :D
http://d.turboupload.com/d/1427404/ffdshow.zip.html
Thank you. Updated.

Excuse me, I forgot to tell you one thing.
I have a plan(may be just a dream) to build a help file(ffdshow.chm).
Microsoft HTML help 1.4 (http://msdn.microsoft.com/library/default.asp?url=/library/en-us/htmlhelp/html/vsconhtfile.asp) supports HTML 3.2.
So I would like to be keep compatibility with HTML 3.2.
As far as I see, it is compatible.:thanks:

[fastplayer]

Since IE is the rendering engine for HTML Help, I think the DOCTYPE should not matter or at least can be changed to something different in case the code is not rendered properly in HTML Help viewer.

Edit: In case the separate style.css is a problem for HTML Help, it can be embedded right into the index.html.

[haruhiko_yamagata]

HTML 3.2 is officially supported, but yes, IE is the rendering engine and XHTML should work in a computer that ffdshow works.

[Spaceman-Spiff]

my concern was not to ask which one is better compared to each one but to find out (because of testing) if different patches are used or what_not...

I ask because it is the same revision in both cases and they should be compiled with the same compilers (for generic use)...

I'm curious if SSE or MMX really makes a difference compared to normal builds.

I saw this on Wikipedia:
A common misconception is that ICL SSE/SSE2 builds will decode video better than "generic" builds. In fact, the video decoders are always compiled in gcc and are usually hand-optimized; it's the ffdshow filters that benefit from ICL.

[Jeremy Duncan]

libavcodec decode:

int numthreads=deci->getParam2(IDFF_numLAVCdecThreads);
if (numthreads>1 && mpeg12_codec(codecId))
libavcodec->avcodec_thread_init(avctx,threadcount=numthreads);
else
threadcount=0;

libavcodec encode:

if (coCfg->numthreads>1 && sup_threads(coCfg->codecId))
libavcodec->avcodec_thread_init(avctx,threadcount=coCfg->numthreads);
else
threadcount=0;

Is this tweak being used ?
Does it improve multithreading ?

[Mangix]

smaller version of that PNG image that eragon4ever made :P

http://img180.imageshack.us/img180/1808/logophpbbsz2.png

[fastplayer]

Before everybody else starts "optimizing" file sizes, CHM files are compressed archives...

[Wedgedkc]

Would it be possible to enable support for RealAudio files? When I try to play them with mplayer it works ok (ffcook/realaud) but when I play it in mplayerc I get this messages, cannot render pins etc:
Audio: RealAudio SIPR 16000Hz mono

AM_MEDIA_TYPE:
majortype: MEDIATYPE_Audio {73647561-0000-0010-8000-00AA00389B71}
subtype: Unknown GUID Name {52504953-0000-0010-8000-00AA00389B71}
formattype: FORMAT_WaveFormatEx {05589F81-C356-11CE-BF01-00AA0055595A}
bFixedSizeSamples: 1
bTemporalCompression: 1
lSampleSize: 5120
cbFormat: 91
Media Type 0:
--------------------------
Audio: RealAudio COOK 44100Hz stereo

AM_MEDIA_TYPE:
majortype: MEDIATYPE_Audio {73647561-0000-0010-8000-00AA00389B71}
subtype: Unknown GUID Name {4B4F4F43-0000-0010-8000-00AA00389B71}
formattype: FORMAT_WaveFormatEx {05589F81-C356-11CE-BF01-00AA0055595A}
bFixedSizeSamples: 1
bTemporalCompression: 1
lSampleSize: 10240
cbFormat: 112

I've tried to upload some samples here (http://www.megaupload.com/se/?d=YRBZGIM6)

[BlindWanderer]

In the "Resize & aspect" page, the 'Multiply by' button text field is 1 pixel too long and overlaps with the text box. :D

[haruhiko_yamagata]

libavcodec decode:

int numthreads=deci->getParam2(IDFF_numLAVCdecThreads);
if (numthreads>1 && mpeg12_codec(codecId))
libavcodec->avcodec_thread_init(avctx,threadcount=numthreads);
else
threadcount=0;

-- omit --

Is this tweak being used ?
Does it improve multithreading ?
I think it's not working.

Thread=1 : User: 29s, kernel: 0s, total: 29s, real: 30s, fps: 41.2, dfps: 40.6
Thread=2 : User: 7s, kernel: 0s, total: 7s, real: 30s, fps: 162.1, dfps: 40.5
Both result "real = 30s". What does this mean?

By the way, what is the merit of using libavcodec-mpeg1/2?
With libmpeg2, HW deinterlacing was not working and it was crashing (or causing artifacts) on seek formerly.
Both of the bugs have been fixed untill now. Libmpeg2 has better performance in most cases. Will it change if the bug(?) of multithreading of libavcodec-mpeg2 is fixed?
Seeking in libavcodec-mpeg2 has issue. It causes artifacts untill key frame comes.

[Jeremy Duncan]

haruhiko_yamagata,

Please see post # 3537, # 3541 in this thread Link. (http://forum.doom9.org/showthread.php?t=98600&page=177)

Here's a link to a FFDshow that uses the tweak I asked about.
Link (http://www.mytempdir.com/1145846)

[_xxl]

Multithreaded libavcodec mpeg1/2 de(en)coding is working for me.
1 thread:
User: 7s, kernel: 0s, total: 7s, real: 7s, fps: 639.9, dfps: 634.7
2 threads:
User: 1s, kernel: 0s, total: 1s, real: 5s, fps: 3008.6, dfps: 912.1
Small encoding test:
1 thread:
1:52s
2 threads:
1:42s
10s faster.

[Romario]

Well, not much faster, only 10 seconds for 2 threads???

That's very strange, it should be at least 50 % faster then 1 thread.

[fastplayer]

That's very strange, it should be at least 50 % faster then 1 thread.
Really? Since when is every line of code "multi-threadable"?

[F_L_C]

I'm curious if SSE or MMX really makes a difference compared to normal builds.

I saw this on Wikipedia:
A common misconception is that ICL SSE/SSE2 builds will decode video better than "generic" builds. In fact, the video decoders are always compiled in gcc and are usually hand-optimized; it's the ffdshow filters that benefit from ICL.

The developers have said that ICL builds may have better performance if you use any of the filters. Resize is not affected but denoise3D, sharpen, and the rest are. Decoding speed has been shown to be nearly identical to generic builds.

[haruhiko_yamagata]

I'm writing faq.htm.
Here's the results.
libavcodec-DivX5 Decoding 720p CinderellaMan_HD.avi
result(sec)=builds
42=ffdshow_rev752_20070108_xxl.exe
41=ffdshow_rev752_20070108_sse_xxl.exe
42=ffdshow_rev752_20070108_sse2_xxl.exe
42=ffdshow_rev756_20070109_clsid_icl9.exe


CinderellaMan_HD.avi + xsharpen
result(sec)=builds
148=ffdshow_rev752_20070108_xxl.exe
90 =ffdshow_rev752_20070108_sse_xxl.exe
147=ffdshow_rev752_20070108_sse2_xxl.exe
102=ffdshow_rev756_20070109_clsid_icl9.exe

CinderellaMan_HD.avi + Denoise 3d
result(sec)=builds
131=ffdshow_rev752_20070108_xxl.exe
130=ffdshow_rev752_20070108_sse_xxl.exe
115=ffdshow_rev756_20070109_clsid_icl9.exe

[_xxl]

Well, not much faster, only 10 seconds for 2 threads???
That's very strange, it should be at least 50 % faster then 1 thread.
Really? Since when is every line of code "multi-threadable"?
What is multithreaded?
(ffmpeg 2772)
multithreaded/SMP motion estimation
multithreaded/SMP encoding for MPEG1/MPEG2/MPEG4/H263

(ffmpeg 2778)
multithreaded spatial complexity calculation for i frames
(ffmpeg 2810)
multithreaded mpeg2 decoding
* DV decoding/encoding now supports MultiThreading for up to 324 CPUs ;-)

Here's the results.
CinderellaMan_HD.avi + xsharpen
result(sec)=builds
148=ffdshow_rev752_20070108_xxl.exe
90 =ffdshow_rev752_20070108_sse_xxl.exe
147=ffdshow_rev752_20070108_sse2_xxl.exe
102=ffdshow_rev756_20070109_clsid_icl9.exe
CinderellaMan_HD.avi + Denoise 3d
result(sec)=builds
131=ffdshow_rev752_20070108_xxl.exe
130=ffdshow_rev752_20070108_sse_xxl.exe
115=ffdshow_rev756_20070109_clsid_icl9.exe
More tests are needed.If SSE(2) compiled by MSVC2003 are faster or not?

[haruhiko_yamagata]

I have added FAQ to the web (http://ffdshow-tryout.sourceforge.net/html/en/faq.htm).
Please fix my English,html and anything.

[fastplayer]

I have added FAQ to the web (http://ffdshow-tryout.sourceforge.net/html/en/faq.htm).
Please fix my English,html and anything.
I'm on it.

FAQ (http://ffdshow-tryout.sourceforge.net/html/en/faq.htm)& Changelog (http://ffdshow-tryout.sourceforge.net/html/en/) updated.

[cc979]

I'm on it.

FAQ (http://ffdshow-tryout.sourceforge.net/html/en/faq.htm)& Changelog (http://ffdshow-tryout.sourceforge.net/html/en/) updated.

maybe you should add hali splitter has to be installed when using timecodec to faq

[fastplayer]

Will do.

[clsid]

Here are 6 builds for filter performance comparison.

download (http://www.mytempdir.com/1177015)

MSVC2003: mmx/sse/sse2
ICL9: mmx/sse/sse2

Are there other filters besides xsharpen and Denoise3d that benefit from optimizations?

[fastplayer]

From the FAQ:
Generic builds (ICL 9.1) work on Pentium-MMX/AMD K6 or later.
Is this true? I thought SSE1 was minimum requirement for ICL.

[clsid]

ICL supports generic builds just fine. In fact, all my ffdshow ICL9 builds on sourceforge are generic (only MMX is required).

[F_L_C]

This is what foxyshadis said about the builds and filters:

Decoding and resizing do not benefit, but deinterlacing and many ffdshow filters (blur, sharpen, levels, properties, and the goofier less common filters) can show a noticeable boost, along with the non-lavc decoders (like libfaad and libdts). Like I said, it only means anything if you use ffdshow filters, instead of straight decoding.

Edit: and nice to see a new ICL build on the download page :)

[fastplayer]

Thanks, I'll add it to the FAQ later.

@clsid:
Please add links to the tryouts changelog and FAQ to the first post in this thread. Thanks!

Edit: FAQ updated. Please do a "spell check" ;-)

[Inventive Software]

Quick request while my mind's on it still: add MS WMV decoding for WMV7, WMV8, WMV9 as well as libavcodec...

[foxyshadis]

The heavy deinterlacers, kernel and tom's, and currently ICL-only in InnoSetup builds. They're pre-compiled, since they rarely change. So you get the benefit even with generic builds. =p

I could have sworn the WMV9 option was there before, I wonder what happened to it. I must have forgotten to add some dll.

drivel_xxl tells me that ICL can correctly compile lavc and might get a speed boost, but ICL hasn't worked right for me since I updated VC to SP1. :(

[Inventive Software]

Bad move foxy, VC SP1's only really necessary if you do a lotta 64-bit programming, and it's a space hogger too. (Read Avery Lee's blog at virtualdub.org and you'll find out more.)

[clsid]

I tried compiling lavc with ICL, but it gives errors in 'opt.c', 'imgconvert.c' and 'dv.c'. The same errors are present when compiling with plain MSVC2003.

[foxyshadis]

I know, I should uninstall SP1 and reinstall ICL fresh, but I'm sort of fed up with it at the moment. Soon.

Odd thing I noticed: VMR9 mixer mode prevents queue now. I'm pretty sure it was working before, but none of the old builds I tested works, so maybe I'm just absent minded. Didn't update anything related to it as far as I know. I don't remember the benefit anyway, so I'll just leave it off.

Devil pointed out to me that ICL and ICC aren't the same engine. Silly me, I should have known, but Intel's documentation hints toward the opposite. Too bad, it'd be much easier that way.

[fatchick1o9]

Good evening gentlemen.

I wanted to start by thanking you for carrying on the development of ffdshow. I came across ffdshow when I joined the AVS forums when I became interested in building a serious HTPC + home theatre. In time I came to realize that regardless of the hardware that was used, a few essential pieces of software were regularly used and highly praised.

ZoomPlayer
ffdshow
Mediaplayer Classic
etc...

Although I have gone through many pieces of hardware and software, ffdshow has always been the first thing I install on any
machine I setup. Be it my machines or friends PC's that are having problems playing audio or video files.

Thank you for all the work you do. ffdshow is always evolving thanks to the dedication of passionate people like the ones @ doom9 and abroad.

Now, when I watch DVDs, satellite tv, listening to my CD's or my mp3's I always know that I am getting the best possible sound & video quality thanks to a hand full of well written software applications, plugins & codecs.

I do have 1 feature request for ffdshow. This is not a major problem for me but I think it may go a long way towards making the use & configuration of ffdshow a little easier.

Would it be possible to make the audio & video configuration panel of ffdshow resizable I realize that codec performance and quality are of the greatest importance, but as a person that constantly uses the config panels I find it a pain sometimes to have to scroll up and down to select the various sections I want to configure.

I understand that I could just remove the filters from the left panel so that only the ones I need are visible. But this is not always the best option as I sometimes forget I have removed a filter or a Preset profile loads up depending on my application and I spend a lot of time trying to remember what I did that caused the filter to disappear.

If making the audio / video configuration panels resizable is too much effort to accomplish, then please forgive my ignorance as I am not well versed in writing code for applications.

Thanks you for you time and consideration. I appreciate all the hard work you guys put into making ffdshow the best application that it ca be.

Have a good day.

eric g.

P.S. Damn, I registered on the forum just to say that and I still had to wait 5 days just to make my 1st post. :(

[haruhiko_yamagata]

Hello, fatchick1o9.
Interesting request, to make vertical size resizable is not an impossible thing. But it may not be easy to code (not too difficult though).
I would like to prioritize debuging and documentation over feature requests. I may reconsider when I have finished debuging and documentation, but please do not expect too much, because they are almost endless.

[fatchick1o9]

Thanks Mr. Yamagata. Your work on the Queueing and SMP is something I follow more closely than anything else.

I recently put together an AMD 64x2 with dual FX7600GT's specifically for using ffdshow's queuing.

I figured that the development team had other priorities but as the saying goes "The squeaky wheel gets the grease". Well I am just waiting to get lubed up.

eric g.

[fastplayer]

If this was Java code, it'd be relatively easy to adjust vertical size: Just change the frame size and the layout manager does its magic - more or less. Dunno how it's handled in C++ though...

[Jeremy Duncan]

May I ask if it's possible to update the deinterlacers in the deinterlace tab ?
Some of them don't work, and the one that does work, tomsmocomp, causes artifacts when used.

Specifically, I'm asking if Tomsmocomp deinterlacer can be updated, please.

Here's tbarries web page. You can look to see if you have the latest version of Tomsmocomp, please.
Link (http://www3.impacthosting.com/trbarry/)

[the_corona]

Hi guys,

I am very happy with FFDSHOW there is just one request for me regarding 4:3 material on my 16:9 screen.
http://www3.impacthosting.com/trbarry/HDTV.htm

So if I may be so bold and request this :-)

Again thanks a ton for your hard work.

Corona

[Syris2k4]

I too have mused over the resizable config idea. For about 2 years infact... I even went and hacked together a small script to force the window to be resizable - however that was purely thru the window management... so it would randomly crash and burn... still - I think it would be a pretty neat thing to throw in :)

Keep going guys, ffdshow is looking great at the moment :)

[SeeMoreDigital]

I am very happy with FFDSHOW there is just one request for me regarding 4:3 material on my 16:9 screen.
http://www3.impacthosting.com/trbarry/HDTV.htm

So if I may be so bold and request this :-)At the end of the day.... a 4:3 image should be displayed on a 16:9 screen like this: -

http://img249.imageshack.us/img249/8855/43on1697ok.png

Displaying them any other way, by distorting their shape is "just plain wrong"....


Cheers

[foxyshadis]

Hi guys,

I am very happy with FFDSHOW there is just one request for me regarding 4:3 material on my 16:9 screen.
http://www3.impacthosting.com/trbarry/HDTV.htm

So if I may be so bold and request this :-)

Again thanks a ton for your hard work.

Corona

Better done as an avisynth script as proof of concept before trying to integrate anything into ffdshow. So for now, give this a shot:

converttoyuy2().WarpedResize(m(2,width*1.15),height,.7,1).converttoyv12()

ratio=16.0/9.0
bordersize=int(height*ratio-width)
bordersize > 0 ? colorborders(bordersize) : last

function colorborders(clip c, int bordersize) {
global border1 = int(bordersize / 4) * 2
global border2 = bordersize - border1
global colory = 0
global clp = c
temp = c.addborders(bordersize,0,0,0)
scriptclip(temp,"""
overlay(blankclip(last,color_yuv=colory),clp,x=border1)
""")
FrameEvaluate("colory = int(AverageLuma())*65536+int(AverageChromaU())*256+int(AverageChromaV())")
}
function m(int r, float x, string "type") { return x<16 ? 16 : int(round(x/float(r))*r) }

The warpresize parameters can be changed to taste. If you're sizing the whole thing up to fit the screen, you'll need to play with them. Ratio needs to be the desired output ratio. Also, it's pillarboxing only, didn't think about changing it to do both.

Further ambience effects are possible, but this functions as a proof of concept only. I don't especially care for the effect.

[Jeremy Duncan]

Here's the newest Tomsmocomp Link (http://www3.impacthosting.com/trbarry/TomsMoComp.zip)

It was released to the public a few months ago.
Please use it and update the FFDshow Tomsmocomp. :)

[the_corona]

Well...plain wrong might be a bit strong. Sure it changes the image but so do many other filters in ffdshow. If it enhances peoples enjoyment why not add them?

The script sounds awesome, I'll give this a try when I get home. Thanks alot for putting it together.

Looking from how little code (well script) it seems to be in avisynth does it mean that it would be possible to introduce this into ffdshow? Does nobody want their 4:3 content being displayed as efficiently as possible (even if loosing a bit of original image) on their 16:9 screen?

Regarding the "black" frame average colorisation, perhaps this would be even better if each side would only take into account its respective, let's say 1/3, of the image? I'm sure this would need to by tried and depends alot on the user's subjective preferences probably. Some people might even find it highly distracting.

Edit: after trying the script I get an error: "There is no function WarpedResize". When I take out the line though the black borders get their averaged colores nicely!

[clsid]

Here are 6 builds for filter performance comparison.

download (http://www.mytempdir.com/1177015)

MSVC2003: mmx/sse/sse2
ICL9: mmx/sse/sse2
Has anyone tested this yet? I am curious if SSE/SSE2 further improves the performance of ICL builds. Some filter benchmarks:
http://ffdshow-tryout.sourceforge.net/html/en/faq.htm

[fastplayer]

I'm not good at testing but I'd gladly integrate the results into the FAQ :)

[foxyshadis]

Well...plain wrong might be a bit strong. Sure it changes the image but so do many other filters in ffdshow. If it enhances peoples enjoyment why not add them?

The script sounds awesome, I'll give this a try when I get home. Thanks alot for putting it together.

Looking from how little code (well script) it seems to be in avisynth does it mean that it would be possible to introduce this into ffdshow? Does nobody want their 4:3 content being displayed as efficiently as possible (even if loosing a bit of original image) on their 16:9 screen?
It's much harder to do something as a filter than as a script. (And harder to change once it's done.) The benefit is that it's often faster running. I'll ask around and see what others think, if it's popular I'll fine-tune it.

Regarding the "black" frame average colorisation, perhaps this would be even better if each side would only take into account its respective, let's say 1/3, of the image? I'm sure this would need to by tried and depends alot on the user's subjective preferences probably. Some people might even find it highly distracting.
That's what I meant about further ambient effects. It'd take a few days to put together different versions and test them to see what actually looks decent. (My hunch was actually smearing some of the video into the borders, fading into average. But I have no idea how nice or ugly it'd look.)

Edit: after trying the script I get an error: "There is no function WarpedResize". When I take out the line though the black borders get their averaged colores nicely!

Requires this plugin (http://www.trbarry.com/SimpleResize.zip) in <c:\program files\avisynth 2.5\plugins> or wherever your install is, for the warpedresize. It should look a little better than no resize or a linear resize.

[wyrd]

Has anyone tested this yet? I am curious if SSE/SSE2 further improves the performance of ICL builds. Some filter benchmarks:
http://ffdshow-tryout.sourceforge.net/html/en/faq.htm
my test result here (http://tirnanog.fate.jp/tmp/comparison/clsid_ffdshow_comparison.txt).
used sample (58MB) (http://trailers.divx.com/Dreamworks/Madagascar_HD.zip)
Thanks :)

[F_L_C]

Which one of the developers makes ICL SSE builds? It is clearly superior when xsharpen is enabled and competitive w/ ICL9 when enabling denoise3D, at least according to wyrd's numbers.

Edit: Oh I get it. You have to download the package from clsid at mytempdir and then rename the ffdshow.ax version that you need. Finally, copy it into the ffdshow root folder overwriting the current ffdshow.ax file. I'll be using ICL9 SSE because I have an overclocked xp mobile CPU and I use all those filters.

[clsid]

If there are enough positive test results, then I probably will enable SSE by default for my ICL builds. SSE vs SSE2 doesn't seem to make much difference.

Denoise3d does not gain much from optimizations, while xsharpen gains a lot.

More tests with other filters are greatly appreciated.

[fastplayer]

These 2 results (in red) are weird. Any explanations?
ffdshow_unicode_sse.ax User: 118s, kernel: 0s, total: 118s, real: 119s, fps: 23.3, dfps: 23.1
ffdshow_unicode_sse.ax User: 76s, kernel: 0s, total: 76s, real: 77s, fps: 36.1, dfps: 35.6
ffdshow_unicode_sse.ax User: 117s, kernel: 0s, total: 117s, real: 118s, fps: 23.4, dfps: 23.3
ffdshow_unicode_sse.ax User: 118s, kernel: 0s, total: 118s, real: 119s, fps: 23.3, dfps: 23.0

ffdshow_unicode_sse2.ax User: 117s, kernel: 0s, total: 117s, real: 118s, fps: 23.5, dfps: 23.3
ffdshow_unicode_sse2.ax User: 75s, kernel: 0s, total: 75s, real: 76s, fps: 36.5, dfps: 36.2
ffdshow_unicode_sse2.ax User: 116s, kernel: 0s, total: 117s, real: 117s, fps: 23.6, dfps: 23.5
ffdshow_unicode_sse2.ax User: 118s, kernel: 0s, total: 118s, real: 118s, fps: 23.3, dfps: 23.3

[wyrd]

I think it strange that this result too.:)
but I cannot yet make a cause clear.
This result sometimes occurs in sse/sse2.

Regards

[fastplayer]

Maybe a bug in timecodec.exe?
Anyway, thanks for the benchmark results, wyrd!
For the FAQ, I will simply ignore both results and calculate the average of the remaining three. Might take a while though...

[clsid]

I don't think it is really needed to add all these results to the FAQ. It is better to make some conclusions based on the results and put those in the FAQ.

Q: Which build has the best decoding performance?
A: All builders use the GCC compiler to build libavcodec, which is the main library used for decoding. This library contains a lot of hand optimized assembly code. Using additional compiler optimizations or advanced instruction sets (such as SSE/SSE2) doesn't give any significant performance increases. So when comparing pure decoding performance, there is almost no difference between builds.

Q: Which build has the best performance if certain filters are used?
A: Optimized builds of ffdshow.ax will have better performance when certain filters are used in ffdshow. This only applies to a few filters. A list of filters is given below. Both ICL9 MMX and GCC SSE builds can give a nice performance boost. ICL9 SSE builds are the fastest. Using SSE2 has no added benefit over SSE.
The following filters benefit from optimizations:
* xsharpen
* denoise3d
* todo: add more

Q: What is the biggest difference between builds?
A: The main difference between builds is the compiler that is used for compiling ffdshow.ax. There are four different compilers being used in the different builds: MSVC71, MSVC80, GCC and ICL9.

Q: Which build is the most stable?
A: MSVC builds of ffdshow.ax are probably the most stable. However, there are no known issues with GCC builds either. For ICL builds there is only one known issue, it crashes on files created with a very old and buggy build of x264, but you'll probably encounter those files in real life. So from a stability point of view, the build doesn't matter.

[fastplayer]

I have made a simple bar chart (http://ffdshow-tryout.sourceforge.net/html/images/benchmark.png) with the results from wyrd. I used dfps as the relevant unit. Is that correct?

[clsid]

dfps is correct. To avoid confusion you should add "dfps" to the horizontal axis, add perhaps also "(higher = better)".

Please rename the items on the vertical axis to:
ICL9 (SSE2)
ICL9 (SSE)
ICL9 (generic)
MSVC71 (SSE2)
MSVC71 (SSE)
MSVC71 (generic)

[fastplayer]

OK, I will add your Q&A's to the FAQ too.

[SeeMoreDigital]

Does anybody know whether anybody is working on a E-AC3 audio decoder filter for FFdshow?

By-the-way, the guys over at MPlayer seem to created a patch for their player; -

http://svn.mplayerhq.hu/ffmpeg?view=rev&revision=7687


Cheers

[dragongodz]

the guys over at MPlayer seem to created a patch for their player
actually its from the ffmpeg devs list and is for ffmpeg(note its the ffmpeg repository which is hosted on mplayer). though mplayer uses ffmpeg libavcodec etc anyway. :)

[Mangix]

some of the mplayer developers also work on ffmpeg though :)

[_xxl]

I have added x264 encoder and update it to rev 620.
Is it a good or a bad idea?
http://rapidshare.com/files/13292917/ffdshow_rev816_20070125_xxl.exe.html
EDIT:
Please test this encoder, if it works ok.

[shon3i]

It is good for all vfw users, and this is much better because support cqm's and other cli stuff witch isn't inclded in dts wfv build.

Other ppl's who hate vfw will say is that is bad idea, and whole vfw thing again can make usual flame.

edit, you should incude 2pass encoding

[klinika]

If this is not the right place to bring forth bugs, beg pardon. But in the mean time, there's a tiny one that's been there as long as I remember

http://luuranko.freeshell.org/ffdshow_subs.jpg

The lines get cut off by a few pixels on right side if the font is angled very much. The font in the picture is in italic style.

[_xxl]

Wmv9lib encoder in the original ffdshow project:
http://img140.imagevenue.com/loc1/th_64552_try_122_1lo.jpg (http://img140.imagevenue.com/img.php?image=64552_try_122_1lo.jpg)
In tryouts:
http://img171.imagevenue.com/loc8/th_64551_org_122_8lo.jpg (http://img171.imagevenue.com/img.php?image=64551_org_122_8lo.jpg)
This is a new bug?

[dragongodz]

some of the mplayer developers also work on ffmpeg though
you missed the point. SMD said this was created for mplayer when it was actually done for ffmpeg. mplayer uses ffmpeg just like ffdshow does. so these changes would be an advantage to ffdhow whereas mplayer alone changes would not.

[Keepitsimple]

Hi, Im sorry if this is the wrong thread for this.

When I use ffdshow to decode x.264 films which are 720p there is obviously no need for deinterlacing. However when I play some 576i videos with the format vob I need to deinterlace. Is there some way to make this automatic or do I have to tick the deinterlace option?

Also I use mpc as my player. When I check the filters I recognice ffdshow, ac3filter, haali renderer, Realtek soundcard. However there is a filter called audio switcher. Whats that?

[ExtraEye]

for some reason flash video stopped working on ffdshow for me.
i have the splitter and other videos work. when i check the filter list ffdshow video isn't listed even though it's supposed to load (it isn't selected as "disabled" on codec list and zoom player is set up to load it when playing flash video).
for now i watch flash videos on avisynth.

[SeeMoreDigital]

you missed the point. SMD said this was created for mplayer when it was actually done for ffmpeg. mplayer uses ffmpeg just like ffdshow does. so these changes would be an advantage to ffdhow whereas mplayer alone changes would not.Yes... sorry about that guys!

So it sounds like adding E-AC3 support to FFdshow, could technically be possible!

[clsid]

Remember that ffdshow is only a decoder. If E-AC3 requires splitter/parser changes, which I suspect it does, then you are out of luck.

[haruhiko_yamagata]

May I ask if it's possible to update the deinterlacers in the deinterlace tab ?
Some of them don't work, and the one that does work, tomsmocomp, causes artifacts when used.

Specifically, I'm asking if Tomsmocomp deinterlacer can be updated, please.

Here's tbarries web page. You can look to see if you have the latest version of Tomsmocomp, please.
Link (http://www3.impacthosting.com/trbarry/)
>>Some of them don't work
I can't reproduce this. Please upload a sample.

[LoRd_MuldeR]

Here's the newest Tomsmocomp Link (http://www3.impacthosting.com/trbarry/TomsMoComp.zip)

It was released to the public a few months ago.
Please use it and update the FFDshow Tomsmocomp. :)

The release date is 2003/06/29, so it was released ~4 years ago :sly:

[clsid]

Latest version is from April 22nd 2006.

[LoRd_MuldeR]

Latest version is from April 22nd 2006.

Where you got this information from?

On trbarry's homepage it sasys:
TomsMoComp - 2003/06/29 v0.0.1.7

http://www3.impacthosting.com/trbarry/downloads.htm

[clsid]

This file contains 0.0.1.8:
http://www3.impacthosting.com/trbarry/TomsMoComp.zip

[fastplayer]

This trailer (SVQ3/IMA ADPCM) won't playback with MPC:
http://flash.sonypictures.com/classics/angel-a/AngelA_Trailer_2500.mov [148MB]
MPUI has no problems though. Maybe a splitter problem?

[_xxl]

The same with Black Sheep Trailer.mov.
(AVC1/IMA ADPCM)

[fastplayer]

Yep, here too. If IMA is disabled, then both files play back.

[DeathWolf]

with all 7xx & 8xx builds i've tried, Resize+multiply by 2+always creates problems, that is luma looks ok but chroma broken(all green), and that regardless of colorspace&renderer, can anyone confirm?

[foxyshadis]

Just to add to clsid's post:
// ?? Aug 2005 0.0.1.8 Tom Barry Fix bug so TMC60 script works with
// vertical filter for 60 fps output.
// Also add SE=8 support.

So it's just a minor bugfix release of double fps mode.

[siim04]

Hi,

I have been looking for x64 versions of ffdshow, but the latest I found are from 17. May 2006. Are there any newer verions available?

Also, is there any documentation on how to compile ffdshow from source (what is needed)? Maybe I could just compile my own x64 version (preferrably using VS2005).

With best wishes,
siim04

[MacAddict]

with all 7xx & 8xx builds i've tried, Resize+multiply by 2+always creates problems, that is luma looks ok but chroma broken(all green), and that regardless of colorspace&renderer, can anyone confirm?

Not in front of my PC right to test again but I think I was getting the 'green' when resizing plus adding noise. If I turned off the noise function then it appeared normal. Are you using noise as well by chance?

[DeathWolf]

yep! that's it! noise+resize=green
without noise it's fine

[MacAddict]

This noise bug has been around before but unfortunately I've forgotten who fixed it last year. I could be way off but I thought the code had some sort of rounding error in the code. I definitely confirmed the noise+resize options as a problem on another PC here.

Edit- I dont see the bug occur if I enable 'noise' after the video has already started playing. Easy work around :-)

[wyrd]

This trailer (SVQ3/IMA ADPCM) won't playback with MPC:
http://flash.sonypictures.com/classics/angel-a/AngelA_Trailer_2500.mov [148MB]
MPUI has no problems though. Maybe a splitter problem?

It work fine for me in MPC,ZoomPlayer and Nero show time with nero digital parser.
(I love haali but I'm not using for .mov play)
http://tirnanog.fate.jp/tmp/snap/angelA_t.jpg
(http://tirnanog.fate.jp/tmp/snap/angelA.jpg)
Regards
//EDIT
ahh.. sorry It has don't work for me in ffdshow's ima4.
I misunderstood, because this (http://tirnanog.fate.jp/tmp/sample/mov/surge-1-16-B-ima4.mov) is work fine in ffdshow.

[dragongodz]

just so this is not left hanging ...

Remember that ffdshow is only a decoder. If E-AC3 requires splitter/parser changes, which I suspect it does, then you are out of luck.
well there are already solutions for demuxxing evobs
http://forum.doom9.org/showthread.php?t=120652
so who is to say there wont be splitter/parser changes mad to current mpeg2 program stream splitters, since evob is just a variant of that.

its just something for the future currently anyway.

[DeathWolf]

yeah if noise is after the resize it's fine, but wouldnt that change the order things are applied?
that is i'd rather have the noise before resizing rather than after:)

[MacAddict]

@DeathWolf

Let me clarify now that I'm awake and had some coffee. Deselect noise from the ffdshow config window and just open your movie normally. As the movie/clip starts to play, open the ffdshow config window and select the noise to be enabled and you wont see the bug occur. For me, the bug will not occur if I'm already playing video and then enable noise.

[fastplayer]

drevil_xxl fixed the IMA problems in his latest 844 build (http://downloads.sourceforge.net/ffdshow-tryout/ffdshow_rev844_20070128_xxl.exe). I don't know which one of the changes in rev844 did the trick but both trailers now play fine in MPC. :thanks:

[DeathWolf]

@MacAddict
That still doesnt remove the fact there is a prob;)
I personally like to grain a bit things systematically when watching:)

[fastplayer]

@MacAddict
That still doesnt remove the fact there is a prob;)
I personally like to grain a bit things systematically when watching:)
Have you checked out one of the latest builds? h_yamagata fixed it in rev841:
fix noise + resize = green problem

[gribelu]

I have a quick question about the Letterbox option on the Subtitles page.
I have a 16:10 display so i'm using that ratio.. and it works fine for 16:9 and other widescreen formats.
But with 4:3 files it zooms in instead of leaving it alone?
IMO it should only apply padding when it's needed and otherwise do nothing... or create borders on the sides of the 4:3 file..

[GmorG McRoth]

I would like to suggest adding "Scrap.exe" to list "don't use FFDSHOW in:" Game Scrapland depends on windows implementation of mpeg-4 decoder and seems to be not compatible with FFDSHOW.

[wyrd]

@drevil_xxl
thanks for fix.

-----
I have the following problems.(freeze at end of the file)
previous fastplayer's report:
http://forum.doom9.org/showthread.php?p=924533#post924533
Edit: One odd thing happened though: MPC doesn't rewind even though "Rewind when done playing" is checked. I don't know if this has anything to do with it...

I have tested in gabest's/ndpaser splitter with MPC/ZoomPlayer.
(also i tryed even change both queue output on/off and renderer(system/vmr9),but a same results..)
and i tested with SVQ3 file of here (http://tirnanog.fate.jp/tmp/sample/mov/).

STORMAHEAD_PodsTrailer[SVQ3+MP3].mov : reproduce
StormAhead_TFN[SVQ3+MP3].mov : reproduce
00000397_quicktime_HOOG[SVQ3+AAC].mov : reproduce
t28_large[SVQ3+mp3].mov : reproduce
noitulove[SVQ3+AAC].mov : reproduce
madonna_BbandHi[SVQ3+MP3].mov : reproduce
zgundam_3-trailer480[SVQ3+mp3].mov : reproduce
basic-instinct-640[SVQ3+MP3] .mov : fine
appleseed_large[SVQ3+QDM2].mov : fine
tomb_raider_10th_anniversary_edition_video_1_-_trailer_anglais_29586[SVQ3+twos].mov : fine
gt4_tgs2k3[SVQ3+IMA4].mov : reproduce
[note:video is wrong - It was only MPC+QTalt(non DSHOW) that it can play in my player's(vlc,mplayer..) afaik.]

Can anyone confirm?

@authors
Thanks for your efforts again.

Regards

[fastplayer]

Yes, I can confirm. The funny thing is that the video renderer still keeps counting frames, even though the file has reached the end.
Another thing happened though when playing the basic-instinct-640[SVQ3+MP3] .mov trailer. I get this message as already reported here:
http://forum.doom9.org/showthread.php?p=790882#post790882
The trailer plays back fine though after hitting the close button.

Setup: MPC latest (system default renderer), ffdshow build 844, no Nero/Haali splitters used/installed.

[SeeMoreDigital]

By any chance, can Nero's NDIG FOURCC code be added to the list of "Generic MPEG-4" or "Other MPEG-4" codes...

That said, is there any particular reason why these two fields can't be merged?

[Peuj]

@drevil_xxl
thanks for fix.

-----
I have the following problems.(freeze at end of the file)
previous fastplayer's report:
http://forum.doom9.org/showthread.php?p=924533#post924533


I have tested in gabest's/ndpaser splitter with MPC/ZoomPlayer.
(also i tryed even change both queue output on/off and renderer(system/vmr9),but a same results..)
and i tested with SVQ3 file of here (http://tirnanog.fate.jp/tmp/sample/mov/).

STORMAHEAD_PodsTrailer[SVQ3+MP3].mov : reproduce
StormAhead_TFN[SVQ3+MP3].mov : reproduce
00000397_quicktime_HOOG[SVQ3+AAC].mov : reproduce
t28_large[SVQ3+mp3].mov : reproduce
noitulove[SVQ3+AAC].mov : reproduce
madonna_BbandHi[SVQ3+MP3].mov : reproduce
zgundam_3-trailer480[SVQ3+mp3].mov : reproduce
basic-instinct-640[SVQ3+MP3] .mov : fine
appleseed_large[SVQ3+QDM2].mov : fine
tomb_raider_10th_anniversary_edition_video_1_-_trailer_anglais_29586[SVQ3+twos].mov : fine
gt4_tgs2k3[SVQ3+IMA4].mov : reproduce
[note:video is wrong - It was only MPC+QTalt(non DSHOW) that it can play in my player's(vlc,mplayer..) afaik.]

Can anyone confirm?
....


I've just tried STORMAHEAD_PodsTrailer[SVQ3+MP3].mov and StormAhead_TFN[SVQ3+MP3].mov and I don't have any issues.


Another thing happened though when playing the basic-instinct-640[SVQ3+MP3] .mov trailer. I get this message as already reported here:
http://forum.doom9.org/showthread.php?p=790882#post790882
The trailer plays back fine though after hitting the close button.

Same issue for me.

I use lastest MPC + ffdshow build 844 and Haali renderer.

[clsid]

By any chance, can Nero's NDIG FOURCC code be added to the list of "Generic MPEG-4" or "Other MPEG-4" codes...

That said, is there any particular reason why these two fields can't be merged?
Added and merged.

[_xxl]

Latest snow encoder doesn't work.
http://img103.imagevenue.com/loc145/th_08255_snow_122_145lo.jpg (http://img103.imagevenue.com/img.php?image=08255_snow_122_145lo.jpg)

[SeeMoreDigital]

Added and merged.How cool are you.... Nice one :D


EDIT: Bummer... NDIG, it not there :(

http://img258.imageshack.us/img258/5594/ffdshowjq0.png

[clsid]

I successfully re-encoded 'divx625.avi' with Snow. Could be that it doesn't like certain inputs. It doesn't surprise me, considering the many FIXMEs in the code.

Does anyone have a Snow sample file for me? (encoded with a recent version of mencoder)

NDIG is only missing in the notes (fixed in SVN). Playback should be working.

[wyrd]

Does anyone have a Snow sample file for me? (encoded with a recent version of mencoder)

sample here (http://tirnanog.fate.jp/tmp/sample/snow/)

Encode by MEncoder 1.0pre8-3.4.2 (2006/06/12)

Decode results:
vlc0.8.6a : ok
mplayer pre8,rc1,20061220(celtic_druid) : ok
mplayer 20070126 : ng
ffmpeg(ffplay)rev7333,7375 : ok
ffmpeg(ffplay)rev7430,7579 : ng
latest ffdshow : ng

related post by akupenguin
http://forum.doom9.org/showthread.php?p=918839#post918839

Best regards

[Jeremy Duncan]

>>Some of them don't work
I can't reproduce this. Please upload a sample.

bypass: Doesn't work
Linear interpolation: Works, but does a very poor job, showing comb artifacts, and a rippling line artifact.
Linear blending: Works, but removes a lot of detail
cubic interpolation: Works, but same problem with jagged comb artifacts, and rippling line artifact.
cubic blending: Doesn't work
median: Works, but suffers from a rippling ling artifact
tomsmocomp: Works, Search effort 1, vertical filter on, but suffers from rippling ling artifact
dgbob: Works if Artifact protection is checked and the second and third modes are used
framerate doubler: doesn't work
ffmpeg deinterlacer: Works but has jagged comb artifacts and rippling line artifact.
5 tap lowpass: works
kernal deinterlacer: Doesn't work
kernal bob: doesn't work

In FFDshow. Deinterlacing has to be before resize & aspect to function.

Rippling Line (http://www.slibe.com/fullimage/0428471a-Rippling_Line.png)

No Deinterlacing (http://www.slibe.com/fullimage/83f3de8c-No_deinterlacing.png)

This is using rev 814, Jan 22 2007 17:17:32 msvc 2003, x86 unicode.

The image is a pattern on the hqv test disk, and the whole dvd is interlaced.

[Yong]

sample here (http://tirnanog.fate.jp/tmp/sample/snow/)

Encode by MEncoder 1.0pre8-3.4.2 (2006/06/12)

Decode results:
vlc0.8.6a : ok
mplayer pre8,rc1,20061220(celtic_druid) : ok
mplayer 20070126 : ng
ffmpeg(ffplay)rev7333,7375 : ok
ffmpeg(ffplay)rev7430,7579 : ng
latest ffdshow : ng

related post by akupenguin
http://forum.doom9.org/showthread.php?p=918839#post918839

Best regards

your sample crashing here, mplayer svn-r22069.

snow encoded by ffdshow(clsid) r853 is both play fine in mplayer svn-r22069 and ffdshow too;)

[foxyshadis]

Jeremy, are you sure about the artifacts being ffdshow's fault? I'd try the avisynth versions, because I think the artifacts (stray combing, rippling, etc) are simply common artifacts of certain classes of deinterlacing algorithms. In fact, the HQV disc is specifically made to show those shortcomings in glaring detail. That's why Didée is working on MCBob, one deinterlacer for all things. ;)

Can you be more specific on "doesn't work"? Errors, major arifacts (like discoloration or blocking), simply does nothing at all?

[SeeMoreDigital]

NDIG is only missing in the notes (fixed in SVN). Playback should be working.It does not seem to be: -

http://img222.imageshack.us/img222/7229/ndigdj9.png

Admittedly, except for testing, it going to be "very" rare for users to de-mux NeroDigtal (Part-2) video streams out of the .MP4 container to the .AVI container (and correct the 4CC code).... But, that's me ;)


Cheers

[wyrd]

@yong
thanks for reply.
I've confirmed it. (It works fine with both mplayer and r853)
test sample uploaded, but i do not tuning, so it's bad quality:D
also i could not encode with recent mencoder(rc1 or later)...
off topic:can anyone do it?

//EDIT:
oops, I'm embarrassed:P.
I could encode in rc1 or later.(sorry,my setup was too old...)
This file (http://tirnanog.fate.jp/tmp/sample/snow/sakuratan%5bsnow(mencoder20070126)+mp3%5d.mkv) works fine in r853.
(with mencoder svn-r22069(20070126))
Best Regards

[Leak]

That's why Didée is working on MCBob, one deinterlacer for all things. ;)
Are you sure that's going to be feasible to use in realtime during playback? :D

On another note - would it be feasible to integrate TIVTC into ffdshow, as an IVTC step with optional deinterlacing for still-combed frames? (i.e. as another sub-page of the deinterlacing filter)

That sure would give a smoother result than simply deinterlacing telecined sources - but I'm not sure if ffdshow can handle reducing the number of frames per second and looking/decoding ahead a few images. Any ideas?

And while I'm asking - has anyone ever had a chance to look at DVD decoding? Or rather, what exactly is making it non-functional currently?

(Incidentally, those two things are the only two things I'm still missing to get my anime DVDs to look really smooth on my projector... :p)

[HeadBangeR77]

And while I'm asking - has anyone ever had a chance to look at DVD decoding? Or rather, what exactly is making it non-functional currently?
Have you tried? ;) Then plz do ;) For me it works perfectly, apart from seeking, but that's a splitter problem (MPC) I think. There are artefacts (large 8x8 squares) while seeking, but the playback itself is already ok. :)

Have a look at some screencaps here:
http://forum.doom9.org/showpost.php?p=945832&postcount=37

cheers,
HDBR77

PS. libmpeg2 of course.

[foxyshadis]

Avisynth framerate change works fine as long as you don't do anything with VFR. (Trivia: It interpolates the new timecodes between the old ones, using the difference between input and output framerates, since DShow has no real framerate.) Of course, you wouldn't normally want to IVTC a VFR file anyway, so it's sort of academic.

The only problem with IVTC is that decimation takes most of the processing power at the beginning of the cycle, so there are sharp periodic bursts of CPU usage. A good way to make it look jerky after all the effort not to. ;_; Queuing can help, of course. The weave/deinterlace usage is pretty regular though.

TIVTC could probably be integrated, but all the framerate stuff would have to be modified (and big chunks would be useless for plain viewing); since tritical's always coming out with new fixes it would mean more maintenance effort to stay on top of.

I think the only thing wrong with DVD decoding is 2-thread decoding at the moment, but I might be wrong.

[SeeMoreDigital]

Thanks clsid,

MPEG-4 with NDIG 4CC code works a treat now, with ffdshow_rev858_20070130_clsid.exe


Cheers

[Jeremy Duncan]

bypass: Does not Deinterlace.
Linear interpolation: Works but has combing artifact, and Rippling line Effect.
Linear blending: Works but removes too much detail.
cubic interpolation: Works but has combing artifact, and Rippling line Effect.
cubic blending: Works
median: Works but has Rippling line Effect.
tomsmocomp: Works but has Rippling line Effect. Latest Avisynth version does not have rippling line artifact.
dgbob: Only first mode works. Has combing artifact, and Rippling line Effect.
framerate doubler: Does not Deinterlace.
ffmpeg deinterlacer: Works but has slight combing artifact, and Rippling line Effect.
5 tap lowpass: Works.
kernal deinterlacer: Does not Deinterlace.
kernal bob: Does not Deinterlace.

Here's a picture of Tomsmocomp version 0.0.1.8 used in FFDshow Avisynth, tomsmocomp(1,5,1).
No Rippling (http://www.slibe.com/fullimage/af9352f3-Tomsmocomp.png)

[MacAddict]

libmpeg2 seems to decode DVD's just fine for me but the problem comes in when seeking. I get random crashes/lockups on DVD's when seeking or navigating the DVD menu.

[Leak]

libmpeg2 seems to decode DVD's just fine for me but the problem comes in when seeking. I get random crashes/lockups on DVD's when seeking or navigating the DVD menu.
That's what I meant, basically. Like the first US volume of Ergo Proxy crashing MPC as soon as you click on "Play all"... :(

[lexor]

there is something very wrong with Audio Decoder.

If you open it up (the gui I mean) and enable only the Mixer, set Speaker Configuration to 2/0 (this far everything works fine), but now enable the LFE checkbox and your system will reboot once you try to play back a movie. Setting Speakers to 2/1 fixes it. But still even if 2/0 is a wrong setting, the computer shouldn't reboot. (gray out LFE checkbox in unsupported settings maybe?)

ffdshow 844, xp sp2 + haali splitter (xvid + mp3 input)

[wozio]

Hi,

Is it possible to include option to disable adjusting color space in ffdshow? I would like to have it in native scale as it was from decoder without any recalculations. It shouldn't be much work.

Regards
Piotr

[haruhiko_yamagata]

Revision 864
H.264 MBAFF hardware deinterlacing support.
It was a request from stream47.

Samples originally (http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?t=144&start=41) posted by stream47 (In Japanese).

English translation
ASP top field first
AVI (http://asnxload.dip.jp/files/ppfx00340.avi)
MP4 (http://asnxload.dip.jp/files/ppfx00341.mp4)

ASP bottom field first
AVI (http://asnxload.dip.jp/files/ppfx00346.avi)
MP4 (http://asnxload.dip.jp/files/ppfx00347.mp4)

AVC
MP4 (http://asnxload.dip.jp/files/ppfx00344.mp4)
MKV (http://asnxload.dip.jp/files/ppfx00345.mkv)

MPEG4-ASP hw-deinterlacing is supported only by libavcodec, not xvid.

[fastplayer]

If you open it up (the gui I mean) and enable only the Mixer, set Speaker Configuration to 2/0 (this far everything works fine), but now enable the LFE checkbox and your system will reboot once you try to play back a movie.
Doesn't happen for me. No matter how many times I check/uncheck the LFE box. Maybe the problem is your sound card driver?

[foxyshadis]

Doesn't happen for me. No matter how many times I check/uncheck the LFE box. Maybe the problem is your sound card driver?

It couldn't really be anything but the soundcard driver, since an entirely user-mode thing like ffdshow could never cause a reboot.

lexor, either disable reboot on bug check, or use Event Viewer (eventvwr.msc) to determine the probable cause of the fault. It'll give you a something.sys, you might have to google around to find out just which driver that is.

Revision 864
H.264 MBAFF hardware deinterlacing support.

Awesome, getting AVIVO and Purevideo to work should be much simpler now.

[SeeMoreDigital]

MPEG4-ASP hw-deinterlacing is supported only by libavcodec, not xvid.Finally.... This is excellent news!

EDIT: I've only been banging on about the need to provide proper support for decoding interlaced MPEG-4 Part-2 sources for years (plain old "Bob" would have done in the meantime). I wonder if this move has been promted by DivX's recent announcement regarding their 6.5 direct-show decoder offering support too?!

[fastplayer]

Finally.... This is excellent news!
It gets even better: the "SVQ3 freeze at the end of file"-bug has been fixed, too.
:thanks:

[Seb.26]

Hi world ...

First : thanks to the new dev team for the work !

FFDShow reborn is really a great news !!!

Second, where can I find a list of each FFDShow's filter whith the color space allowed by the filter please ? ...

( This to avoid the YV12 conversion done by some filters ... )

Thanks !
Seb

[_xxl]

I can't play .evo files on my computer properly.
Tested with AMD x2 + 2gb DDR2 + 7900 + ffdshow + MPC.
No Open-Source solution can handle it.:(
Maybe someday libavcodec h264 decoder is going to be multithreaded.
ftp://mplayerhq.hu/MPlayer/samples/evob/Departed%20Trailer.EVO

[fastplayer]

ftp://mplayerhq.hu/MPlayer/samples/evob/Departed%20Trailer.EVO
I hear no sound. GSpot tells me it's AC3... Is a special splitter/decoder needed?

[clsid]

Check out this thread for info about .evo:
http://forum.doom9.org/showthread.php?t=120652

[SeeMoreDigital]

I hear no sound. GSpot tells me it's AC3... Is a special splitter/decoder needed?Yes... HD-DVD and Blu-ray use E-AC3 (http://en.wikipedia.org/wiki/E-AC3)....

[fastplayer]

Thanks for the info guys. Certainly a nice read for a rainy afternoon :D

[Scoty]

which settings needet for best video quality ? i use the latest Build with MPC.

[foxyshadis]

Scoty, see this thread (http://forum.doom9.org/showthread.php?t=115727) for starters. There is no one setting that will magically be best for everything, though, and it's pretty cpu intensive.

[fastplayer]

Would it be possible to make the priority over which type of subtitle gets loaded by default configurable? For example, like a drop-down box where you can choose between the subtypes like SRT, VOB, SUB etc.

[SeeMoreDigital]

Are there any "Revision 864" builds (with interlaced support) available yet?

The most up-to date builds on SourceForge stop at 30 Jan 07.


Cheers

[clsid]

I'll upload latest build asap.

VP5/VP6 decoding seems to be broken in current SVN revision. Picture freezes at the beginning. Sample file (http://www.zshare.net/video/vp62-avi.html).

Edit: revision 865 broke it

[cc979]

got error building rev866

make -C ffmpeg
make[1]: Entering directory `/home/user/svn/ffdshow-tryout/trunk/src/ffmpeg'
make[1]: *** No rule to make target `libavcodec/lzo.h', needed by `libavcodec/cscd.o'. Stop.
make[1]: Leaving directory `/home/user/svn/ffdshow-tryout/trunk/src/ffmpeg'
make: *** [FFMPEG] Error 2

[_xxl]

Can you compile latest GCC for MinGW?

[cc979]

Can you compile latest GCC for MinGW?

i tried other day, still not compile ffdshow with gcc-4.2 or gcc-4.3 still patches needed i think

[_xxl]

Maybe GCC prerelease-4.1.2-20070128?

[clsid]

@cc979, make clean and remove cscd.d

[JarrettH]

Revision 867

-added libavcodec vc-1

Will this open my DRMd WMV-HD files?:sly:

[Scoty]

Revision 867

-added libavcodec vc-1

Will this open my DRMd WMV-HD files?:sly:

link ?

[haruhiko_yamagata]

Would it be possible to make the priority over which type of subtitle gets loaded by default configurable? For example, like a drop-down box where you can choose between the subtypes like SRT, VOB, SUB etc.
A case ffdshow is inputing multiple formats of subtitles?
What kind of case do you assume?

[fastplayer]

A case ffdshow is inputing multiple formats of subtitles?
What kind of case do you assume?
For example: Consider the following 4 files in one directory:

video.avi
video.idx
video.sub
video.srt

Now when you start playing the AVI file, ffdshow opens the Vobsub subtitle (IDX/SUB) by default. Is it possible to make this default behavior configurable? Like making SRT files to be opened by default.

[haruhiko_yamagata]

For example: Consider the following 4 files in one directory:

video.avi
video.idx
video.sub
video.srt

Now when you start playing the AVI file, ffdshow opens the Vobsub subtitle (IDX/SUB) by default. Is it possible to make this default behavior configurable? Like making SRT files to be opened by default.
OK, it should not be too difficult.

// EDIT
added at rev 870.
Ideally the extensions in the dialog should be dragable so that user can drag and re-order.
Considering the users who use this feature would be experts, the simplified dialog may be enough.

[fastplayer]

OK, it should not be too difficult.

// EDIT
added at rev 870.
Ideally the extensions in the dialog should be dragable so that user can drag and re-order.

(:thanks:)^10
Wow, this feature required more changes than I expected... :o
I hope this feature is not just useful for me but for others as well!
Considering the users who use this feature would be experts, the simplified dialog may be enough.
I would've been happy even if this feature was exposed via the registry :D

[Leak]

Avisynth framerate change works fine as long as you don't do anything with VFR. (Trivia: It interpolates the new timecodes between the old ones, using the difference between input and output framerates, since DShow has no real framerate.)
Yeah, just changing the framerate works, but all hell (well, not really... ;)) breaks loose once you try to actually return less frames than the original source has:

Info()
SelectEvery(30,0)
AssumeFPS(1)should really give you a 1 FPS slideshow but plays normally - except for the frame numbers printed by Info which increment by 30... :(

Obviously AviSynth's seeking ahead does nothing - shouldn't the ffdshow source decode ahead 30 frames in that case? Referencing past frames (within limits) is handled by AviSynth's cache usually, but of course it can't look into the future... :D

And as far TIVTC is concerned, TFM works of course, but TDecimate doesn't, for similar reasons.

As for TIVTC producing load spikes for each cycle, like you said earlier - I rather doubt it. After all, it bases it's decisions on metrics calculated by comparing each frame with it's previous and next frame, for each frame in the cycle, which can be done as soon as each frame is coming in, which distributes the load nicely. At least I didn't get any weird CPU spikes when decoding a ripped VOB file and using the AviSynth filter with just TFM() followed by TDecimate() - CPU usage stayed somewhere between 20 and 30 % according to Task Manager.

When figuring out which frame to drop, it just looks at the metrics that have been calculated for the past frame, which shouldn't be noticeable performance-wise. Of course, what you do need to do is decoding ahead 5 (or 10) frames before returning the first frame, then always buffering those frames and returning frames from that buffer - but that'd only give you a slight hiccup/load spike at the very beginning or after seeking (unless you ignore seeking and let TDecimate figure out what to do - shouldn't be treated much different than any other scene change), which shouldn't be too much of a problem.

So here's my question - is it possible to request some frames ahead via DirectShow when ffdshow is not doing the decoding (dunno how the graph thing works exactly)? I guess it should at least be possible as soon as ffdshow itself is used as the decoder - but it's not done currently for AviSynth, as far as I can see...

EDIT: Yeah, looking at the source for Tffdshow_source, the passed in frame number (n) isn't used, ever.

np: Mira Calix - Belonging (No Longer Mix) (Eyes Set Against The Sun)

[fastplayer]

about rev870:
When setting the vertical subtitle position from 90% (default) to 95%, then only the first row of subtitles with 2+ lines is shown.
In case of vobsubs, nothing is displayed.
The only way to bypass this, is to reduce font size resp. the scale factor when dealing with vobsubs but this makes the subtitles harder to read... :o

[JarrettH]

I'm just waiting on a new ICL from clsid :D

Will these VC-1 codecs be able to playback DRMd WMV-HD stuff?

[Kado]

I've been waiting for the wmv9 codec to be added as a decoding alternative to libavcodec. Thanks.
However after doing some tests I have registered the following cpu usage (dual-core cpu):
Internal wmv9 decoder => 10 to 25% cpu usage (no artifacts).
ffdshow libavcodec => 25 to 45% cpu usage (some artifacts like trails or something)
ffdshow wmv9 => 20 to 35% cpu usage (no artifacts).
Internal wmv9 + ffdshow raw processing => 20 to 35% cpu usage (no artifacts).

Is there any special reason to ffdshow use more cpu?

Details:
Video: Windows Media Video 9 1280x720 29.97fps 3.8Mbit (VBR)
Media Player Classic v6.4.9.0 v3.2+ (same settings for all codecs)
CPU: Intel Pentium D 930@3.6ghz
ffdshow build 875 with no extras like post-processing active.

Kado

[haruhiko_yamagata]

If this is not the right place to bring forth bugs, beg pardon. But in the mean time, there's a tiny one that's been there as long as I remember

http://luuranko.freeshell.org/ffdshow_subs.jpg

The lines get cut off by a few pixels on right side if the font is angled very much. The font in the picture is in italic style.
Thank you for report and sory to be late.
fixed at rev 883.
Some other bugs related to italic characters are fixed at the same time.

[_xxl]

http://img165.imagevenue.com/loc72/th_79303_Capture_122_72lo.jpg (http://img165.imagevenue.com/img.php?image=79303_Capture_122_72lo.jpg)
Anybody knows what's wrong with mingw gcc-build-4.1.2-121479?

[Yong]

http://img165.imagevenue.com/loc72/th_79303_Capture_122_72lo.jpg (http://img165.imagevenue.com/img.php?image=79303_Capture_122_72lo.jpg)
Anybody knows what's wrong with mingw gcc-build-4.1.2-121479?

gcc4x cant find the standard include dir.
i have the same problem too, when the ffdshow source code was in my drive D:, mingw gcc was in C:,
then i moved the ffdshow source code back to drive C: and the problem is gone :p

[cweb]

gcc4x cant find the standard include dir.
i have the same problem too, when the ffdshow source code was in my drive D:, mingw gcc was in C:,
then i moved the ffdshow source code back to drive C: and the problem is gone :p

gcc4 isn't stable under mingw, that's why there are no builds using gcc4 in mingw.

[HeadBangeR77]

about rev870:
When setting the vertical subtitle position from 90% (default) to 95%, then only the first row of subtitles with 2+ lines is shown.
In case of vobsubs, nothing is displayed.
The only way to bypass this, is to reduce font size resp. the scale factor when dealing with vobsubs but this makes the subtitles harder to read... :o
It may sound obvious, but did you enabled letterboxing for subs? I also have the vertical position set to 95%, since I like the first line to be displayed at the bottom of the main screen, and the 2nd line beneath. With no letterboxing the subs have simply no place to be displayed on. ;)

[fastplayer]

It may sound obvious, but did you enabled letterboxing for subs? I also have the vertical position set to 95%, since I like the first line to be displayed at the bottom of the main screen, and the 2nd line beneath. With no letterboxing the subs have simply no place to be displayed on. ;)
No, it's off. Something has definitely changed in the way subs are positioned. I have to change font size from my usual '17' to '9' in order to get 2 lines displayed. But that makes it unreadable... With the pre-870 builds I can go down to 100% with vertical subtitle position and nothing gets snipped off...

Edit: I've taken a look at Haruhiko's implementation and I can't see anything regarding positioning that has been changed.
Me looking at C++ code means nothing on the other hand... :D

[HeadBangeR77]

I see, then enabling letterboxing might be a good temporary solution, since you wouldn't have to change the fontsize then (hopefully).

Was it after the mentioned implementation (one page earlier), when the bug occured? I'm asking, since I'm still using clsid's svn 843 (icl 9, sse) and won't upgrade to a version with buggy subtitles, because I use them too often.

Btw. Whole lot of good work on ffdshow, guys! Much appreciated. :)

[fastplayer]

I see, then enabling letterboxing might be a good temporary solution, since you wouldn't have to change the fontsize then (hopefully).
I will check it out later and report back.
Was it after the mentioned implementation (one page earlier), when the bug occured?
Yes. I always use the exact same settings for subs and noticed the issue right away.
I'm asking, since I'm still using clsid's svn 843 (icl 9, sse) and won't upgrade to a version with buggy subtitles, because I use them too often.
Most important part of ffdshow for me apart from the decoding :D
Btw. Whole lot of good work on ffdshow, guys! Much appreciated. :)
Yep, the download numbers reflect this :)

[haruhiko_yamagata]

As long as I can see, beta1 has the same behavior, but yes, it should be fixed anyway.

[HeadBangeR77]

Hello,

I've already exchanged PMs with foxyshadis on this matter. I will try to give as much info as possible, and if it's necessary, I will cut a short sample of my encode.

Bug: reendering of fog and sky in some dark scenes; happens only from time to time, but looks really nasty.
At the top there's also some sort of distortion, looks like the effect caused by FillMargins() sometimes, when we want to add some artificial pixels after cropping, to get mod16 resolution.

Source: MPEG-2 -> XviD 1.1.2 encode with GK, packed bitstream and Qpel (this may matter, I think), CQM Didee's SixOfNine.
I've also got many short samples, including the scene with buggy decoding, encoded with various CQMs, either with GK+VDM or with the latest VDub 1.7.0. In all of them the bug is present, so it's not only this one encode.

Player: MPC, the latest celtic druid's build i.e. mplayerc.rev611-3.2kxp.
At the time I first encountered the bug I was using the latest official 6.4.9.0, so I would rule out the player in this case.

ffdshow revison: I marked it first with some old drevil_xxl's sse revision from December, reproduced with now with clsid's svn 843 ICL 9 sse, so it seems it might be a permanent bug, imo.

The bug occurs, while decoding the mentioned XviD encode(s) with libavcodec and idct set to auto. The auto-setting defaults to libmpeg2 every time.

The bug doesn't occur while decoding with XviD 1.1.2 via ffdshow, with idct set to auto. The auto setting defaults as above (always).

The bug doesn't occur while decoding with libavcodec and idct set manually to XviD MMX!

So :
libavcodec + auto/libmpeg2 -> bug
libavcodec + manual XviD MMX -> fine

Gonna post screenshots soon (libavcodec decoding is always one frame further then XviD, I've taken exactly the same frames, checked many times).

cheers,
HDBR77

[fastplayer]

@HeadBangeR77:
Just checked with clsid's latest 885 build and with letterboxing enabled the issue remains. I tried all aspect ratios.

[cc979]

gcc4x cant find the standard include dir.
i have the same problem too, when the ffdshow source code was in my drive D:, mingw gcc was in C:,
then i moved the ffdshow source code back to drive C: and the problem is gone :p

i did not get them problems when i compiled ffdshow with gcc-4.1.2, it compiled fine but ffdshow.ax would crash

i have d:/msys/1.0, d:/mingw and d:\msys\1.0\dx (with libs and include) and also and empty include dir d:\msys\1.0\include (because there is a problem when compiling gcc if it is not there)

apart from my compiles are fine

[HeadBangeR77]

@HeadBangeR77:
Just checked with clsid's latest 885 build and with letterboxing enabled the issue remains. I tried all aspect ratios.
Sorry to hear that ... I've checked also, and with the mentioned above clsid's svn 843 (icl 9, sse) everything is fine.

***

I'm gonna post screenshots in this post soon.

1) Elizabeth's face, background textures

a) XviD + libmpeg2
http://www.hidebehind.com/thumbs3/D03A6090.png (http://www.hidebehind.com/D03A6090)

b) libavcodec + libmpeg2
http://www.hidebehind.com/thumbs3/92AC06.png (http://www.hidebehind.com/92AC06)

c) liibavcodec + XviD MMX
http://www.hidebehind.com/thumbs3/B160BAC0.png (http://www.hidebehind.com/B160BAC0)

2) The sky, the mist, and the mountain

a) http://www.hidebehind.com/thumbs3/1B331800.png (http://www.hidebehind.com/1B331800)

b) http://www.hidebehind.com/thumbs3/BE9A6778.png (http://www.hidebehind.com/BE9A6778)

c) http://www.hidebehind.com/thumbs3/5054EB6B.png (http://www.hidebehind.com/5054EB6B)

Plz save them to your hdd, and vie at full screen. I've got more, but I think the above will suffice.

[chros]

Grabbing frames bug ?

I have tried to grab a single 1 frame to png through ffdshow, but MPC has crashed ... Doesn't matter if I use VMR 9 renderless or the haali's renderer ... It was a h264 video which ffdshow decoded ...

Can anybody confirm this ?
Thanks

[clsid]

So :
libavcodec + auto/libmpeg2 -> bug
libavcodec + manual XviD MMX -> fine
This is not really a bug, but a general iDCT problem. If the iDCT doesn't match the DCT that was used by the encoder, then the video won't be decoded properly. The auto detection in ffdshow needs improvement.

[HeadBangeR77]

This is not really a bug, but a general iDCT problem. If the iDCT doesn't match the DCT that was used by the encoder, then the video won't be decoded properly. The auto detection in ffdshow needs improvement.
So as I thought, this is not the libavcodec itself, but the improper iDCT - that's why I checked libavcodec + XviD MMX iDCT ...

Now, wait a sec - I can't reproduce the bug with XviD + libmpeg2 set manually (?!)

@ chros
Yeah, I can confirm that with MPC, ffdshow, and H.264 in general - I use MPC's save image instead.

[Yong]

i did not get them problems when i compiled ffdshow with gcc-4.1.2, it compiled fine but ffdshow.ax would crash

i have d:/msys/1.0, d:/mingw and d:\msys\1.0\dx (with libs and include) and also and empty include dir d:\msys\1.0\include (because there is a problem when compiling gcc if it is not there)

apart from my compiles are fine

hmm, my gcc was obtain from here
http://oss.netfarm.it/mplayer-win32.php
4.0.3, 4.1.1 and 4.2.0
cant use them to compile directx related program like xvid decoder and gpac directx plugin, duno how to fix the error :p
and the directory/files not found problem only occur for me when the compiler was not in same drive as thee source...
gcc4x still problematic though.

[SeeMoreDigital]

Can someone please confirm.... Has support for "interlaced" MPEG-4 Part-2 sources been added. It does not seem to be working for me. Or have I missed something?


Cheers

[clsid]

Now, wait a sec - I can't reproduce the bug with XviD + libmpeg2 set manually (?!)The iDCT setting is for libavcodec only.

[SeeMoreDigital]

The iDCT setting is for libavcodec only.Hmmmm...

Are you able to play the following sample correctly: -

http://www.mytempdir.com/1204197


Cheers

[cc979]

hmm, my gcc was obtain from here
http://oss.netfarm.it/mplayer-win32.php
4.0.3, 4.1.1 and 4.2.0
cant use them to compile directx related program like xvid decoder and gpac directx plugin, duno how to fix the error :p
and the directory/files not found problem only occur for me when the compiler was not in same drive as thee source...
gcc4x still problematic though.

i compiled mine myself, and noticed path finding issues maybe a msys problem with linux style pathnames

i will look into it further

[clsid]

Are you able to play the following sample correctly: -

http://www.mytempdir.com/1204197

CheersYes. It plays correctly.

[SeeMoreDigital]

Yes. It plays correctly.Really!

I'm obviously doing something very wrong... I think I would benefit from "step-by-step" set-up instructions... For dummies (li-ka-me)

[clsid]

I haven't done anything special to make it play. It should just work out-of-the-box.

Screenshot:
http://i15.tinypic.com/2rdu748.jpg

[fastplayer]

For me it's heavily interlaced with the 866 build...

[foxyshadis]

Well, you need to turn deinterlacing on (it's field-shifted). Wait, was this question about whether the file plays at all, or whether it automatically triggers the deinterlacing? I don't remember the latter being added.

[Leak]

Well, you need to turn deinterlacing on (it's field-shifted). Wait, was this question about whether the file plays at all, or whether it automatically triggers the deinterlacing? I don't remember the latter being added.
Heh. That'd be another reason to integrate TIVTC - just turn on TFM to fix field shifts without giving the image the full deinterlacing smackdown... :D

I guess I'll have to take a look at the AviSynth filter in the near future - but I'd be happy if someone could tell me if it's possible to read a few frames ahead in the input stream for the filter, or if it's at least possible to just kind-of return nothing from the filter while buffering up a few input frames. (Or if that's not possible - would returning the same frame with the same timestamp a few times work?)

Either of that should make it possible to implement limited support for seeking (within a configurable range) in AviSynth and also for dropping frames. Of course, ideally TIVTC should be integrated as a "real" ffdshow filter with a config dialog for ease of use...

But I guess I should try to get ffdshow to build first...

[haruhiko_yamagata]

Can someone please confirm.... Has support for "interlaced" MPEG-4 Part-2 sources been added. It does not seem to be working for me. Or have I missed something?


Cheers
H.264 hardware deinterlacing support was added recently.
Hardware deinterlacing support for MPEG4-ASP works even with 20051115.
libavcodec works, xvid does not.

[haruhiko_yamagata]

Hello,

I've already exchanged PMs with foxyshadis on this matter. I will try to give as much info as possible, and if it's necessary, I will cut a short sample of my encode.

Bug: reendering of fog and sky in some dark scenes; happens only from time to time, but looks really nasty.

Interesting. Improving auto detection of IDCT looks very hard, but worth while. Please post the sample movie.

[SeeMoreDigital]

H.264 hardware deinterlacing support was added recently.
Hardware deinterlacing support for MPEG4-ASP works even with 20051115.
libavcodec works, xvid does not.Thanks, I have "libavcodec" selected...

Is the de-interlacing "automated" (ie: does FFdshow detect the interlaced and field position from header in the MPEG-4 stream) or is some "manual" intervention required?


Cheers

[Peuj]

Hmmmm...

Are you able to play the following sample correctly: -

http://www.mytempdir.com/1204197


Cheers

something wrong with me too.
I sue the build 885.
IDTC is set to libmpeg2.
(if I enable deinterlacing the video is ok.)

http://img409.imageshack.us/img409/9021/samplesk0.jpg (http://imageshack.us)

[haruhiko_yamagata]

Thanks, I have "libavcodec" selected...

Is the de-interlacing "automated" (ie: does FFdshow detect the interlaced and field position from header in the MPEG-4 stream) or is some "manual" intervention required?


Cheers
Manual intervention is required.
ffdshow detects intercaced video, but does not detect "Bob" or "Weave".
If Output->Use overlay mixer->HW deinterlacing is checked (and Method "Bob" is selected), ffdshow detects interlaced video and set interlaced flag in output media type. If the video is true progressive, you don't have any side effect.
The problem is, if the video is interlaced "Weave" (like film source DVD), you'll have to uncheck HW deinterlacing or select Method "Weave".

[foxyshadis]

Heh. That'd be another reason to integrate TIVTC - just turn on TFM to fix field shifts without giving the image the full deinterlacing smackdown... :D

I guess I'll have to take a look at the AviSynth filter in the near future - but I'd be happy if someone could tell me if it's possible to read a few frames ahead in the input stream for the filter, or if it's at least possible to just kind-of return nothing from the filter while buffering up a few input frames. (Or if that's not possible - would returning the same frame with the same timestamp a few times work?)

Either of that should make it possible to implement limited support for seeking (within a configurable range) in AviSynth and also for dropping frames. Of course, ideally TIVTC should be integrated as a "real" ffdshow filter with a config dialog for ease of use...

But I guess I should try to get ffdshow to build first...

Read your other post too, haven't answered because while I looked, I don't have any good answers or any really useful insight at this point, and I think I'm going to have to make a special version just to poke and prod at it for a while.

You might have better ideas of how to hack vfw into the directshow framework, as well. Another method is just to inject global per-frame variables into the environment and have tfm/tivtc modify them directly (ignoring them when used in pure avisynth). Requiring direct plugin support is annoying, but everything else I've thought of is a really lousy hack.

[ilpippo80]

I have a small problem with hw deinterlacing in mpeg-2 video:
ffdshow only: OK
MPC internal decoder only: OK
DScaler filter only: OK
MPC internal decoder + ffdshow (uncompressed): OK
DScaler + ffdshow (uncompressed): doesn't work
I tried to change the settings of Dscaler and ffdshow, but no matter what I did, I had no way to enable hardware deinterlacing when dscaler was connected with ffdshow...

[SeeMoreDigital]

Manual intervention is required.
ffdshow detects intercaced video, but does not detect "Bob" or "Weave".
If Output->Use overlay mixer->HW deinterlacing is checked (and Method "Bob" is selected), ffdshow detects interlaced video and set interlaced flag in output media type. If the video is true progressive, you don't have any side effect.
The problem is, if the video is interlaced "Weave" (like film source DVD), you'll have to uncheck HW deinterlacing or select Method "Weave".Okay thanks.... I've just tried it. And it does not seem to be working for me. To confirm, here are my settings: -

http://img228.imageshack.us/img228/8683/ffdshow866tx5.png


Cheers

[Peuj]

Okay thanks.... I've just tried it. And it does not seem to be working for me.....

Cheers

Well it works for me (enable HW + Bob) but you have to close ffdshow to load the change

[SeeMoreDigital]

Well it works for me (enable HW + Bob) but you have to close ffdshow to load the changeIndeed...

I fiddled about with Media Player Classics "Output" filters and discovered that when set to "System Default" the de-interlacing works okay. But when set to "VMR7 or VMR9 (Renderless)" it does not.... Well not for me... Bummer!

[sillKotscha]

But when set to "VMR7 or VMR9 (Renderless)" it does not.... Well not for me... Bummer!

this whole render discussion(s) is drivin' me nuts... for myself I came to the conclusion, if I set the renderer to VMR7 (windowed) at least on my system I have always highest interoperability...

[clsid]

I used the software deinterlacer in ffdshow. That will work with VMR-9 too.

[chros]

@ chros
Yeah, I can confirm that with MPC, ffdshow, and H.264 in general - I use MPC's save image instead.
Thanks for the answer ... The problem is that MPC can't save into png ... :(

[fastplayer]

Sorry to hear that ... I've checked also, and with the mentioned above clsid's svn 843 (icl 9, sse) everything is fine.
Issue fixed by Yamagata in rev888. Latest 890 build is living proof :D
:thanks:

[Hans Ohlo]

hi,
i have big problems with skyoneHD streams to decode with ffshow (h264 - libavcodec). i use the latest haali splitter for the transport stream splitting and ffshow as decoder for transcoding to x264. with graphedit i setup filter graphs to use ffdshow, cyberlinks h264 decoder and coreavc. for me ffdshow gives the best picture so i want to go with it. coreavc has still the color issues on these kind of 1080 streams (paff i think). cyberlink has kind of a more blury look.

problem is that at some points in the stream there are i think little errors which all the other decoder eat up but only ffdshow chokes on. on older builds there were heavy blocky images on these locations but with the latest builds 850+ ffdshow crashes asking for a jit debugger (i have vs 2005 installed). i have uploaded a little sample of these errors so that you guys can have a look at them.

the latest ffdshow build i tried is 890, i set IDCT to auto, encoder workarounds to autodetect, error resilience to compliant and error concealment to none.

please have a look, because this is starting to cost very much sleep for me ;)

sample:
http://www.megaupload.com/?d=0Z6EP0RW
http://www.mytempdir.com/1205486 (mirror)

[clsid]

That sample also crashes FFplay revision 7768.

Submit your sample to the FFmpeg developers:
http://ffmpeg.mplayerhq.hu/bugreports.html

[SeeMoreDigital]

Decoding seems to be a little more stable if you re-mux the video stream into the .MP4 container (along with re-encoded AC3-to-AAC audio)...

Cheers

[Hans Ohlo]

yes, but remuxing introduces new errors on other places of the stream due to skipping of frames while demuxing (mplayer, h264to etc. tested, all the same result, so no remuxing is the way i chose).

i will report the decoding bug

[HeadBangeR77]

this whole render discussion(s) is drivin' me nuts... for myself I came to the conclusion, if I set the renderer to VMR7 (windowed) at least on my system I have always highest interoperability...
That's my conclusion too :) + no tv-scale bug + taking screenshots is possible.

@ chros
Yep, but after I've taken whole lot of them, I convert them to pngs - that's twice the work, but saves a lot of hdd space (especially when running more passes during the conversion - 2nd one already does some extra packing, the difference is rather large).

Update!
Revision 891 - Directory Listing
Modified Tue Feb 6 16:43:33 2007 UTC (8 hours, 40 minutes ago) by clsid2

Fixed grabbing to PNG.
Ta. :)

@ fastplayer
Thanks very much for the info. :)
Unfortunatelly the latest revision of clsid's sse icl9 is 885 (poped out today I think), so there's no reason for me to upgrade, since I can confirm that sse icl9 utilizes less cpu with various filters than any other compilation. I've gotta wait a while then. ;)

[HeadBangeR77]

Interesting. Improving auto detection of IDCT looks very hard, but worth while. Please post the sample movie.
I'm gonna cut a short clip, where I've marked the described errors (libav + auto/libmpeg2), directstreamcopy the video and upload tomorrow. I've scanned for similar errors thorugh the movie, and they appear from time to time, yet the mentioned part (see my screenshots approx. 2 pages earlier) is the worst.

PS. When I play with settings, sometimes the auto mode works all right, sometimes not, I've just marked. Nevertheless libav + XviD MMX iDCT is the only one 100% without bugs (apart from crappy decoding of packed bitstream, but that's another story).

cheers,
HDBR77

[Lancerevo]

Hi guys, I was just wondering whether is there any ffdshow build that supports multithreaded h264 playback? I have a 1920x1080 video that does 50% of CPU Usage and I've checked in Task manager that it maxes out 1 core and leaves the other un-used.

[_xxl]

Ask ffmpeg developers.Many people won't be happy if ffmpeg h264 decoder is multithreaded, but more people will be.If you understand what I mean...

[SeeMoreDigital]

yes, but remuxing introduces new errors on other places of the stream due to skipping of frames while demuxing (mplayer, h264to etc. tested, all the same result, so no remuxing is the way i chose).Why would a de-muxer "skip frames"?.... If it did skip frames, then this anomaly should be reported also!

By all accounts, the MP4 container is the perfect partner for MPEG-4 AVC streams. And has far less overhead.

[Hans Ohlo]

i also like the mp4 container, but if i demux the stream and pack it into mp4 (MP4Box -add part1_video.h264#video:fps=25 part1.mp4) the video is not playing smoothly (haali splitter or gabest splitter with coreavc, ffdshow...). in a mkv container it plays fine but shows some errors on other locations than the ffdshow crash (which remains at the same location btw. no matter what container).

i do not seem to get answers on my bugreport on the ffmpeg list :/

[SeeMoreDigital]

Hmmm...

Just to be certain.... Are you using the newest versions of MP4Box. And are you making sure the frame rate speed of your MP4 matches the sources frame rate speed?

[Hans Ohlo]

Hmmm...

Just to be certain.... Are you using the newest versions of MP4Box. And are you making sure the frame rate speed of your MP4 matches the sources frame rate speed?
MP4Box - GPAC version 0.4.3-DEV
GPAC Copyright: (c) Jean Le Feuvre 2000-2005
(c) ENST 2005-200X

and yes 25fps is the source framerate. i also tried without interleaving but the juddery playback stays

[SeeMoreDigital]

MP4Box - GPAC version 0.4.3-DEV
GPAC Copyright: (c) Jean Le Feuvre 2000-2005
(c) ENST 2005-200X

and yes 25fps is the source framerate. i also tried without interleaving but the juddery playback staysThanks.... I suppose it did no harm to check.

Sadly I can't test/play such high resolutions AVC sources with my "slow" PC.... Which is a real bummer :(

[HeadBangeR77]

@ haruhiko_yamagata:

I've cut a short sample, just uploading it + some more png screenshots. I will update my post, when I'm done.

http://xasonline.info/headbanger/ffdshow/POTC.Black.Pearl.Sample.avi

http://xasonline.info/headbanger/ffdshow/ffdshow.rar

cheers,
HDBR77

[fastplayer]

I can't notice the difference between IDCT set to libmpeg2 or Xvid MMX... The distortions look bad to me with both settings (look at Keira's chest @ 0:24 and 0:33min). Am I missing something here?

[HeadBangeR77]

I can't notice the difference between IDCT set to libmpeg2 or Xvid MMX... The distortions look bad to me with both settings (look at Keira's chest @ 0:24 and 0:33min). Am I missing something here?
I can't reproduce the distortions with libav + Xvid MMX (I've also tried Xvid-Xvid alone), only with libav + libmpeg2 (+eyebrow, Orlando's also). The auto function for iDCT still works as it wants to - I've got both distorted and proper decoding, depending on its mood. :D

PS. If it's the chest alone, then you could be looking at XviD encoding of dark scenes in its worst, though CC was added, and Didee's SixOfNine throws quants 2-3/3-4.

[fastplayer]

OK, I was just wondering because you had the error pinned down at some specific scenes/frames.

[HeadBangeR77]

OK, I was just wondering because you had the error pinned down at some specific scenes/frames.
The time-range you've given is full of decoding errors with libav + libmpeg2 - she looks as if an alien was trying to get out from beneath her skin.

[fastplayer]

PS. If it's the chest alone, then you could be looking at XviD encoding of dark scenes in its worst, though CC was added, and Didee's SixOfNine throws quants 2-3/3-4.
Even the sea fog scene at the end of the sample looks terrible - with both IDCT modes. To be honest though, I've bever seen "good" fog in Xvid/DivX encodes. There's always some sort of "wobbling" effect...

[HeadBangeR77]

Even the sea fog scene at the end of the sample looks terrible - with both IDCT modes. To be honest though, I've bever seen "good" fog in Xvid/DivX encodes. There's always some sort of "wobbling" effect...
Have you tried it with XviD? Looks way better, and not so bad for an XviD encode (?)

Blah - you can compare the screens, I've uploaded them, so you will know for sure what I'm talking about all the time. :D

[fastplayer]

Have you tried it with XviD? Looks way better, and not so bad for an XviD encode (?)
Less noise in the Keira-in-bed scene and less "wobbling" in the sea fog at the end of the sample (using Xvid 1.1.2 via ffdshow and "standalone").

[_xxl]

Maybe xvid 1.1.2 is "more" standard-compliant decoder.

[HeadBangeR77]

Maybe xvid 1.1.2 is "more" standard-compliant decoder.
Still, there's one thing which has just come to my mind:
1) I can decode the above sample properly via ffdshow using either XviD 1.1.2 library or libavcodec with manually set XviD MMX iDCT - clsid's svn 843, sse, icl9.
2) Fastplayer can decode the sample properly only when using XviD library. Libavcodec + libmpeg2 idCT and libavcodec + XviD MMX are both causing artifacts - rev. ???

Another difference between builds? Shouldn't be - if both the decoding library and iDCT are set manually. :confused:

[haruhiko_yamagata]

:confused: Becasue QPEL was introduced in MPEG4,
libmpeg2-idct even does not know about QPEL.
So libmpeg2-idct must be excluded when QPEL is detected. :confused:

Is this correct? I'm n00b to these things, any advice is wellcome.

[fastplayer]

2) Fastplayer can decode the sample properly only when using XviD library. Libavcodec + libmpeg2 idCT and libavcodec + XviD MMX are both causing artifacts - rev. ???
No, he can't. Shoot him! :angry:
I did a frame comparison and came up with the same conclusions like you:
http://simpleupload.net/download/44120/idct.rar.html

[KoD]

HeadBangeR77, what encoder options have you used for that sample file ?

[HeadBangeR77]

HeadBangeR77, what encoder options have you used for that sample file ?

2 max. consecutive B-frames (ratio=1.62, offset=0), packed bitstream, qpel, to name the ones that should matter while decoding, no GMC.
To name the ones that shouldn't matter at all: MSP 6, VHQ 4, VHQ for b-frames, Chroma ME, trellis, chroma optimizer, CC 15/6.

If the matrice (Didee's SixOfNine) could have an influence on the bug, I've ruled that out: the sample is cut from my 2-pass encode, but I've seen the same things happening with my 1-pass samples, while I was testings CQMs, with 11 other matrices, including H.263 quantization and MPEG standard. :(

@ haruhiko: I haven't got the faintest idea, unfortunatelly.

@ fastplayer: sh**, I can't view files at filefactory - my ISP's issue, 1/4 of webpages doesn't load; might be sth with proxy :confused:
Thanks :)

[fastplayer]

@ fastplayer: sh**, I can't view files at filefactory - my ISP's issue, 1/4 of webpages doesn't load; might be sth with proxy :confused:
http://simpleupload.net/download/44120/idct.rar.html

[igor1st]

I can't reproduce the distortions with libav + Xvid MMX (I've also tried Xvid-Xvid alone), only with libav + libmpeg2 (+eyebrow, Orlando's also). The auto function for iDCT still works as it wants to - I've got both distorted and proper decoding, depending on its mood. :D

I made quick frame to frame comparison with your sample:
lavc xvid/auto (yes, auto works always as it should) identical to xvid 1.1.2.
lavc libmpeg2 not identical to xvid 1.1.2.

[clsid]

:confused: Becasue QPEL was introduced in MPEG4,
libmpeg2-idct even does not know about QPEL.
So libmpeg2-idct must be excluded when QPEL is detected. :confused:

Is this correct? I'm n00b to these things, any advice is wellcome.
From what I could find with the search, the Xvid encoder will use a different DCT when QPEL is activated. So that would explain why a different iDCT is needed during decoding.

http://forum.doom9.org/showthread.php?t=104143&highlight=qpel+idct

[fastplayer]

I can confirm this: Setting IDCT to 'auto' produces same result as Xvid and lavc+Xvid MMX. I've updated link with frame comparisons (see above).

[Jeremy Duncan]

Hello, I'm hoping that leakkerneldeint (http://neuron2.net/kerneldeint/leakkerneldeint154.zip) can be put into the Deinterlacers section of FFDshow in the next rev.

[clsid]

I can confirm this: Setting IDCT to 'auto' produces same result as Xvid and lavc+Xvid MMX. I've updated link with frame comparisons (see above).
You mean that 'auto' choses the correct iDCT? In that case we might want to consider making 'auto' the default again.


I just did a benchmark with a non-interlaced H.264 file. Having HW deinterlacing enabled in the output settings hurts performance a bit. In this one test about 16% lower dfps. But it will probably depend on the power of the graphics card. Mine is pretty weak. So my recommendation to everyone is too keep it turned off (the default) if you mostly play progressive files and to use a software deinterlacer when needed.

[fastplayer]

You mean that 'auto' choses the correct iDCT? In that case we might want to consider making 'auto' the default again.
In this particular case, yes. I remember the IDCT being 'auto' by default some time ago. I don't know why it has been changed to 'libmpeg2'.

By the way, is there an option in ffdshow that allows me to see what IDCT is being currently used (when it's set to 'auto')?

[Jeremy Duncan]

haruhiko_yamagata,

I tried the HW Deinterlacing in the Output Tab.
- I set the output to YV12 only, and FFDshow Video decoder would not show up, but FFDshow Audio would.
- If I set it to YUY2, FFDshow Video decoder showed up, but the HW deinterlacing was off.
I tried other and all Output Colorspaces, but they all set it to Software mode deinterlacing.

I have a few questions.
- What Output colorspace do I need to USE HW Deinterlacing: Bob ?
- Why can't I just use YV12 ?

I'm using the Nvidia Video Codec, and Media Player Classic.
I've tried VMR9 renderless, and default in the MPC output tab.
I've tried fiddling with the video card properties.
I'm using FFDshow rev 814

:angry:

[Hans Ohlo]

hi,
a little question, when are ffmpeg patches in the ffdshow builds? do they use the same repository?

[ouchi]

Big thanks to the developers of FFDShow.

I'm not sure that this is the right place to talk about the difficulties from using FFDShow.
And English is not my mother tongue.
Your patience will be really appreciated.

Unlike PowerDVD 7, v5 has no compatability issue with FFDShow.
But in recent FFDShow builds after the rev657_20061209_clsid, there is no way that make it work with my PowerDVD 5 but to stop using the blacklist which is very useful for me.
Excluding PowerDVD.exe from the list does nothing in my PC and I don't know why.

Is this designed intentionally? If so, how can I bypass it?
Or is there something that I overlooked or misunderstood in configuration?

[Bathrone]

Flac audio playback is broken. When installing builds with the flac option enabled, and playing it back in wmp11 no playback occurs and an error message is generated "wmp cannot play the file.......not supported...."

Also, why are the cslid and drevilxxx builds different for default installs? I see vorbis tremor has different defaults.

[vortex_hl]

Flac audio playback is broken. When installing builds with the flac option enabled, and playing it back in wmp11 no playback occurs and an error message is generated "wmp cannot play the file.......not supported...."

you need flac source filter. this already noted in audio codecs section in ffdshow.

[haruhiko_yamagata]

You mean that 'auto' choses the correct iDCT? In that case we might want to consider making 'auto' the default again.
I think 'auto' is best for most users.

[haruhiko_yamagata]

By the way, is there an option in ffdshow that allows me to see what IDCT is being currently used (when it's set to 'auto')?
I'm working on it :D .

[haruhiko_yamagata]

haruhiko_yamagata,

I tried the HW Deinterlacing in the Output Tab.
- I set the output to YV12 only, and FFDshow Video decoder would not show up, but FFDshow Audio would.
- If I set it to YUY2, FFDshow Video decoder showed up, but the HW deinterlacing was off.
I tried other and all Output Colorspaces, but they all set it to Software mode deinterlacing.

I have a few questions.
- What Output colorspace do I need to USE HW Deinterlacing: Bob ?
- Why can't I just use YV12 ?

I'm using the Nvidia Video Codec, and Media Player Classic.
I've tried VMR9 renderless, and default in the MPC output tab.
I've tried fiddling with the video card properties.
I'm using FFDshow rev 814

:angry:
YV12 and HW deinterlacing is not supported. I don't know why, but milan judged to disable when HW deinterlacing was checked. He should have good reason to do so, it is not safe to enable it without testing broadly. Because we don't have that many video cards, it is better to leave it as it is.

HW deinterlacing should work with YUY2, if it does not work, it should be nVidia's issue.

[haruhiko_yamagata]

Big thanks to the developers of FFDShow.

I'm not sure that this is the right place to talk about the difficulties from using FFDShow.
And English is not my mother tongue.
Your patience will be really appreciated.

Unlike PowerDVD 7, v5 has no compatability issue with FFDShow.
But in recent FFDShow builds after the rev657_20061209_clsid, there is no way that make it work with my PowerDVD 5 but to stop using the blacklist which is very useful for me.
Excluding PowerDVD.exe from the list does nothing in my PC and I don't know why.

Is this designed intentionally? If so, how can I bypass it?
Or is there something that I overlooked or misunderstood in configuration?
Thank you for report.
It was intentionally designed, I didn't know PowerDVD5 worked though. There is no way to bypass currently. Is there way to check version of PowerDVD?

[_xxl]

Unlike PowerDVD 7, v5 has no compatability issue with FFDShow.
But in recent FFDShow builds after the rev657_20061209_clsid, there is no way that make it work with my PowerDVD 5 but to stop using the blacklist which is very useful for me.
Excluding PowerDVD.exe from the list does nothing in my PC and I don't know why.
Is this designed intentionally? If so, how can I bypass it?
Or is there something that I overlooked or misunderstood in configuration?
You could rename Powerdvd.exe to Powerdvd5.exe.

[fastplayer]

hi,
a little question, when are ffmpeg patches in the ffdshow builds? do they use the same repository?
Our devs here only apply selected changes from the ffmpeg repo.
I'm working on it :D .
Haha, my subliminal message worked :D

[Hans Ohlo]

Our devs here only apply selected changes from the ffmpeg repo.
i am asking because of my last bugreport regarding the h264 decoder crash with sykoneHD streams. i hope they fix it soon and the changes migrate into ffdshow.

[KoD]

If the stream is buggy, then there's no bug in the decoder. At best it's a "feature request": please show mercy to buggy streams. In either case, it's the ffmpeg user list where this has to be posted.

[haruhiko_yamagata]

If the stream is buggy, then there's no bug in the decoder. At best it's a "feature request": please show mercy to buggy streams. In either case, it's the ffmpeg user list where this has to be posted.
I don't think so. He's doing great job there.

[fastplayer]

@Yamagata:
STDMETHODIMP_(const char*) TffdshowDecVideo::get_cuurent_idct(void)
Is this just a typo or did you ran out of operation names? :D I suppose it's the latter otherwise VS would complain but you never know... :D

Edit: Where in the libavcodec code is the part that sets the IDCT when 'auto' is selected? I guess lavc reads out the FOURCC and sets the IDCT based on a pre-defined list but I can't find the part...

[clsid]

There have been a few changes to the H.264 decoding code lately to handle 'bad' data better. So I am pretty confident that the FFmpeg devs will take a look at the sample that Hans submitted to their mailing list.

[haruhiko_yamagata]

@Yamagata:
STDMETHODIMP_(const char*) TffdshowDecVideo::get_cuurent_idct(void)
Is this just a typo or did you ran out of operation names? :D I suppose it's the latter otherwise VS would complain but you never know... :D

Edit: Where in the libavcodec code is the part that sets the IDCT when 'auto' is selected? I guess lavc reads out the FOURCC and sets the IDCT based on a pre-defined list but I can't find the part...
it's typo, fixed.

h263dec.c
if(s->codec_id == CODEC_ID_MPEG4 && s->xvid_build && avctx->idct_algo == FF_IDCT_AUTO && (mm_flags & MM_MMX)){
avctx->idct_algo= FF_IDCT_XVIDMMX;

There are some other. Search FF_IDCT_AUTO.

[fastplayer]

There are some other. Search FF_IDCT_AUTO.
Yes, thank you. I was particularly interested in how ffdshow deals with IDCT and H.264. As far as I can see, H.264 uses a reference IDCT implementation (h264idct.c). So I suppose ffdshow "forwards" to this IDCT instead of to the others.

[clsid]

DivX 4/5/6, 3ivX and FLV1 now use iDCT "Simple MMX".
Xvid now uses iDCT "Xvid MMX".
Other formats show "unknown". Either because the iDCT is hardcoded (like for VP3/5/6) or they don't use a DCT algo.

[_xxl]

http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2007-February/052237.html
ffmpeg doesn't support PAFF interlacing.

[Hans Ohlo]

http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2007-February/052237.html
ffmpeg doesn't support PAFF interlacing.

if you mean my problem. there are only few locations in these streams where ffdshow/ffmpeg fails and crashes. the rest of the streams play perfectly. there was a build where the decoder did not crash (i do not know the exact build number) but showed extreme decoding errors where other decoders (cyberlink, mediaconcept, coreavc) do not show such a behavior.

[ouchi]

Thank you for report.
It was intentionally designed, I didn't know PowerDVD5 worked though. There is no way to bypass currently. Is there way to check version of PowerDVD?

Thanks for your confirmation, haruhiko_yamagata.
Now I'm free of thinking what's wrong with my machine.
Thanks again.

As you already know, v5,v6 and v7 are major version numbers.
Upgrade between those is not free while upgrade between minor builds is free.
So I've just known it's v5 since I purchased it..

Also PowerDVD.exe, the main executable of PowerDVD shows its own build number on the properties -> version tab.
Mine says it's 5.0.0.2027, or PowerDVD 5 build 2027.
But to be precise, 2027 is only for the PowerDVD.exe.
It should be 2214 for the whole PowerDVD, because I installed update patch 2214.

[ouchi]

You could rename Powerdvd.exe to Powerdvd5.exe.

drevil_xxl, thanks for your help.
Unfortunately, it seems that PowerDVD.exe doesn't like to be renamed.
It says "your license is expired...." and refusing to start.
So I had to go back to the original name and everything is OK.

[Thunderbolt8]

could always having deblocking enabled actually have negative quality aspects on videos, where there is no deblocking needed ?

[Leak]

Hello, I'm hoping that leakkerneldeint (http://neuron2.net/kerneldeint/leakkerneldeint154.zip) can be put into the Deinterlacers section of FFDshow in the next rev.
I very much doubt that that will be neccessary, seeing as Milan already included it when he was still working on ffdshow - just select "Kernel deinterlacer" or "Kernel bob" on the "Deinterlacing" page... :D

np: Bola - AMNiON (Soup)

[cc979]

drevil_xxl, thanks for your help.
Unfortunately, it seems that PowerDVD.exe doesn't like to be renamed.
It says "your license is expired...." and refusing to start.
So I had to go back to the original name and everything is OK.

maybe its possible to read version details from registry or file, if someone can code it

[clsid]

Blocking "powerdvd.exe" is currently hardcoded in ffdshow. If Haruhiko removes that, then editing the blacklist should be enough.

[haruhiko_yamagata]

@HeadBangeR77
The bug occurs, while decoding the mentioned XviD encode(s) with libavcodec and idct set to auto. The auto-setting defaults to libmpeg2 every time.
Can you confirm this again?
At least, the sample you posted is OK. Maybe you tested on different sample when you first find this issue.
clsid's latest build is available to see IDCT on OSD.

[haruhiko_yamagata]

@ouchi
rev 908: Everything including PowerDVD7 works fine. Please wait for next builds :D .

//EDIT
Please remove powerdvd.exe from blacklist.

[Jeremy Duncan]

Leak,
Neuron2 made kernel deinterlacer, AND Leak Kernel Deinterlacer. Link (http://neuron2.net/kerneldeint/leakkerneldeint154.zip)
I want the Leak Kernel Deinterlacer in the Deinterlacer section of FFDshow. Pretty please, somebody please put it in there. :)

haruhiko_yamagata,

I can't see the FFDshow video decoder icon when I use YUY2 and use the HW bob deinterlacing. What output colorspace do I use to both see the icon and use HW deinterlacing ?

[dragongodz]

Leak,
Neuron2 made kernel deinterlacer, AND Leak Kernel Deinterlacer.
neuron2 made kernel deint of course but it was actually Leak who made the changes to make Leak kernel deint. the name change, especially the "Leak" part, should have given you a big clue to that.

http://forum.doom9.org/showthread.php?t=81322

[Jeremy Duncan]

Ok.
...

[ouchi]

@ouchi
rev 908: Everything including PowerDVD7 works fine. Please wait for next builds :D .

//EDIT
Please remove powerdvd.exe from blacklist.

@haruhiko_yamagata
Great!
Of course, I can wait..
Thank you.

[haruhiko_yamagata]

I coded for the feature request "Use ffdshow only in:".
http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?p=1000#1000

I think the default should be off, but I would like to add exe names as many as possible so that users can do by just checking it.

Please add as many as possible (Send Private Messages for me or reply there (http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?p=1000#1000)).

[_xxl]

I coded for the feature request "Use ffdshow only in:".
http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?p=1000#1000

I think the default should be off, but I would like to add exe names as many as possible so that users can do by just checking it.

Please add as many as possible (Send Private Messages for me or reply there (http://ffdshow-tryout.sourceforge.net/phpBB2/viewtopic.php?p=1000#1000)).
Why off?
FFdshow should work with some media players.No more games trouble or other untested applications.More like a software compatibility list.

[haruhiko_yamagata]

Why off?
FFdshow should work with some media players.No more games trouble or other untested applications.More like a software compatibility list.
OK, discussion is wellcome. What other people think?
To be on by default, the list has to be comprehensive.

[Keepitsimple]

When is the next stable release expected?

[KoD]

OK, discussion is wellcome. What other people think?
To be on by default, the list has to be comprehensive.

How does this list work ?

- ffdshow can be loaded only by applications on the list ?

It should be easy for the user to add new entries to the list or to remove some. A text box like the one for blacklists that the user can edit on its own would allow such user editing.

[haruhiko_yamagata]

How does this list work ?

- ffdshow can be loaded only by applications on the list ?

It should be easy for the user to add new entries to the list or to remove some. A text box like the one for blacklists that the user can edit on its own would allow such user editing.
http://ffdshow-tryout.sourceforge.net/html/images/useonlyin.png
like this, the text box is small. Adding one is easy, while removing one is not.
Currently the list is alphabetically sorted automatically (Should it be sorted?).

[fastplayer]

The text box certainly wouldn't win the 1st prize in a usability contest but since there's not much space, I think this is the best option.
Sorting makes entries definitely easier to find (especially when you can't scroll).

[clsid]

Edit: I have removed the sections for each builder on the tryouts forums. There is not really a need for those. The only real activity takes place in bugs/features/other.

Shall I also remove language forums that are currently empty?

http://ffdshow-tryout.sourceforge.net/phpBB2/

[HeadBangeR77]

@HeadBangeR77
Can you confirm this again?
At least, the sample you posted is OK. Maybe you tested on different sample when you first find this issue.
clsid's latest build is available to see IDCT on OSD.

With the svn I'm still using (svn 843, clsid, ICL9, sse):

Yes, I can confirm the above. For me it works correctly if:
- I either choose XviD as a decoding library,
- or I choose libavcodec + XviD MMX iDCT manually.

Libav + libmpeg2 is buggy, as shown on screens and tested by the others (thanks, fastplayer), and the auto setting is still playing tricks on me, but it seems only on me, so don't worry - must be some software's witchcraft in my case. ;) ;) ;)

I've just downloaded rev 910 by xxl and rev 895 by clsid (ICL9, sse - since this is what I usually use). I guess the 910 also has the above feature. Gonna test this week, when I find enough time.

cheers,
HDBR77

PS. This is the sample I've cut out of my full encode, and this is the only one I've used for testing so far (the others are short one-pass samples with various cqms, but I haven't tried them with the given revision).

EDIT:
Revision 902 - Directory Listing
Modified Sat Feb 10 12:06:28 2007 UTC (36 hours, 8 minutes ago) by h_yamagata

new OSD item IDCT(libavcodec)
Gonna try auto iDCT with 910 and report back.

[fastplayer]

Shall I also remove language forums that are currently empty?
Yes, if requested they can be re-opened again.

[HeadBangeR77]

OK, discussion is wellcome. What other people think?
To be on by default, the list has to be comprehensive.

The proposed list with applications with which ffdshow will be working (and only with them) is a good idea imo. :)
I don't care if it's on or off by default, however it should containt as many players' executables as possible, if it should be enabled by default, as you stated.

Here is a thread that contains some exotic ones (at least for me) too, so you might have a look at this:
http://forum.doom9.org/showthread.php?t=120112

[clsid]

A lot of video conversion progs should also be added to the list.

Perhaps instead of a tiny textbox, it should get its own page (together with "don't use ffdshow in ..."). Then a scrollable textarea could be used, with 1 item on each row.

[haruhiko_yamagata]

Edit: I have removed the sections for each builder on the tryouts forums. There is not really a need for those. The only real activity takes place in bugs/features/other.

Shall I also remove language forums that are currently empty?

http://ffdshow-tryout.sourceforge.net/phpBB2/
Please remove "Stable build nominations" as well.

[HeadBangeR77]

A lot of video conversion progs should also be added to the list.

Perhaps instead of a tiny textbox, it should get its own page (together with "don't use ffdshow in ..."). Then a scrollable textarea could be used, with 1 item on each row.
This sounds good imo, at least for the sake of transparency for a user like me. :)
Good luck with your work, guys!

[Jeremy Duncan]

Besides checking Bob in HW Deinterlacing in the Output tab. What output colorspace do I use, and why doesn't the FFDshow video decoder icon show up when I use the FFDshow HW deinterlacing feature.

[B.F.]

Can you tell me why deblocking filters don't work in that file?
http://akross.ruweb.net/videos/ac2006/[AKROSS_Con_2006]_AlexDaikou_-_DANSU_preview.avi

[_xxl]

You could use official VP6 decoder.
http://www.free-codecs.com/download_soft.php?d=96&s=73

[haruhiko_yamagata]

Besides checking Bob in HW Deinterlacing in the Output tab. What output colorspace do I use, and why doesn't the FFDshow video decoder icon show up when I use the FFDshow HW deinterlacing feature.
It's the spec of your video card's device driver. Ask the maker of video card rather than here.
All ffdshow does is set interlaced flag on output samples and disable YV12.

[haruhiko_yamagata]

A lot of video conversion progs should also be added to the list.

Perhaps instead of a tiny textbox, it should get its own page (together with "don't use ffdshow in ..."). Then a scrollable textarea could be used, with 1 item on each row.This sounds good imo, at least for the sake of transparency for a user like me. :)
Good luck with your work, guys!
OK, I'll implement the full course.

[B.F.]

You could use official VP6 decoder.
http://www.free-codecs.com/download_soft.php?d=96&s=73

I know.
But why ffdshow deblocking don't do anything?

[ilpippo80]

I have a small problem with hw deinterlacing in mpeg-2 video:
ffdshow only: OK
MPC internal decoder only: OK
DScaler filter only: OK
MPC internal decoder + ffdshow (uncompressed): OK
DScaler + ffdshow (uncompressed): doesn't work
I tried to change the settings of Dscaler and ffdshow, but no matter what I did, I had no way to enable hardware deinterlacing when dscaler was connected with ffdshow...

@ haruhiko
Do you think that also my problem might be hardware specific?
Does anyone else has the same issue?
Thanks

[haruhiko_yamagata]

@ilpippo80
I don't have time to look into now.
Does DScaler without ffdshow support HW deinterlacing?
If not, ffdshow can't. Technically it is possible to label interlaced regardless of the input though.

[ilpippo80]

I just checked again to be sure...
Yes, DScaler alone supports hardware deinterlacing, it's only when I connect DScaler with ffdshow that I hw deinterlacing doesn't work

[_xxl]

I know.
But why ffdshow deblocking don't do anything?
Please try:
http://www.mytempdir.com/1214748

[ilpippo80]

I tried again to tweak the filters settings, and I finally succeded in having dscaler filter and ffdshow connected and working with hardware deinterlacing: I had to set "Force Weave" in dscaler's deinterlaced mode and set "Bob" in ffdshow's hardware deinterlacing. That's weird, though.
I just did some more tests, here is what I discovered:
- if I enable "Set interlaced flag in output media type" in mplayerc internal mpeg2 decoder, and use ffdshow for postprocessing, hw deinterlacing doesn't work in any way;
- in graphedit, if I connect two instances of ffdshow in a row, hw deinterlacing doesn't work in any way.
Is this normal? Is this expected?

[haruhiko_yamagata]

@ilpippo80
Thank you the information. That's very weird and will help debugging.

[Jeremy Duncan]

It's the spec of your video card's device driver. Ask the maker of video card rather than here.
All ffdshow does is set interlaced flag on output samples and disable YV12.

What I am asking for is what steps do I need to do to enable HW deinterlacing in FFDshow.
- I check Bob.
- What Output colorspace do I check.
- Do I need to check anything else.

[Leak]

@ilpippo80
Thank you the information. That's very weird and will help debugging.
Maybe it's just me, but as far as I know most of the graphics card's deinterlacers want an interlaced flag and NV12 colorspace.

At least that's what I gleaned from the DScaler message boards...

np: Death Cab For Cutie - Song For Kelly Huckaby (Facts Version) (You Can Play These Songs With Chords)

[B.F.]

Please try:
http://www.mytempdir.com/1214748

It all work now.
:thanks:

[oddball]

I'm having some issues with black levels in ffdshow when playing back some HD x264 content. I am using VMR9 windowless in Zoom Player 5 and find the black levels are too bright/washed out. If I use COreAVC 1.2 the blacks are OK. I have had to compensate for the black levels in ffdshow with the gamma levels down to 0.88

ffdshow tryout 896

[ExtraEye]

sounds like your outputting to YV12 or something... outputting to RGB should fix this.

[oddball]

Yup! In the pinouts it shows ffdshow as outputting YV12. How do I force RGB?

EDIT: Nevermind found the option to force it. Why doesn't ffdshow output RGB when in VMR mode automatically? I Had to enable overlay and deselect all the options except the RGB tick boxes to get it to display black levels correctly.

EDIT2: Hmm. This is odd. Forcing RGB causes other odd issues. Like some jerking and slow motion playback. It's not taxing the CPU (E6600 CPU) so it's not that. But it does playback very erratically. Not always totally smooth. x264 1080P clip.

EDIT3: CoreAVC has an option called 'Fix VMR9 Color Range' where is the equivalent in ffdshow? I can't get anything to playback with the correct black levels AND smoothly if I force RGB.

[Peuj]

Yup! In the pinouts it shows ffdshow as outputting YV12. How do I force RGB?

go in the Ouput section of ffdshow, just keep selected RGB32

[oddball]

That does not work. Zoom Player locks up completely.

EDIT: OK it locks up in VMR9 windowed, plays but with no picture in renderless and plays but still jittery in windowed.

[foxyshadis]

Do you have the most up-to-date video drivers? All ffdshow does is hand frames off to them. YUY2 is sometimes a good compromise, sometimes it has the same issues as YV12 with bad drivers.

Or you can just use overlay renderer, which has no such black-levels problem.

[vlada]

I have a question regarding DVD support - there is still the checkbox for DVD with the appendix (Not working yet). The DVD decoding works for me. What are the limitations of DVD playback in ffdshow? Why is it not officially supported?

[oddball]

I am using one of the later beta drivers for various reasons. CoreAVC works correctly and displays blacks as blacks with the VMR9 box ticked. If I untick it it looks like ffdshow does in YV12 mode. It looks like ffdshow needs it's own 'Fix VMR9 Color Range' option.

[clsid]

It already has: levels

[B.F.]

I'm having some issues with black levels in ffdshow when playing back some HD x264 content. I am using VMR9 windowless in Zoom Player 5 and find the black levels are too bright/washed out. If I use COreAVC 1.2 the blacks are OK. I have had to compensate for the black levels in ffdshow with the gamma levels down to 0.88

ffdshow tryout 896

I had the same problems on my friend computer.
It has a Rageon X1600 card.
I think it's a driver bug.

On my home GF7600GT all OK.

[Syris2k4]

Tried disabling all the HW acceleration?

Have a pretty beefy PC, never found that the nv/amd drivers really sorted anything out quality wise, just seemed to give issues from time to time.

I disabled them from within ATT (ATi Tray Tool, radeon2.ru)... Cant see a noticable difference between VMR9 and Haali's Render... however, on my older nV card I do see the levels are lighter in VMR9...

My config (http://matthew.rucus.net/VAcc.png) Oops, host seems to hate life... will imageshack it later

[Bathrone]

ffdshow does not take over decoding of Windows Media unless you specifically enable it yourself. I recommend to just use the Microsoft codecs.

Thanks, and clsid, let me again thank all the devs for the great software :)

This is what I do for WinXP - every install:

Disable LPCM which is enabled by default and overwrites the MS one
Disable MP3, MP1 and MP2 which is enabled by default and overwrites the MS one
Remove explorer.exe from the dont use ffdshow rule in both audio and video
Enable vorbis tremor if Im using Drevil's build because unlike clsid's build its not enabled by default
Enable FLAC

It seems to me it does actually overwrite LPCM, MP1, MP2 and MP3 atleast by default.

Now that I have upgraded to Vista I honestly dont know if there are other ones now installed by default that overwrite MS ones.

[Bathrone]

Also, on the multithreaded decoder side MPEG4-ASP and AVC arent multithreaded?

[clsid]

Also, on the multithreaded decoder side MPEG4-ASP and AVC arent multithreaded?No they are not.

It seems to me it does actually overwrite LPCM, MP1, MP2 and MP3 atleast by default.Yes, ffdshow takes over decoding of those formats by default, since most people prefer it over the standard Windows decoders. Overwite is the wrong word to use in this context. ffdshow simply has a higher merit that the standard decoders.

[clsid]

Resizing --> Soft resize to 1280x1024Out if curiosity. What should happen?

[Mentar]

Out if curiosity. What should happen?

On most CPUs, it will slow down the playback of normal-fps anime encodes to unusable levels.

[Keepitsimple]

When will we see the next stable tryouts? I see daily development but as a noob I feel very hesistant to upgrade. Right now most stuff works great. The only thing I would appreciate is better documentation what everything does.

Ps. And perhaps a little faster x264 decoding.. :)

[Dark_Angel_PT]

Aside from the CCCP and ffdshow tryouts discussion, i personally think this is a great project, and i give my respect to the developers of the new fork.
However, i believe the new fork suffers some of the problems the old one had.

- Release a stable, regularly updated build.
People like to have something they feel is safe and stable. Something to act as a reference, something to compare everything else against, like "this version is faster/slower than the stable one"
It is a reality that people still keep downloading and using milan's old builds, because they are considered the stable one.

Also, it gives the project a goal, a milestone, something to plan and work for.
Currently, i feel ffdshow just keeps constantly getting new features, and the occasional bugs..., over and over.
There is a need for milestones!! Something to wait for.

When the beta1 for a stable build came up, i was excited.
This project desperately needs a bigger audience, and only a stable, well spread and announced build can provide that.
How long hasn't the ffdshow beta1 been discussed here??
When someone reaches the sourceforge page, finding the usual "CLICK TO DOWNLOAD" button is impossible. The average user expects this, and it is essential for the wide spread of this project, for testing and feedback.

There is a need for an "official build" or recommended one. Of course a stable version could provide this, but further than that...for the testing versions too.
I think it would be a good idea to reach a decision on which combination of settings and compliers provides the best results. Even if not the fastest, a good allaround compatible build for all systems.
Then, other builds for specific instructions or different compliers could be mentioned for experienced users looking for extra speed or something else.

Taking over the old project is also a good thing, although i think in respect to milan, only a warning should be made in the homepage, everything else left as is, and development should continue in ffdshow-tryouts.


-Get the website organized, and USE IT!
I still think it is a waste to use Doom9 so much to discuss ffdshow, with a perfectly good forum and frontpage (the www.ffdshow.info ).
The discussion should be redirected from here to the forum, and the URL for the new fork much more divulged and announced.I don't expect a "SpreadFfdshow" campaign like Firefox....but at least concentrate things in a single place, and spread the word!!

Gathering people in a single place to report and discuss bugs, features and stuff is more more resource-efficient than have them spread over several places.

Having the forum active and with a developer stopping by sometimes, is good too.
I for example made a feature request, and didn't get a reply. A simple "too hard to implement ATM" or "don't expect to see it soon, not a priority" would have made me happy, as i would at least known what to expect.

The resources are there, just use them.

- Start Documentation
Make ffdshow more "for everyone".
Getting a stable version out, or a reference build would be a great step.
Concentrating efforts in one place, and make it know to the public....even better!
But, ffdshow has so many options it is impossible to just know what they all do by playing with them.
To start a wiki of some sorts, to slowly start to explain the workings of ffdshow would be a HUGE step.
An offline help file is good too, and i have heard something about that around here.



These are just suggestions and thoughts of a longtime ffdshow user, very grateful for everyones effort in making this software what it is

[haruhiko_yamagata]

When will we see the next stable tryouts? I see daily development but as a noob I feel very hesistant to upgrade. Right now most stuff works great. The only thing I would appreciate is better documentation what everything does.

Ps. And perhaps a little faster x264 decoding.. :)
It's fair statement. We have a lot to do before beta2. Please wait one or two months at least.

[ExtraEye]

Is X264 encoding speed up to you? I thought that actually depended on the FFMPEG guys...

[haruhiko_yamagata]

Aside from the CCCP and ffdshow tryouts discussion, i personally think this is a great project, and i give my respect to the developers of the new fork.
However, i believe the new fork suffers some of the problems the old one had.

- Release a stable, regularly updated build.
-Get the website organized, and USE IT!
- Start Documentation

As for stable release, we can't do so often. Because ffdshow is a huge project, testing it is very hard. We needed long time to test beta1. I think 4 times a year is the max.
As for take over, in respect to milan, I would like to continue on the tryouts.

As for the website, I agree with you. There should be more active, I hope. Especially for feature requests, it will be buried in, if they are posted here. And sorry for not replying yet.

Documentation is a big target. I completely agree with you. I've just begun but have no time to update. It is greatly appreciated if anyone write a single page for it. If one don't have skill in html, please post a simple text.

[fastplayer]

I see daily development but as a noob I feel very hesistant to upgrade.
There's no need to be "hesitant" using the latest build. Just try it! :)
The devs make - mostly - small incremental changes in each build and if they screw things up, they fix em in a matter of minutes.
When will we see the next stable tryouts?
Good question! Maybe the devs could shed some light on this one.

[_xxl]

I thought that actually depended on the FFMPEG guys.
True.Please ask ffmpeg developers to add multithreaded h264 decoding support.ffmpeg,mplayer and ffdshow can't handle h264 evobs.Tested with AMD X2.

[check]

Just looking at your first post:
ffdshow raw filter doesn't work in ZoomPlayer. (reported by midiboy)
This is a ZP 'feature', called smartplay. It seems to remove any raw processing filters, and also includes a blacklist where you can add evil filters (morgan, antifreeze, et al). Disabling it lets everything work as normal.

re: CCCP doing bug reports and helping code. No can do. We test our own version, which means bug reports would simply become more work for us, and we are mostly a bunch of patch monkeys, not Real Coders, so we can't help with dev :).

re: documentation. I'd be happy to write some of the ffdshow documentation, where does it belong? I've written a fair amount of the megui documentation, so I know what I'm getting in to.

[Hans Ohlo]

True.Please ask ffmpeg developers to add multithreaded h264 decoding support.ffmpeg,mplayer and ffdshow can't handle h264 evobs.Tested with AMD X2.
add paff decoding to the list. i 'dared' to report a bug/crash to the ffmpeg list and ask later if it has been confirmed and if it will be fixed... do not ask me what i got into doing so ;)

[ExtraEye]

"It is greatly appreciated if anyone write a single page for it. If one don't have skill in html, please post a simple text."
Post a simple text where?

[Yong]

can anyone comfirm this?
ffdshow.ax crash if noise filter+SSE2 is enabled.
im using P4 2.4Ghz.

tested with clsdi build, no crash.
i was compiled ffdshow with SSE=1 SSE2=1 flags...

[haruhiko_yamagata]

"It is greatly appreciated if anyone write a single page for it. If one don't have skill in html, please post a simple text."
Post a simple text where?
https://sourceforge.net/tracker/?group_id=173941&atid=867362

[haruhiko_yamagata]

Just looking at your first post:
ffdshow raw filter doesn't work in ZoomPlayer. (reported by midiboy)
This is a ZP 'feature', called smartplay. It seems to remove any raw processing filters, and also includes a blacklist where you can add evil filters (morgan, antifreeze, et al). Disabling it lets everything work as normal.

re: CCCP doing bug reports and helping code. No can do. We test our own version, which means bug reports would simply become more work for us, and we are mostly a bunch of patch monkeys, not Real Coders, so we can't help with dev :).

re: documentation. I'd be happy to write some of the ffdshow documentation, where does it belong? I've written a fair amount of the megui documentation, so I know what I'm getting in to.
You may report bugs as long as it is reproducible in the builds from tryouts. We'll fix them if we can.

Writing a documentation for ffdshow is wellcome, but you know I'm not feeling comfortable with CCCP. Before I can act cozy with you,

Open the source code immediately (Am I the copyright holder for the patches?).
Remove the insulting words in your WEB/forum (I know it's not you who wrote that).

[_xxl]

can anyone comfirm this?
ffdshow.ax crash if noise filter+SSE2 is enabled.
im using P4 2.4Ghz.
tested with clsdi build, no crash.
i was compiled ffdshow with SSE=1 SSE2=1 flags...
MinGW GCC 4.x.x can't compile SSE2 builds.
That's why you couldn't find them anymore.

[ExtraEye]

Mentar, your "example" with the resizing is kind of silly. I mean, the fact that the user can change the way the video is processed in accordance to his tastes/pc's abilities is definitly not a bad thing. "trying to use software in ways they are not INTENDED to be used" - what?
It's fine to argue here and there and it's fine you want to support cccp for not using ffdshow tryouts but is it possible for you to stop talking so rudely? It's totally unnecessary and doesn't help this discussion. Actually this hole debate *flame* doesn't suit this thread at all - may as well continue them in cccp forums instead of wasting the development thread's pages on it.

[check]

Writing a documentation for ffdshow is wellcome, but you know I'm not feeling comfortable with CCCP. Before I can act cozy with you,
You don't need to act cozy with me to tell me a good place to write up documentation :confused: edit: and in any case, I'm not doing this work on behalf of the CCCP. In any case, I've started the thing over on the mewiki (which is turning into a more general AV wiki by the day) here: http://mewiki.project357.com/wiki/Ffdshow_reference
I won't have more time to work on it until sunday, so feel free to add/copy/etc as you see fit. GPL fun: all work posted there is licenced under the GFDL, but I dual licence my first two revisions and release them into the public domain. If you want to start the reference somewhere else, that should be all you need.

[ilpippo80]

@haruhiko
I've tested hw deint in clsid's ffdshow_rev926_20070215_clsid_sse_icl9.exe.
Here's my results using graphedit:

MPV decoder only: ok with VMR7 and VMR9
DScaler only: ok with VMR7 and VMR9
ffdshow(libmpeg2) only:
- with VMR7 hw deint is always off and the image is horizontally stretched (I've never seen this bug before)
- with VMR9 hw deint is always on
MPV+ffdshow(raw): ok with VMR7 and VMR9 (hw deint depends only on MPV setting)
DScaler+ffdshow(raw):
- with VMR7 hw deint depends only on DScaler settings
- with VMR9 hw deint is always off
ffdshow(libmpeg2)+ffdshow(raw): hw deint is always off

[Yong]

MinGW GCC 4.x.x can't compile SSE2 builds.
That's why you couldn't find them anymore.

ok thx for the info;)
btw discovered 2 new bugs:
1. resize filter, eg, if i wana resize my video from 320x240 to 640x480, picture is resized, but it only show 1/4 on the screen.
same as 640x480 resize to 320x240.
2. resize filter crash if using rgb32 out.

just a quick report.:p

[HeadBangeR77]

@ clsid:
Your last 926 icl9 sse build works like charm, thanks. :)

@ haruhiko_yamagata
I cannot reproduce the old bug with auto iDCT & XviD decoding with the above build. Auto works fine for me now.

Btw. Is XviD MMX DCT really always used while encoding?

PS. * Using words of French origin seems to be complicated. ;)

[haruhiko_yamagata]

@haruhiko
I've tested hw deint in clsid's ffdshow_rev926_20070215_clsid_sse_icl9.exe.
Here's my results using graphedit:

MPV decoder only: ok with VMR7 and VMR9
DScaler only: ok with VMR7 and VMR9
ffdshow(libmpeg2) only:
- with VMR7 hw deint is always off and the image is horizontally stretched (I've never seen this bug before)
- with VMR9 hw deint is always on
MPV+ffdshow(raw): ok with VMR7 and VMR9 (hw deint depends only on MPV setting)
DScaler+ffdshow(raw):
- with VMR7 hw deint depends only on DScaler settings
- with VMR9 hw deint is always off
ffdshow(libmpeg2)+ffdshow(raw): hw deint is always off
Oops! True. Please wait.

[ExtraEye]

drevil_xxl: Exactly the point. Nothing is coming out of this - he even made a negative post in which he asks to be ignored and that's probably the best idea too. Let's just move on.

[foxyshadis]

All CCCP discussion moved to http://forum.doom9.org/showthread.php?t=122341

Please keep it there, this thread is discussion for the development of tryouts only.

[cc979]

just be using the resizer in ffdshow rev926, i got it set to 720/576 bicubic and resize allways and 0% border set

thing is - viewing the video on my desktop with normal 100% zoom using overlay/vm7 renderless/vmr9 renderless/haali- normal viewspace, but when in fullscreen there is about an inch of black border on the sides - is that normal ?

[haruhiko_yamagata]

Oops! True. Please wait.
Oops again, I replied too quickly. I tested with wrong settings.
All the case you pointed works OK for me.

[ilpippo80]

That's bad...
Which wrong settings? Maybe I had something wrong too.

[haruhiko_yamagata]

Resize was checked. :stupid:

[clsid]

just be using the resizer in ffdshow rev926, i got it set to 720/576 bicubic and resize allways and 0% border set

thing is - viewing the video on my desktop with normal 100% zoom using overlay/vm7 renderless/vmr9 renderless/haali- normal viewspace, but when in fullscreen there is about an inch of black border on the sides - is that normal ?
That can happen if you have "keep aspect ratio" enabled.

[foxyshadis]

You don't need to act cozy with me to tell me a good place to write up documentation :confused: edit: and in any case, I'm not doing this work on behalf of the CCCP. In any case, I've started the thing over on the mewiki (which is turning into a more general AV wiki by the day) here: http://mewiki.project357.com/wiki/Ffdshow_reference
I won't have more time to work on it until sunday, so feel free to add/copy/etc as you see fit. GPL fun: all work posted there is licenced under the GFDL, but I dual licence my first two revisions and release them into the public domain. If you want to start the reference somewhere else, that should be all you need.

I think that's a great idea and a great place for it, since you already have a well-known wiki set up. Thanks!

[ilpippo80]

ok, my bad, I just rechecked, I didn't understand that I had to disconnect and reconnect ffdshow in the graph after modifying the settings! also the bug with ffdshow(libmpeg2) only with VMR7 went away!
Everything works as it should now, thanks haruhiko! :)

[ilpippo80]

ehm, well, the truth is that there's another small bug, I noticed it during my tests...
with gabest's mpeg2 splitter and vmr9, with ffdshow mpeg2 decoding (but also with dscaler, not with mpv decoder) after seeking the field order seems to be wrong and the hw deinterlaced video becomes "bumpy" (goes 2 frames forward and 1 backward).
I can't tell if it works well with ms mpeg2 splitter because it doesn't allow me to seek at all in graphedit.
I don't know if you can reproduce this and if you have any interest in fixing it...

[haruhiko_yamagata]

ehm, well, the truth is that there's another small bug, I noticed it during my tests...
with gabest's mpeg2 splitter and vmr9, with ffdshow mpeg2 decoding (but also with dscaler, not with mpv decoder) after seeking the field order seems to be wrong and the hw deinterlaced video becomes "bumpy" (goes 2 frames forward and 1 backward).
I can't tell if it works well with ms mpeg2 splitter because it doesn't allow me to seek at all in graphedit.
I don't know if you can reproduce this and if you have any interest in fixing it...

http://support.microsoft.com/kb/924523/en-us
I can't reproduce untill now.
Is this related? If it is not reproducible in VMR7, it's possible.
Which version of direct X do you use?

[ilpippo80]

I only see it when hw deinterlacing is on, and it's the same artifact I see in progDVB (a program to watch DVB tv channels) after seeking.
I have it only with VMR9, not with VMR7.
The directx version reported in dxdiag is 9.0c (4.09.0000.0904).
Maybe it's just related with my hardware and video card drivers...

[ilpippo80]

I just noticed that I don't have it if I enable 'YUV mixing mode' in VMR9 renderer. Unfortunately it seems to be off by default, and I don't think there's a way to enable it globally...
I can enable it in MPC, though. I guess I could just keep using VMR9 in MPC and VMR7 in progDVB...

[ilpippo80]

Some more tests: what I wrote before is true only if ffdshow is used for mpeg2 decoding only.
If I have ffdshow(mpeg2)+ffdshow(uncompressed) or dscaler+ffdshow(uncompressed) I have this problem with VMR7 and VMR9, no matter if I check or uncheck YUV mixing mode and if I reconnect all the filters. No problems at all with MPV decoder also connecting it to ffdshow uncompressed.
Would you like me to upload one of the small mpeg2 captures I'm using for these tests? Maybe it could be of help...

[multiblitz]

Are there any plans that ffdshow can be used to playback hd-dvd ? Now that AnydvdHD is available...that would bump up the pq and give the PC agian the lead...

[_xxl]

FFmpeg, mplayer and ffdshow can't handle h264 HD-DVD evobs.No multithreaded decoder support.Ms VC-1 codec has multitheaded and DXVA, but no support yet for vc1 HD-DVD evobs in ffdshow.

[cc979]

That can happen if you have "keep aspect ratio" enabled.

so is it a small bug then ?

[haruhiko_yamagata]

Some more tests: what I wrote before is true only if ffdshow is used for mpeg2 decoding only.
If I have ffdshow(mpeg2)+ffdshow(uncompressed) or dscaler+ffdshow(uncompressed) I have this problem with VMR7 and VMR9, no matter if I check or uncheck YUV mixing mode and if I reconnect all the filters. No problems at all with MPV decoder also connecting it to ffdshow uncompressed.
Would you like me to upload one of the small mpeg2 captures I'm using for these tests? Maybe it could be of help...
Thank you for testing.
It's not reproducible for me. It may depend on the sample or your hardware. Please upload the sample.

Revision 933 - Directory Listing
Modified Sat Feb 17 13:58:33 2007 UTC (4 minutes, 35 seconds ago) by h_yamagata
Raw video and HW deinterlacing
enable interlaced flag only when ffdshow is inputing VIH2-interlaced format.
Though it was not symptomatic for me, it might not be safe to test dwTypeSpecificFlags without checking the connection type.

[ilpippo80]

@haruhiko
I sent you the sample...

[clsid]

so is it a small bug then ?
Only if there also are black bars at the top and bottom.

But why resize to 720x576 and then go to fullscreen (which does another resize operation)? Is there perhaps a PAL TV connected via TV-out?

[cc979]

Only if there also are black bars at the top and bottom.

But why resize to 720x576 and then go to fullscreen (which does another resize operation)? Is there perhaps a PAL TV connected via TV-out?

i was converting a hd vid and want see what options looked best

it was 'Elephants_Dream_HD.avi'

i will upload a snippet if you like

[cc979]

after testing more testing, with a new clip 'transformers.1080p.divx.avi'

its got no borders and its 1904x784, i have re-encoded it 640x274

resize off on both clips
using ffdshow not in fullscreen normal
using ffdshow in fullscreen normal

resize set to 720x576 on both clips
using ffdshow not in fullscreen only borders added to top and bottom
using ffdshow in fullscreen black borders on the sides

i did not matter which render i used either, its looking more like a bug to me

[clsid]

A sample would be helpful, because I cannot reproduce it.


But I think I have found another bug:
I re-encode a 640x352 sample with VirtualDub and ffdshow VFW and enable image processing. If I use the crop filter to remove a bit from top and bottom, then the actual video will be cropped, but not the frame size. So the output is still 640x352, but with black bars top and bottom, equal to the size I cropped.
And another one:
If I resize to 640x274 with keep aspect ratio on, then the frame size will stay 640x352 instead of resizing and added black bars to the sides. Without aspect ratio correction it works ok.

[cc979]

here is the a sample

http://www.mytempdir.com/1220860

i've just found if the subtitles if ticked even if it has none - it displays correctly

subtitles off - black borders on the sides

[Yong]

i think resize filter bugged since rev850, or may be earlier rev.
and i tried the resizer in zoomplayer, its only works in fullscreen mode,
windowed mode has same problem in mpc.

im using ffdshow_rev926_20070215_clsid_sse_icl9

[clsid]

I still cannot reproduce it :(

[_xxl]

New beta testers are required for ffdshow-tryouts.

[Yong]

Compiling error with r934:
ffdshowRemoteAPI.cpp: In member function 'LRESULT Tremote::remoteWndProc(HWND__*, UINT, WPARAM, LPARAM)':
ffdshowRemoteAPI.cpp:307: error: 'struct IffdshowDecA' has no member named 'grabNow'
make: *** [ffdshow_all.o] Error 1

[Yong]

http://img151.imageshack.us/img151/3088/clipboard01da0.gif
1. rundll32.exe crash if i delete the loaded preset image.
2. im getting "Reconnect failed, please restart video application" if using resize filter with VMR9 windowed/renderless mode.
3. the weird resize problem seems only happended if using overlay mixer/haali video renderer, im not sure though, need more testing :p
4. ffdshow crash with this sample, http://y0ngc6.googlepages.com/KoimomoDEMO300.mp4
, encoded with x264 r622, he aac 64kbps

[haruhiko_yamagata]

1. rundll32.exe crash if i delete the loaded preset image.
2. im getting "Reconnect failed, please restart video application" if using resize filter with VMR9 windowed/renderless mode.
3. the weird resize problem seems only happended if using overlay mixer/haali video renderer, im not sure though, need more testing :p
4. ffdshow crash with this sample, http://y0ngc6.googlepages.com/KoimomoDEMO300.mp4
, encoded with x264 r622, he aac 64kbps

1. OK, loaded preset should be protected.
2. Do you use rev 928 or later? Do you use queue?
4. It's not reproducible for me.

Thanks.

[haruhiko_yamagata]

@ilpippo80
Confirmed. Thank you for the sample.

[ilpippo80]

Good news. I hope it's not too hard to fix...

[Yong]

1. OK, loaded preset should be protected.
2. Do you use rev 928 or later? Do you use queue?
4. It's not reproducible for me.

Thanks.

2. now im using rev 933 gcc4.0.3, yup disable queue and the resize works again,
3. the resize+overlay mixer problem is confirmed?
4. this sample very wierd, its can only play once after reinstall ffdshow, then cant play anymore, disable all simd instruction doesnt help.
5. new crash:rolleyes: , ffdshow crash with msvcrt.dll whenever i click the subtitle filter, even without activated the filter.
6. ffdshow.ax crash if sse2+flickering(noise filter) > 2
http://img218.imageshack.us/img218/1921/clipboard01xk6.gif
7. gcc compiled flt_ffdshow.dll seems to be unuseable.

is it gcc4.0.3 is bugged?

[wyrd]

4. ffdshow crash with this sample, http://y0ngc6.googlepages.com/KoimomoDEMO300.mp4
, encoded with x264 r622, he aac 64kbps
i can reproduce with haali.(do you use ?)
gabest(mpc internal) and nero splitter are fine.
Thanks.

[Yong]

i can reproduce with haali.(do you use ?)
gabest(mpc internal) and nero splitter are fine.
Thanks.

omg:rolleyes: yup im using latest haali media splitter.
and ffdshow only crash if video resolution is 320x240.
thx ;)

[cc979]

I still cannot reproduce it :(

i have just re-tested the clip with ffdshow-rev933 - still happens but if i select 720x576 as the fullscreen res in mpc - it displays without the said side black borders

[_xxl]

ffdshow.ax crash if sse2+flickering(noise filter) > 2

Is ffdshow compiled by MinGW GCC?

[Yong]

Is ffdshow compiled by MinGW GCC?
Yup, mingw gcc4.0.3.

[_xxl]

MinGW GCC 4.x.x can't compile SSE2 version of ffdshow.
It is a compiler bug.

[Yong]

MinGW GCC 4.x.x can't compile SSE2 version of ffdshow.
It is a compiler bug.

ok, got it:p

btw is it possible to compile ffdshow with msvc 2005 express edition or msvc 2003 toolkit?
or with command-line msvc2003/2005 compiler?
------ Rebuild All started: Project: FLT_ffdshow, Configuration: Release Win32 ------
Deleting intermediate and output files for project 'FLT_ffdshow', configuration 'Release|Win32'
Compiling...
ffdebug.c
D:\Program Files\Microsoft Platform SDK\Include\winnt.h(222) : error C2061: syntax error : identifier 'PVOID64'
D:\Program Files\Microsoft Platform SDK\Include\winnt.h(222) : error C2059: syntax error : ';'
D:\Program Files\Microsoft Platform SDK\Include\winnt.h(5940) : error C2061: syntax error : identifier 'PVOID64'
D:\Program Files\Microsoft Platform SDK\Include\winnt.h(5942) : error C2059: syntax error : '}'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5056) : error C2146: syntax error : missing ')' before identifier 'aSegmentArray'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5056) : error C2081: 'FILE_SEGMENT_ELEMENT' : name in formal parameter list illegal
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5056) : error C2061: syntax error : identifier 'aSegmentArray'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5056) : error C2059: syntax error : ';'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5056) : error C3409: empty attribute block is not allowed
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5056) : error C2059: syntax error : ','
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5060) : error C2059: syntax error : ')'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5067) : error C2146: syntax error : missing ')' before identifier 'aSegmentArray'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5067) : error C2081: 'FILE_SEGMENT_ELEMENT' : name in formal parameter list illegal
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5067) : error C2061: syntax error : identifier 'aSegmentArray'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5067) : error C2059: syntax error : ';'
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5067) : error C3409: empty attribute block is not allowed
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5067) : error C2059: syntax error : ','
D:\Program Files\Microsoft Platform SDK\Include\winbase.h(5071) : error C2059: syntax error : ')'
im getting similar error with other dll too, eg ff_x264.dll

[haruhiko_yamagata]

@Yong
3. the resize+overlay mixer problem is confirmed?
What is the problem? If you are talking about on the fly resize setting change, overlaymixer would not accept it.

[Yong]

@Yong

What is the problem? If you are talking about on the fly resize setting change, overlaymixer would not accept it.
Yup, on the fly setting change,
is it a bug?
its weird if overlay mixer wouldnt accept it.
only mpc have this problem.

[_xxl]

4)I have line in 222 of winnt.h fixed from

typedef void * POINTER_64 PVOID64;

--------->

typedef void * POINTER_64;
typedef void * PVOID64;
2)
#ifndef _BASETSD_H_

#define _BASETSD_H_

#if !defined(_MAC) && (defined(_M_MRX000) || defined(_M_AMD64) || defined(_M_IA64)) && (_MSC_VER >= 1100) && !(defined(MIDL_PASS) || defined(RC_INVOKED))

#define POINTER_64 __ptr64 // <-- This code is not called for some reason.

typedef unsigned __int64

-------->

I simply put this definition over the POINTER_64 typedef like this:

#define POINTER_64 __ptr64 // <<---- ORROR !!!

typedef void *PVOID;

typedef void * POINTER_64 PVOID64;

http://forums.microsoft.com/MSDN/ShowPost.aspx?PageIndex=2&SiteID=1&PostID=1196188
http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=63785&SiteID=1

[clsid]

ffdshow crash with this sample, http://y0ngc6.googlepages.com/KoimomoDEMO300.mp4
, encoded with x264 r622, he aac 64kbpsSubmit the sample to Haali and explain that it only crashes with his splitter and not with Gabest/Nero splitters. Then hopefully he can fix it.

[haruhiko_yamagata]

Yup, on the fly setting change,
is it a bug?
its weird if overlay mixer wouldnt accept it.
only mpc have this problem.
No, it's not a bug. Overlaymixer usually(exept RGB out) connects to old video renderer which does not accept the change in format during playback. It is written in Microsoft's document.

This problem has been discussed over and over.
Yes, it's our fault that it is not written in the document.

I think overlaymixer (exept RGB out) should not be tried a reconnect. From this point of view, it's a bug.

[Yong]

http://forums.microsoft.com/MSDN/ShowPost.aspx?PageIndex=2&SiteID=1&PostID=1196188
http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=63785&SiteID=1
ok, thx alot :D
one more problem, the ffdshow.ax compiled by msvc 2005 express seems to be unuseable:confused:

Submit the sample to Haali and explain that it only crashes with his splitter and not with Gabest/Nero splitters. Then hopefully he can fix it.
got it, will do ;)

No, it's not a bug. Overlaymixer usually(exept RGB out) connects to old video renderer which does not accept the change in format during playback. It is written in Microsoft's document.

This problem has been discussed over and over.
Yes, it's our fault that it is not written in the document.

I think overlaymixer (exept RGB out) should not be tried a reconnect. From this point of view, it's a bug.
thx for the info,
btw since when the resize filter cannot change on the fly with overlay mixer?
i remember the resize filter work with very well in older revs.

[_xxl]

one more problem, the ffdshow.ax compiled by msvc 2005 express seems to be unuseable
Can't register .ax?Maybe ffdshow.ax.manifest, microsoft.vc80.crt.manifest and msvcr80.dll are missing?

[Yong]

Can't register .ax?Maybe ffdshow.ax.manifest, microsoft.vc80.crt.manifest and msvcr80.dll are missing?
lol at myselft, i forget to register the ffdshow.ax :p
thx ;)

[vlada]

I use ffdshow-tryouts beta1. I just noticed one problem. I have MOV video with h.263 video compression and g.711 (ULAW) audio compression. I have Mulaw/Alaw in ffdshow enabled. I suppose it should be the same codec, but the sound doesn't play - player says it cannot connect audio pin. Video is O.K. The file plays fine in VLC.

Strange thing happens if I convert (using MP4Cam2AVI) the video to AVI with original compressions. The new AVI file plays fine including sound. Another strange thing is, that MediaInfo reports the audio in MOV to be PCM.

Is this a bug or ULAW in MOV is not supported in ffdshow?

[boban10]

Im sorry if i repeat something.
I have strange things going on.
When i try to play a .MOV video (from my cam) it play fine with audio in media player classic, yet there is one thing that i miss: No control of postprocessing/picture properties. So you will say that it dont support mov files ? Well the strange thing is that if i play the same movie with Wmplayer 10 it play nice with audio, and MY settings from ffdshow works with Wmplayer 10 and not with Mplayer classic, why is that ?
I hope someone understand what is going on. Thanks.

One more thing, i love FFDShow. Thanks.

[Yong]

Submit the sample to Haali and explain that it only crashes with his splitter and not with Gabest/Nero splitters. Then hopefully he can fix it.

wait, i tried this sample with msvc 2005 compiled ffdshow,
the crash is gone :D
gcc4.0.3 is really bugged:rolleyes:

[_xxl]

No control of postprocessing.
H264 PP was removed a long time ago.
gcc4.0.3 is really bugged
MinGW GCC 4.0.x can compile only (MMX and) SSE ffdshow versions.

[vlada]

boban10
I'd bet that you use QuickTime libraries to play your MOV in MPC. Switch MOV playback to DirectShow in MPC's settings.

[Inventive Software]

Any particular reason why clsid's ICL9 builds are larger than his generic ones?

[cc979]

wait, i tried this sample with msvc 2005 compiled ffdshow,
the crash is gone :D
gcc4.0.3 is really bugged:rolleyes:

have you tried gcc 4.0.4 ?

[Leak]

Any particular reason why clsid's ICL9 builds are larger than his generic ones?
Maybe more aggressive code inlining/loop unrolling done by Intel's compiler?

np: Styrofoam - Psychic Friends Network (The Point_Misser)

[cc979]

@Yong: i have just played 'KoimomoDEMO300.mp4' in mpc with haali splitter played with ffdshow svn933 (compiled with gcc 4.0.4) - no crash here

[boban10]

vlada: Thanks very much, im so happy now. Incredible that soo small fix can change things soo much. At first i didnt know where to change that, but happily i found it. THANKS.

[Yong]

@Yong: i have just played 'KoimomoDEMO300.mp4' in mpc with haali splitter played with ffdshow svn933 (compiled with gcc 4.0.4) - no crash here
im downloading ur mingw gcc 4.0.4 :D @4.5kbps:devil:
my mingw folder now is full different build of mingw gcc :p

[Yong]

@Yong: i have just played 'KoimomoDEMO300.mp4' in mpc with haali splitter played with ffdshow svn933 (compiled with gcc 4.0.4) - no crash here
i think there will be no crash if ffdshow was compiled without sse2.
flt_ffdshow.dll still not working if compiled with gcc.:sly:

[JarrettH]

What does yuy mixing mode do?

To make this post more exciting you can download some shots I captured: http://rapidshare.com/files/17145276/2001Shots.rar.html

four shots...

resize + yuy mixing
seesaw dvd + yuy mixing
resize + yuy mixing off
seesaw dvd + yuy mixing off

ffdshow is outputting RGB32...

[foxyshadis]

YUV mixing is mostly a speed boost. If you're outputting rgb, it either does nothing at all, or it'll waste everyone's time by converting back to yuv and reintroducing all the yuv range output issues again. Then all you get for your trouble is the whole YUV->RGB->YUV speed drop.

[JarrettH]

They look much better with it off:devil:

[wyrd]

@Yong,in my case:
with haali splitter(latest)
(ffdshow.ax,libavcodec.dll)
clsid rev926 ms2003, gcc4.0.4 : fine
clsid rev926 icl(sse), gcc4.0.4 : crash
clsid rev895 icl(sse), gcc4.0.3 : crash
clsid rev802 icl, gcc4.0.3 : crash
xxl rev938 ms2003, gcc4.1.2 : fine
xxl rev927 ms2003, gcc4.1.2 : fine
xxl rev899 gcc4.0.4, gcc4.0.4 : fine
xxl rev801 ms2003(sse2),gcc4.0.3 : fine
paehl rev926 ms2003, gcc4.1.1 : fine

but weird... my re-encodefile (http://tirnanog.fate.jp/tmp/koimomo/)(with x264-rev620/622) plays fine in clsid-rev926-icl9 + haali for me.
Regards:).

[Yong]

@Yong,in my case:
with haali splitter(latest)
(ffdshow.ax,libavcodec.dll)
clsid rev926 ms2003, gcc4.0.4 : fine
clsid rev926 icl(sse), gcc4.0.4 : crash
clsid rev895 icl(sse), gcc4.0.3 : crash
clsid rev802 icl, gcc4.0.3 : crash
xxl rev938 ms2003, gcc4.1.2 : fine
xxl rev927 ms2003, gcc4.1.2 : fine
xxl rev899 gcc4.0.4, gcc4.0.4 : fine
xxl rev801 ms2003(sse2),gcc4.0.3 : fine
paehl rev926 ms2003, gcc4.1.1 : fine

but weird... my re-encodefile (http://tirnanog.fate.jp/tmp/koimomo/)(with x264-rev620/622) plays fine in clsid-rev926-icl9 + haali for me.
Regards:).
hmm, what setting did u use to re-encode?
here are my settings:
both pass:
--progress --no-ssim --8x8dct --b-pyramid --mixed-refs --direct auto --weightb -f -3:-6 --bframes 2 --bitrate 250
first pass:
-r 1 -A none -m 1 -p 1 -o nul:
second pass:
-r 5 -p 2 --me umh --no-fast-pskip

ur samples play fines here.
my samples are play fines too as long as i dont use gcc compiled ffdshow.ax:p

msvc 2005 compiled ffdshow.ax(subtitle) still crash for me:mad:
gcc 4.0.4 compiled ff_wmv9.dll still show garbled characters in vdub.

[_xxl]

my samples are play fines too as long as i dont use gcc compiled ffdshow.ax

ffdshow compiled by MinGW GCC 4.0.x SSE2 crashes.SSE build is working ok.
msvc 2005 compiled ffdshow.ax(subtitle) still crash for me
Please provide a sample for that bug.

Bug reporting:
1).ffdshow version, rev, compiler, ansi or unicode
2).ffdshow settings and used filters, ex:queue, resize, sharpening, noise, levels...
3).media player used and settings ex: mpc, media player2...
4).output colorspaces "YUY2", "YV12", "RGB32", "RGB24"...
5).internal/external splitters ex: Haali Media Splitter, Gabest's splitters...
6).video renderers ex:Haali renderer, overlay mixer, VMR9, VMR7...
7).video/audio samples <10MB
8).what you were trying to accomplish, the problem you encountered...anything else that could be useful in reproducing that bug.

Do you really need MinGW GCC SSE2 builds?
Maybe a generic build is working with SSE2?

SSE2 colorize
SSE2 hue and saturation
SSE2 bitmap blending and soften
SSE2 luminance offset and gain
SSE2 warpsharp bump
SSE2 gradual denoise
SSE2 xsharpen
SSE2 unsharp
SSE2 mplayer noise
...

[_xxl]

gcc 4.0.4 compiled ff_wmv9.dll still show garbled characters in vdub.
Is ffdshow.ax compiled with unicode support?
Is ff_wmv9.dll compiled with unicode?
If you have an ansi ffdshow and an unicode ff_wmv9 then it is wrong.

[Yong]

ffdshow compiled by MinGW GCC 4.0.x SSE2 crashes.SSE build is working ok.

Please provide a sample for that bug.

Do you really need MinGW GCC SSE2 builds?
Maybe a generic build is working with SSE2?
Nope, im just playing with the gcc compiler settings,
so far im found out libmpeg2, noise filter and ffdshow.ax(see my previous post)
have problems with gcc compiled SSE2 code,

about the subtitle,
mpc crash with msvcrt80.dll(msvcrt.dll if compiled by gcc) if i configure the subtitle settings during video playback, even there is no subtitle loaded.
zoomplayer is better, no crash, but the ffdshow decoder properties "disappear" without notice.

Is ffdshow.ax compiled with unicode support?
Is ff_wmv9.dll compiled with unicode?
If you have an ansi ffdshow and an unicode ff_wmv9 then it is wrong.
didnt know that, thx for the info ;)

minor cosmetics bug:
http://img293.imageshack.us/img293/2692/clipboard01kr9.gif

[_xxl]

mpc crash with msvcrt80.dll(msvcrt.dll if compiled by gcc) if i configure the subtitle settings during video playback, even there is no subtitle loaded.

MPC is compiled by MSVC2005.Latest?
ffdshow is crashing with msvcrt80.dll(msvcrt.dll if compiled by gcc)?

[Yong]

MPC is compiled by MSVC2005.Latest?
ffdshow is crashing with msvcrt80.dll(msvcrt.dll if compiled by gcc)?
dont know if its latest :p
msvc++ 2005 version 8.0.50727.42 no service pack 1 installed.

[wyrd]

what setting did u use to re-encode?
oh..sorry,my shortage of information.
CQ:
--qp 26 --ref 3 --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 6 --trellis 1
--analyse all --8x8dct --threads auto --thread-input --progress --no-psnr --no-ssim
ABR:
--bitrate 700 --ref 3 --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 6 --trellis 1
--analyse all --8x8dct --vbv-maxrate 25000 --threads auto --thread-input --progress --no-psnr --no-ssim

also,I've used celtic_druid's MP4BOX-20070128(with YAMB) for demux/mux.
Regards:)

[Yong]

Generic build still same :sly:
noise filter still crash :D
my sample crash...
subtitle setting also crash...
compiled by gcc 4.0.4
im gonna try ffdshow with msvc 2003 :p

[Yong]

oh..sorry,my shortage of information.
CQ:
--qp 26 --ref 3 --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 6 --trellis 1
--analyse all --8x8dct --threads auto --thread-input --progress --no-psnr --no-ssim
ABR:
--bitrate 700 --ref 3 --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 6 --trellis 1
--analyse all --8x8dct --vbv-maxrate 25000 --threads auto --thread-input --progress --no-psnr --no-ssim

also,I've used celtic_druid's MP4BOX-20070128(with YAMB) for demux/mux.
Regards:)
ive tired settings above, no crash,
my 2 pass setting still crash ffdshow.ax + if the video resolution is 320x240:sly:
VMR9+resize++queue is finally works, thx :D

[_xxl]

Generic build still same
noise filter still crash
my sample crash...
subtitle setting also crash...
compiled by gcc 4.0.4
im gonna try ffdshow with msvc 2003
Did you removed old version of ffdshow before installing this one?
And please don't use anymore GCC to compile ffdshow.This is not the best way.Only libavcodec and libmplayer should be compiled by GCC.
And please provide samples for every crash so we could help you.

[Yong]

Did you removed old version of ffdshow before installing this one?
And please don't use anymore GCC to compile ffdshow.This is not the best way.Only libavcodec and libmplayer should be compiled by GCC.
And please provide samples for every crash so we could help you.
the crash happended every single of my video files :p
i was using cc979 compiled gcc for compiling ffdshow.
from now i will using msvc 2005 to compile ffdshow.

are you guys using msvc 2003 full edition?

[haruhiko_yamagata]

the crash happended every single of my video files :p
i was using cc979 compiled gcc for compiling ffdshow.
from now i will using msvc 2005 to compile ffdshow.

are you guys using msvc 2003 full edition?
At least professinal edition is required. Standard edition can build a debug version.

[cc979]

using haali splitter

just tested various videos using dscaler plugins in ffdshow - no problems, svn938 compiled everything with gcc4.0.4

in dscaler4 tried playing using ffdshow, resize + subtitles - no problems

[Yong]

At least professinal edition is required. Standard edition can build a debug version.
im using msvc 2003 toolkit + codeblocks ...
this "toolkits" is not very usefull :p

using haali splitter

just tested various videos using dscaler plugins in ffdshow - no problems, svn938 compiled everything with gcc4.0.4

in dscaler4 tried playing using ffdshow, resize + subtitles - no problems
can i test ur ffdshow? :D

and make sure u are using gcc compiled flt_ffdshow.dll :p
not the one that come from ffdshow installer.

[_xxl]

FFdshow should be compiled:
MSVC71:
ffdshow.ax
ff_libdts.dll
ff_libmad.dll
ff_libfaad2.dll
ff_liba52.dll
ff_realaac.dll
libmpeg2_ff.dll
ff_samplerate.dll
ff_theora.dll
ff_tremor.dll
ff_x264.dll
ff_unrar.dll
ICL 9.1:
kerneldeint.dll
TomsMoComp_ff.dll
MinGW GCC 4.1.2:
libavcodec.dll
libmplayer.dll
MinGW GCC should not compile SSE2 versions.

[cc979]

@Yong: my builds of ffdshow and flt_ffdshow.dll

svn938
http://www.mytempdir.com/1223252
http://www.mytempdir.com/1223259

svn944
http://www.mytempdir.com/1223264
http://www.mytempdir.com/1223268

'and make sure u are using gcc compiled flt_ffdshow.dll
not the one that come from ffdshow installer.' what do you mean, make makes all that src?

[Yong]

@Yong: my builds of ffdshow and flt_ffdshow.dll

svn938
http://www.mytempdir.com/1223252
http://www.mytempdir.com/1223259

svn944
http://www.mytempdir.com/1223264
http://www.mytempdir.com/1223268
Thx for the build ;)
'and make sure u are using gcc compiled flt_ffdshow.dll
not the one that come from ffdshow installer.' what do you mean, make makes all that src?
please correct me if im wrong ;
regardless what compiler u use to compile ffdshow,
the installer always install these:
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/bin/distrib/InnoSetup/msvc71/
instead from ffdshow/bin folder.

[_xxl]

Thx for the build ;)
please correct me if im wrong ;
regardless what compiler u use to compile ffdshow,
the installer always install these:
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/bin/distrib/InnoSetup/msvc71/
instead from ffdshow/bin folder.
True.
MinGW GCC should only compile libavcodec & libmplayer.
Intel Compiler ff_kerneldeint.dll, TomsMoComp_ff.dll and the rest should be compiled by MSVC71(80).

[_xxl]

@cc979
Can you compile latest GCC 4.1.2?

[cc979]

@cc979
Can you compile latest GCC 4.1.2?

its now gcc 4.1.3, will upload soon

[ianken]

While I'm sure most of you don't care about Vista compatibility:

With H.264 and non-square pixel content the video is rendered wider than it should be using the current B1 build. I'm using WMP but see the same thing in MPC.

I've not had a chance to investigate further. But thought I'd toss this out in case someone else has seen it and wishes to comment.

[Bathrone]

ianken please supply a sample - Im not seeing it on 32bit vista but maybe my samples arent the same.

This begs the question when are we going to see 64bit vista builds for this great software?

[sillKotscha]

what about adding a 4th mp3 decoder to ffdshow?... as lame v3.98a11 was released with a working DS decoder would it make any sense to incorporate it into ffdshow? ... I guess 85% of all current mp3 files (if not more) are encoded using lame thus it stands to reason to let lame decode those streams...

what do you think :)

[Yong]

correct me if im wrong,
Its a DS Encoder.
btw, lame is using mpglib as thier main decoder,
which is same as ffdshow mp3lib decoder.

[tekNerd]

Why ffdshow can't decode Lagarith? or it can and I am missing something?

[foxyshadis]

Because there's no Lagarith decoder in ffmpeg, and so far you're the first person to my knowledge to express an interest in it. Since it's an intermediate format, it's generally a given that anyone who has such a file created and thus can decode it.

[TheShadowRunner]

Using CLSID generic builds, i noticed that after build rev906, there is a new 2 seconds delay when opening videos which use ffdshow. 906 works fine. the 2 builds after show this effect.
I hope it can be fixed.
Thanks.

TSR

[sillKotscha]

correct me if im wrong,
Its a DS Encoder.
btw, lame is using mpglib as thier main decoder,
which is same as ffdshow mp3lib decoder.

at rarewares it reads like... "LAME 3.98 alpha 11 DirectShow Filter: ICL9.1 compile. " and I thought the *.ax file is the decoder/ not encoder...

about dependencies between mpglib and mp3lib used by ffdshow I don't know - although relying on the same library I thought that the Lame folks may included some *special* whatever into their decoder...

that's why I'm asking...

[Yong]

at rarewares it reads like... "LAME 3.98 alpha 11 DirectShow Filter: ICL9.1 compile. " and I thought the *.ax file is the decoder/ not encoder...
Yeah the title and the *.ax are confusing ppl :p
about dependencies between mpglib and mp3lib used by ffdshow I don't know - although relying on the same library I thought that the Lame folks may included some *special* whatever into their decoder...

that's why I'm asking...
lame is using "lite" version of mpglib, which many mmx/3dnow codes has removed,
while ffdshow still using "full" version of mpglib(renamed to mp3lib).

[haruhiko_yamagata]

Using CLSID generic builds, i noticed that after build rev906, there is a new 2 seconds delay when opening videos which use ffdshow. 906 works fine. the 2 builds after show this effect.
I hope it can be fixed.
Thanks.

TSR
I can't reproduce this.
Please tell me about the settings (queue, application, video renderer, etc).
Does it depend on sample (especially anamorphic AR).
Do you use resize (and the settings)?

[ilpippo80]

@haruiko
I tried clsid's rev945, and now the problem of hardware deinterlacing and seeking with the mpeg2 video has almost disappeared: I still have it though using ffdshow for decoding and VMR9.
I see it using both libmpeg2 or libavcodec, but with lbavcodec it gets fixed in a few seconds, not with libmpeg2.
Do you think it's a libmpeg2 bug?
Thanks!

[TheShadowRunner]

I can't reproduce this.
Please tell me about the settings (queue, application, video renderer, etc).
Does it depend on sample (especially anamorphic AR).
Do you use resize (and the settings)?

Hello haruhiko,
it happens with EVERY .avi files i tried that contain a stream decoded by ffdshow like xvid, divx, dx50 etc..
The application is Zoom Player, video renderer is VMR9.
It doesn't depend on the sample.
And i don't use resize.
(i only use Subtitles)
Regards,

TSR

edit: i tried disabling "Subtitles" on build rev945, the problem is still there.

[TheShadowRunner]

Re- The issue was fixed in new build rev 952, great! ;)
Later,

TSR

[cc979]

i've started getting these warning with nsis with ffdshow svn952

2 warnings:
SetCompressor ignored due to previous call with the /FINAL switch (D:\msys\1.0\home\User\svn\ffdshow-tryout\ffdshow-tryout\bin\distrib\ffdshow.nsi:100)
install function "GetParent" not referenced - zeroing code (0-21) out

not sure what the second is?

[haruhiko_yamagata]

@haruiko
I tried clsid's rev945, and now the problem of hardware deinterlacing and seeking with the mpeg2 video has almost disappeared: I still have it though using ffdshow for decoding and VMR9.
I see it using both libmpeg2 or libavcodec, but with lbavcodec it gets fixed in a few seconds, not with libmpeg2.
Do you think it's a libmpeg2 bug?
Thanks!
I can reproduce this. It's the same with libmpeg2(ffdshow), libavcodec(ffdshow) and libmpeg2(MPC). libavcodec(ffdshow) sometimes screws up and fixes the field order issue, but it seems to be accidental. Not likely a bug of ffdshow/MPC. Maybe only Microsoft can fix it. Thanks.

[Yong]

Bug report:
1.
http://img120.imageshack.us/img120/4108/clipboard01os7.gif
Loop filter options for mpeg4-sp/asp? ;)
2. another way to make vdub crash:D
eg, go to "select video compression" > ffvfw > input > "configure"(ffdshow image processing) > close the "select video compression" window > close vdub > Oops, Vdub mpeg2 has crashed...

[Bathrone]

Lets talk about flac!!

1. Why it is not enabled by default

2. Why when I have installed the flac install package from sourceforge, and enabled it in ffdshow wmp11 on Vista still wont play it.

Much love to all ffdshow contributors.

[_xxl]

2. another way to make vdub crash
eg, go to "select video compression" > ffvfw > input > "configure"(ffdshow image processing) > close the "select video compression" window > close vdub > Oops, Vdub mpeg2 has crashed...
No crash here and please test more...
1. Why it is not enabled by default
Because you need a .flac splitter.
2. Why when I have installed the flac install package from sourceforge, and enabled it in ffdshow wmp11 on Vista still wont play it.
Why aren't you using winamp?Why do you need to play .flac in WMP11?

[Inventive Software]

No crash here and please test more...

Because you need a .flac splitter.

Any chance that splitters for, say, FLAC and WavPack can be integrated into ffdshow, either within the installer or within the code?

[clsid]

You'll have to manually install splitters or use a codec pack. ffdshow is for decoding only and I don't think that will ever change.

[foxyshadis]

Any chance that splitters for, say, FLAC and WavPack can be integrated into ffdshow, either within the installer or within the code?

Haruhiko has proposed an ffformat (?) project to go with ffdshow, based on lavf, but it's just an idea at this point. You can always create a little mini-pack with gabest's splitters and a batch file.

In the meantime, you could ask haali, it seems every other version of his splitter supports a new format. :p

[haruhiko_yamagata]

Haruhiko has proposed an ffformat (?) project to go with ffdshow, based on lavf, but it's just an idea at this point. You can always create a little mini-pack with gabest's splitters and a batch file.

In the meantime, you could ask haali, it seems every other version of his splitter supports a new format. :p
I have not propossed the idea.

[Inventive Software]

I have not propossed the idea.

Then who the hell did?! :D

I did suggest something similar a while ago, to maybe have libavformat integrated into ffdshow so that it's a more complete solution, with specialised splitters for WavPack and FLAC for instance.

[ExtraEye]

hmmm will a splitters' filter be difficult to make or is it an easy option?

[clsid]

Difficult or not, it would take a LOT of time to make, even if code from ffmpeg is used. I don't think there is currently anyone here willing to take on that task.

Furthermore, there are already good DirectShow splitters available for all popular formats, including FLAC and WavPack.

[Yong]

No crash here and please test more...

ok im sorry :D
i now simplify it,
in vdub, go to "select video compression" > "configure" ffvfw > close the "select video compression" window, dont close ffvfw encoder > close vdub > Oops, Vdub mpeg2 has crashed...

[_xxl]

No crash.
I can't close vdub if ffvfw encoder is active.

[Inventive Software]

I can confirm this happens. What you need to do drevil_xxl, is when the ffvfw encoder config box is active, click the "Select video compression" window in VirtualDub, click OK, then close VirtualDub, and said crash occurs.

This is the full crash log dump from VirtualDub 1.7.1:

VirtualDub crash report -- build 27021 (release)
--------------------------------------

Disassembly:
004a5e40: f8 clc
004a5e41: 0100 add [eax], eax
004a5e43: 00e8 add al, ch
004a5e45: 17 pop ss
004a5e46: 0bf7 or esi, edi
004a5e48: ff8b45e883b8 dec dword ptr [ebx-477c17bb]
004a5e4e: 50 push eax
004a5e4f: 0100 add [eax], eax
004a5e51: 0000 add [eax], al
004a5e53: 7429 jz 004a5e7e (VDProjectUI::Detach+e5)
004a5e55: 6a00 push 00h
004a5e57: 8b4de8 mov ecx, [ebp-18h]
004a5e5a: 81c14c010000 add ecx, 0000014c
004a5e60: e8857a0000 call 004ad8ea (?$vdrefptr::(special))
004a5e65: 8b45e8 mov eax, [ebp-18h]
004a5e68: ffb050010000 push dword ptr [eax+150]
004a5e6e: ff1594465a00 call dword ptr [005a4694]
004a5e74: 8b45e8 mov eax, [ebp-18h]
004a5e77: 83a05001000000 and dword ptr [eax+150], 00h
004a5e7e: 8b45e8 mov eax, [ebp-18h]
004a5e81: 83b83001000000 cmp dword ptr [eax+130], 00h
004a5e88: 7419 jz 004a5ea3 (VDProjectUI::Detach+10a)
004a5e8a: 8b45e8 mov eax, [ebp-18h]
004a5e8d: ffb030010000 push dword ptr [eax+130]
004a5e93: ff1594465a00 call dword ptr [005a4694]
004a5e99: 8b45e8 mov eax, [ebp-18h]
004a5e9c: 83a03001000000 and dword ptr [eax+130], 00h
004a5ea3: 8b45e8 mov eax, [ebp-18h]
004a5ea6: 83b83c01000000 cmp dword ptr [eax+13c], 00h
004a5ead: 744e jz 004a5efd (VDProjectUI::Detach+164)
004a5eaf: 8b45e8 mov eax, [ebp-18h]
004a5eb2: 83b83401000000 cmp dword ptr [eax+134], 00h
004a5eb9: 741f jz 004a5eda (VDProjectUI::Detach+141)
004a5ebb: 8b45e8 mov eax, [ebp-18h]
004a5ebe: ffb034010000 push dword ptr [eax+134]
004a5ec4: e867dd0a00 call 00553c30 (VDGetILogWindowControl)
004a5ec9: 59 pop ecx
004a5eca: 8945e0 mov [ebp-20h], eax
004a5ecd: 6a00 push 00h
004a5ecf: 8b45e0 mov eax, [ebp-20h]
004a5ed2: 8b00 mov eax, [eax] <-- FAULT
004a5ed4: 8b4de0 mov ecx, [ebp-20h]
004a5ed7: ff5010 call dword ptr [eax+10h]
004a5eda: 8b45e8 mov eax, [ebp-18h]
004a5edd: ffb03c010000 push dword ptr [eax+13c]
004a5ee3: ff1594465a00 call dword ptr [005a4694]
004a5ee9: 8b45e8 mov eax, [ebp-18h]
004a5eec: 83a03c01000000 and dword ptr [eax+13c], 00h
004a5ef3: 8b45e8 mov eax, [ebp-18h]
004a5ef6: 83a04401000000 and dword ptr [eax+144], 00h
004a5efd: 8b45e8 mov eax, [ebp-18h]
004a5f00: 83b84001000000 cmp dword ptr [eax+140], 00h
004a5f07: 744e jz 004a5f57 (VDProjectUI::Detach+1be)
004a5f09: 8b45e8 mov eax, [ebp-18h]
004a5f0c: 83b83801000000 cmp dword ptr [eax+138], 00h
004a5f13: 741f jz 004a5f34 (VDProjectUI::Detach+19b)
004a5f15: 8b45e8 mov eax, [ebp-18h]
004a5f18: ffb038010000 push dword ptr [eax+138]
004a5f1e: e80ddd0a00 call 00553c30 (VDGetILogWindowControl)
004a5f23: 59 pop ecx
004a5f24: 8945dc mov [ebp-24h], eax
004a5f27: 6a00 push 00h
004a5f29: 8b45dc mov eax, [ebp-24h]
004a5f2c: 8b00 mov eax, [eax]
004a5f2e: 8b4ddc mov ecx, [ebp-24h]
004a5f31: ff5010 call dword ptr [eax+10h]
004a5f34: 8b45e8 mov eax, [ebp-18h]
004a5f37: ffb040010000 push dword ptr [eax+140]
004a5f3d: ff15 call dword ptr [ebp]
004a5f3f: 94 xchg eax, esp

Built on KOS-MOS on Tue Feb 06 21:24:09 2007 using compiler version 1400

Windows 5.1 (Windows XP build 2600) [Service Pack 2]

EAX = 00000000
EBX = 00000000
ECX = 0011026c
EDX = 7c90eb94
EBP = 0012d278
ESI = 004b7e9c
EDI = 00000000
ESP = 0012d250
EIP = 004a5ed2
EFLAGS = 00010246
FPUCW = ffff027f
FPUTW = ffffffff

Crash reason: Access Violation

Crash context:
An out-of-bounds memory access (access violation) occurred in module 'VirtualDub'...

...reading address 00000000.

Pointer dumps:

EDX 7c90eb90: 90909090 24a48dc3 00000000 0024648d 90909090 24548d90 c32ecd08 9cec8b55
ESI 004b7e98: c3c9c032 6aec8b55 e56168ff a1640059 00000000 25896450 00000000 8170ec83
ESP 0012d250: 00000000 001102a0 00000000 00b5ba88 00c0bd20 0012d334 77d70467 0012d288
0012d270: 00b5ba88 001102a0 0012d2b4 004b829a 0012d2b8 00c0be28 00b5ac28 00000000
0012d290: 77d49491 00b5ac30 00c0be28 00000000 0012d294 00c0be28 0012d338 0059e57b
0012d2b0: 00000000 0012d344 004b7fb6 0011029c 00000001 00000001 5ad7153d 00323930
EBP 0012d278: 0012d2b4 004b829a 0012d2b8 00c0be28 00b5ac28 00000000 77d49491 00b5ac30
0012d298: 00c0be28 00000000 0012d294 00c0be28 0012d338 0059e57b 00000000 0012d344
0012d2b8: 004b7fb6 0011029c 00000001 00000001 5ad7153d 00323930 0012d2e0 77d4b96b
0012d2d8: 77d4b933 00000082 0011029c 004b7e9c 0012d2d8 00324000 00000000 00324028

Thread call stack:
004a5ed2: VDProjectUI::Detach()
004b829a: VDUIFrame::DetachNow()
77d49491: USER32!GetWindowLongA [77d40000+945d+34]
004b7fb6: VDUIFrame::StaticWndProc()
5ad7153d: uxtheme!0000153d
77d4b96b: USER32!IsWindow [77d40000+b933+38]
5ad73a5a: uxtheme!DrawThemeText [5ad70000+3031+a29]
5ad73995: uxtheme!DrawThemeText [5ad70000+3031+964]
5ad73a36: uxtheme!DrawThemeText [5ad70000+3031+a05]
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d4d05b: USER32!EnumDisplayMonitors [77d40000+cf63+f8]
77d4b4c0: USER32!DefWindowProcW [77d40000+b33c+184]
77d4dabd: USER32!IsWindowInDestroy [77d40000+da74+49]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d494be: USER32!GetWindowLongA [77d40000+945d+61]
7c832f29: kernel32!LocalUnlock [7c800000+32eb1+78]
77d62fcb: USER32!DrawFrame [77d40000+2299b+630]
77d4b3f9: USER32!DefWindowProcW [77d40000+b33c+bd]
77d4b393: USER32!DefWindowProcW [77d40000+b33c+57]
004b7bca: VDUIFrame::DefProc()
004a9480: _catch$?MainWndProc@VDProjectUI@@IAEJIIJ@Z$0()
77d4882a: USER32!GetDC [77d40000+86c7+163]
77d4b89b: USER32!GetParent [77d40000+b72f+16c]
74724636: MSCTF!DllGetClassObject [74720000+438b+2ab]
77d4eb3e: USER32!CallNextHookEx [77d40000+eb03+3b]
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
004a8f5b: VDProjectUI::WndProc()
004b8145: VDUIFrame::StaticWndProc()
747309a9: MSCTF!TF_UninitSystem [74720000+10469+540]
77f16c25: GDI32!GdiDrawStream [77f10000+6bd1+54]
77d4ea9e: USER32!EnableMenuItem [77d40000+ea2f+6f]
77d4eaad: USER32!EnableMenuItem [77d40000+ea2f+7e]
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
77d4b4c0: USER32!DefWindowProcW [77d40000+b33c+184]
77d4b50c: USER32!DefWindowProcW [77d40000+b33c+1d0]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d494be: USER32!GetWindowLongA [77d40000+945d+61]
77d484fc: USER32!000084fc
77d4ba0e: USER32!IsWindow [77d40000+b933+db]
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d4b3f9: USER32!DefWindowProcW [77d40000+b33c+bd]
5ad73c20: uxtheme!DrawThemeText [5ad70000+3031+bef]
5ad8e300: uxtheme!GetThemeTextMetrics [5ad70000+1de50+4b0]
5ad71ac7: uxtheme!00001ac7
5ad71b3d: uxtheme!00001b3d
77d4bb15: USER32!GetPropW [77d40000+badb+3a]
004b7bca: VDUIFrame::DefProc()
004a9480: _catch$?MainWndProc@VDProjectUI@@IAEJIIJ@Z$0()
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
004a8f5b: VDProjectUI::WndProc()
004b8145: VDUIFrame::StaticWndProc()
5ad74301: uxtheme!GetThemeBackgroundContentRect [5ad70000+3f9f+362]
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
77d4b89b: USER32!GetParent [77d40000+b72f+16c]
77d4b903: USER32!SendMessageW [77d40000+b8ba+49]
5ad9881f: uxtheme!GetThemeTextMetrics [5ad70000+1de50+a9cf]
5ad71ac7: uxtheme!00001ac7
5ad71b3d: uxtheme!00001b3d
77d4bb15: USER32!GetPropW [77d40000+badb+3a]
004b7bca: VDUIFrame::DefProc()
004a9480: _catch$?MainWndProc@VDProjectUI@@IAEJIIJ@Z$0()
77d4eaf2: USER32!EnableMenuItem [77d40000+ea2f+c3]
77d4ea9e: USER32!EnableMenuItem [77d40000+ea2f+6f]
77d4eaad: USER32!EnableMenuItem [77d40000+ea2f+7e]
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
004a8f5b: VDProjectUI::WndProc()
004b8145: VDUIFrame::StaticWndProc()
77d4d8d3: USER32!GetAncestor [77d40000+d8c7+c]
77d4eb3e: USER32!CallNextHookEx [77d40000+eb03+3b]
0048209d: VDModelessDialogHookW32()
77d618f4: USER32!UnhookWinEvent [77d40000+2187d+77]
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d489cd: USER32!GetWindowLongW [77d40000+88a6+127]
77d485a4: USER32!000085a4
77d48a10: USER32!DispatchMessageW [77d40000+8a01+f]
77d5e2b9: USER32!CallMsgFilterW [77d40000+1e0a6+213]
77d561c6: USER32!DrawStateW [77d40000+15fd4+1f2]
77d56208: USER32!DialogBoxIndirectParamAorW [77d40000+161d2+36]
77d6205e: USER32!DialogBoxIndirectParamW [77d40000+22043+1b]
01a62d79: ffdshow!configureEnc [01980000+d6a80+c2f9]
7752b1f7: ole32!CreateGenericComposite [774e0000+4a50b+cec]
7c910992: ntdll!RtlFreeAnsiString [7c900000+10976+1c]
7c85a477: kernel32!OutputDebugStringW [7c800000+5a42d+4a]
01985ce9: ffdshow!00005ce9
7752b767: ole32!CreateGenericComposite [774e0000+4a50b+125c]
774ffa51: ole32!OleInitialize [774e0000+1f6da+377]
7c36b864: MSVCR71!fputs [7c340000+2b71f+145]
7c36c958: MSVCR71!wscanf [7c340000+2c8ee+6a]
7c36c9b7: MSVCR71!wscanf [7c340000+2c8ee+c9]
7c36d180: MSVCR71!wscanf [7c340000+2c8ee+892]
7c36d1d1: MSVCR71!wscanf [7c340000+2c8ee+8e3]

-- End of report

[Yong]

@Inventive Software:
thx :D
btw, where is the loop filter for libavcodec h263+ encoder? :p
ive tested the loopfilter in mencoder, and its work for me,

[clsid]

I can confirm the VDub crash. That is not a ffdshow bug, but a VDub bug.

[cc979]

i looking into this resize problem i got, so did some in WiMP/MPC both same

test file:

http://www.mytempdir.com/1220860

results:

1. resize(bicubic) to 720/480 no subs - no problems

http://img410.imageshack.us/img410/627/1clip1fs720x480qo9.th.png (http://img410.imageshack.us/my.php?image=1clip1fs720x480qo9.png)

2. resize(bicubic) to 720/576 no subs - side borders added, problem

http://img515.imageshack.us/img515/3070/2clip1fs720x576as9.th.png (http://img515.imageshack.us/my.php?image=2clip1fs720x576as9.png)

3. resize(bicubic) to 720/576 with subs letterbox (4/3) - no problems

http://img515.imageshack.us/img515/720/3clip1fs720x576subslb4xtn2.th.png (http://img515.imageshack.us/my.php?image=3clip1fs720x576subslb4xtn2.png)

4. resize(bicubic) to 720/576 with subs letterbox (5/4) - side borders added, problem

http://img92.imageshack.us/img92/4748/4clip1fs720x576subslb5xds4.th.png (http://img92.imageshack.us/my.php?image=4clip1fs720x576subslb5xds4.png)

5. resize(bicubic) to 720/576 with subs letterbox (16/9) - no problems

http://img158.imageshack.us/img158/2159/5clip1fs720x576subslb16wa4.th.png (http://img158.imageshack.us/my.php?image=5clip1fs720x576subslb16wa4.png)


so i'm thinking the problem could linked with subs letterbox - aspect ratio

anybody get this problem?

cheers

[clsid]

I cannot reproduce it. Perhaps because my fullscreen resolution is different (1280x1024).

[cc979]

I cannot reproduce it. Perhaps because my fullscreen resolution is different (1280x1024).

i'm running 1024x768, but will try 1280x1024

cheers

edit: does not happen using 1280x1024 desktop, odd??

[LoRd_MuldeR]

Furthermore, there are already good DirectShow splitters available for all popular formats, including FLAC and WavPack.

What about adding a few splitters to the ffdshow installer?

Of course only the most important ones and only the ones that are not supported by Haali's splitter.
I don't want this to be a "Codec Package", but a few (small) splitters won't hurt...

[cc979]

@clsid:
just been doing some calculations, the clip in question is 1904x784(ar 2.42) and my desktop is 1024x768(ar 1.25) so in ffdshow using 720x576(ar 1.25) but want ffdshow to keep the original ar of 2.42 the output should be 720x296

whats happening is
ffdshow is using 1904/784=ar.242 to make the x=more than 720,y=576 (is that right??)

the calculation i used

src x1=1904
src y1=784

dest x2=720
dest y2=(y1/x1)*x2

and just noticed in ffdshow - can use 720x0(auto) which does that calculation, me fool - doh

so ffdshow was correct all the time

maybe the auto y size thing could be displayed on resize tab somewhere in ffdshow or 720xauto in a preset - along with 720x576, as i only noticed moving the mouse of the y box and tooltip came up

edit: could it be possible to use 720xauto then add vertical borders to make it up 720x576 ??


cheers, anyway

[Inventive Software]

@clsid: It's NOT a VDub bug, but an ffdshow dialog bug, since it doesn't happen with DivX, Xvid, or any other VFW encoder I know of.

[_xxl]

VirtualDub 1.7.1 has this issue.Vb 1.6.11 can't be closed if ffdshow encoder is active.Xvid 1.1.2 seems to be working ok.

[Inventive Software]

Same issue with 1.6.17. Crash log:

VirtualDub crash report -- build 24469 (release)
--------------------------------------

Disassembly:
00499060: 5b pop ebx
00499061: c20400 ret 0004
00499064: 8b06 mov eax, [esi]
00499066: 8bce mov ecx, esi
00499068: ff5004 call dword ptr [eax+04h]
0049906b: 5f pop edi
0049906c: 5e pop esi
0049906d: 5d pop ebp
0049906e: 32c0 xor al, al
00499070: 5b pop ebx
00499071: c20400 ret 0004
00499074: 90 nop
00499075: 90 nop
00499076: 90 nop
00499077: 90 nop
00499078: 90 nop
00499079: 90 nop
0049907a: 90 nop
0049907b: 90 nop
0049907c: 90 nop
0049907d: 90 nop
0049907e: 90 nop
0049907f: 90 nop
00499080: 56 push esi
00499081: 57 push edi
00499082: 8bf1 mov esi, ecx
00499084: 8b4608 mov eax, [esi+08h]
00499087: 33ff xor edi, edi
00499089: 57 push edi
0049908a: 50 push eax
0049908b: ff1534645300 call dword ptr [00536434]
00499091: 57 push edi
00499092: 8bce mov ecx, esi
00499094: e857abffff call 00493bf0 (VDProject::SetUICallback)
00499099: 8b8e2c010000 mov ecx, [esi+12c]
0049909f: 3bcf cmp ecx, edi
004990a1: 7405 jz 004990a8 (VDProjectUI::Detach+28)
004990a3: 8b11 mov edx, [ecx]
004990a5: ff5204 call dword ptr [edx+04h]
004990a8: 89be2c010000 mov [esi+12c], edi
004990ae: 8b8630010000 mov eax, [esi+130]
004990b4: 3bc7 cmp eax, edi
004990b6: 53 push ebx
004990b7: 8b1db8665300 mov ebx, [005366b8]
004990bd: 7409 jz 004990c8 (VDProjectUI::Detach+48)
004990bf: 50 push eax
004990c0: ffd3 call ebx
004990c2: 89be30010000 mov [esi+130], edi
004990c8: 55 push ebp
004990c9: 8bae3c010000 mov ebp, [esi+13c]
004990cf: 3bef cmp ebp, edi
004990d1: 742a jz 004990fd (VDProjectUI::Detach+7d)
004990d3: 8b8634010000 mov eax, [esi+134]
004990d9: 3bc7 cmp eax, edi
004990db: 7411 jz 004990ee (VDProjectUI::Detach+6e)
004990dd: 50 push eax
004990de: e83ddbfcff call 00466c20 (VDGetIVideoWindow)
004990e3: 8b10 mov edx, [eax] <-- FAULT
004990e5: 83c404 add esp, 04h
004990e8: 57 push edi
004990e9: 8bc8 mov ecx, eax
004990eb: ff5210 call dword ptr [edx+10h]
004990ee: 55 push ebp
004990ef: ffd3 call ebx
004990f1: 89be3c010000 mov [esi+13c], edi
004990f7: 89be44010000 mov [esi+144], edi
004990fd: 8bae40010000 mov ebp, [esi+140]
00499103: 3bef cmp ebp, edi
00499105: 742a jz 00499131 (VDProjectUI::Detach+b1)
00499107: 8b8638010000 mov eax, [esi+138]
0049910d: 3bc7 cmp eax, edi
0049910f: 7411 jz 00499122 (VDProjectUI::Detach+a2)
00499111: 50 push eax
00499112: e809dbfcff call 00466c20 (VDGetIVideoWindow)
00499117: 8b10 mov edx, [eax]
00499119: 83c404 add esp, 04h
0049911c: 57 push edi
0049911d: 8bc8 mov ecx, eax
0049911f: ff5210 call dword ptr [edx+10h]
00499122: 55 push ebp
00499123: ffd3 call ebx
00499125: 89be40010000 mov [esi+140], edi
0049912b: 89be48010000 mov [esi+148], edi
00499131: 8b8634010000 mov eax, [esi+134]
00499137: 3bc7 cmp eax, edi
00499139: 5d pop ebp
0049913a: 7409 jz 00499145 (VDProjectUI::Detach+c5)
0049913c: 50 push eax
0049913d: ffd3 call ebx
0049913f: 89be34010000 mov [esi+134], edi
00499145: 8b8638010000 mov eax, [esi+138]
0049914b: 3bc7 cmp eax, edi
0049914d: 7409 jz 00499158 (VDProjectUI::Detach+d8)
0049914f: 50 push eax
00499150: ffd3 call ebx
00499152: 89be38010000 mov [esi+138], edi
00499158: 8b8628010000 mov eax, [esi+128]
0049915e: 3bc7 cmp eax, edi

Built on KOS-MOS on Sat Nov 04 21:36:36 2006 using compiler version 1200

Windows 5.1 (Windows XP build 2600) [Service Pack 2]

EAX = 00000000
EBX = 77d4daea
ECX = 77d49491
EDX = 7c90eb94
EBP = 000704ae
ESI = 00acf738
EDI = 00000000
ESP = 0012d634
EIP = 004990e3
EFLAGS = 00010246
FPUCW = ffff027f
FPUTW = ffffffff

Crash reason: Access Violation

Crash context:
An out-of-bounds memory access (access violation) occurred in module 'VirtualDub'...

...reading address 00000000.

Pointer dumps:

EBX 77d4daea: 001163b8 0300ba00 12ff7ffe 900004c2 90909090 8b55ff8b 758b56ec 3e836608
ECX 77d49491: 900008c2 ff909090 8effffff 9777d70b 8b77d70b cd042444 0004c22b 90909090
EDX 7c90eb90: 90909090 24a48dc3 00000000 0024648d 90909090 24548d90 c32ecd08 9cec8b55
ESI 00acf738: 00547c14 00ac0178 00080322 00000000 00000000 00000000 00000000 00000000
ESP 0012d630: 00000000 000704a4 00000082 00000000 00acf840 00acf578 004a36b5 00080322
0012d650: 00acf578 00000001 0012d684 00532715 00000000 004a3534 00080322 00acf840
0012d670: 00000001 00000000 004a33f0 0012d6b8 00000000 0012d710 005326fb ffffffff
0012d690: 77d48734 00080322 00000082 00000000 00000000 004a33f0 dcbaabcd 00000000

Thread call stack:
004990e3: VDProjectUI::Detach()
004a36b5: VDUIFrame::DetachNow()
004a3534: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d4d05b: USER32!EnumDisplayMonitors [77d40000+cf63+f8]
77d4b4c0: USER32!DefWindowProcW [77d40000+b33c+184]
77d4dabd: USER32!IsWindowInDestroy [77d40000+da74+49]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d494be: USER32!GetWindowLongA [77d40000+945d+61]
004a34a4: VDUIFrame::StaticWndProc()
77d4b3f9: USER32!DefWindowProcW [77d40000+b33c+bd]
77d4b393: USER32!DefWindowProcW [77d40000+b33c+57]
004a31e1: VDUIFrame::DefProc()
0049bb58: VDProjectUI::MainWndProc()
77d4882a: USER32!GetDC [77d40000+86c7+163]
747309a9: MSCTF!TF_UninitSystem [74720000+10469+540]
77d618f4: USER32!UnhookWinEvent [77d40000+2187d+77]
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
0049b67e: VDProjectUI::WndProc()
004a34a4: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
77d4b4c0: USER32!DefWindowProcW [77d40000+b33c+184]
77d4b50c: USER32!DefWindowProcW [77d40000+b33c+1d0]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d494be: USER32!GetWindowLongA [77d40000+945d+61]
77d484fc: USER32!000084fc
77d4ba0e: USER32!IsWindow [77d40000+b933+db]
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d4b3f9: USER32!DefWindowProcW [77d40000+b33c+bd]
5ad73c20: uxtheme!DrawThemeText [5ad70000+3031+bef]
5ad8e300: uxtheme!GetThemeTextMetrics [5ad70000+1de50+4b0]
5ad71ac7: uxtheme!00001ac7
5ad71b3d: uxtheme!00001b3d
77d4bb15: USER32!GetPropW [77d40000+badb+3a]
004a31e1: VDUIFrame::DefProc()
0049bb58: VDProjectUI::MainWndProc()
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
0049b67e: VDProjectUI::WndProc()
004a34a4: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
77d4b89b: USER32!GetParent [77d40000+b72f+16c]
77d4b903: USER32!SendMessageW [77d40000+b8ba+49]
5ad9881f: uxtheme!GetThemeTextMetrics [5ad70000+1de50+a9cf]
5ad71ac7: uxtheme!00001ac7
5ad71b3d: uxtheme!00001b3d
77d4bb15: USER32!GetPropW [77d40000+badb+3a]
004a31e1: VDUIFrame::DefProc()
0049bb58: VDProjectUI::MainWndProc()
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
0049b67e: VDProjectUI::WndProc()
004a34a4: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d489cd: USER32!GetWindowLongW [77d40000+88a6+127]
77d485a4: USER32!000085a4
77d48a10: USER32!DispatchMessageW [77d40000+8a01+f]
77d5e2b9: USER32!CallMsgFilterW [77d40000+1e0a6+213]
77d561c6: USER32!DrawStateW [77d40000+15fd4+1f2]
77d56208: USER32!DialogBoxIndirectParamAorW [77d40000+161d2+36]
77d6205e: USER32!DialogBoxIndirectParamW [77d40000+22043+1b]
00f32d79: ffdshow!configureEnc [00e50000+d6a80+c2f9]
00460020: HexEditor::Extract()
7c910992: ntdll!RtlFreeAnsiString [7c900000+10976+1c]
7c85a477: kernel32!OutputDebugStringW [7c800000+5a42d+4a]
00e55ce9: ffdshow!00005ce9
7752b767: ole32!CreateGenericComposite [774e0000+4a50b+125c]
774ffa51: ole32!OleInitialize [774e0000+1f6da+377]
7c36b864: MSVCR71!fputs [7c340000+2b71f+145]
7c36c958: MSVCR71!wscanf [7c340000+2c8ee+6a]
7c36c9b7: MSVCR71!wscanf [7c340000+2c8ee+c9]
7c36d180: MSVCR71!wscanf [7c340000+2c8ee+892]
7c36d1d1: MSVCR71!wscanf [7c340000+2c8ee+8e3]
7c36c9b7: MSVCR71!wscanf [7c340000+2c8ee+c9]
77f18f29: GDI32!GetMapMode [77f10000+8dfa+12f]
77f191cd: GDI32!SetDIBitsToDevice [77f10000+903c+191]
77f18f29: GDI32!GetMapMode [77f10000+8dfa+12f]
77f191cd: GDI32!SetDIBitsToDevice [77f10000+903c+191]
77f1918f: GDI32!SetDIBitsToDevice [77f10000+903c+153]
77f191ac: GDI32!SetDIBitsToDevice [77f10000+903c+170]
77f1aa66: GDI32!CreateDIBitmap [77f10000+a935+131]
7c9105c8: ntdll!RtlFreeHeap [7c900000+1043d+18b]
7c910551: ntdll!RtlFreeHeap [7c900000+1043d+114]
7c91056d: ntdll!RtlFreeHeap [7c900000+1043d+130]
77f16e56: GDI32!DeleteObject [77f10000+6c2d+229]
77f16f43: GDI32!DeleteDC [77f10000+6e98+ab]
77f16eaa: GDI32!DeleteDC [77f10000+6e98+12]
77d52874: USER32!EnumDisplayDevicesA [77d40000+1217f+6f5]
77d526f7: USER32!EnumDisplayDevicesA [77d40000+1217f+578]
77d526df: USER32!EnumDisplayDevicesA [77d40000+1217f+560]
77d533b0: USER32!CreateIconIndirect [77d40000+12d02+6ae]
77d53377: USER32!CreateIconIndirect [77d40000+12d02+675]
77d53382: USER32!CreateIconIndirect [77d40000+12d02+680]

-- End of report

[Inventive Software]

With 1.6.15, the same issue (apologies for double post, forum post length restricts it to 16000 characters):

VirtualDub crash report -- build 24442 (release)
--------------------------------------

Disassembly:
00498ec0: 5b pop ebx
00498ec1: c20400 ret 0004
00498ec4: 8b06 mov eax, [esi]
00498ec6: 8bce mov ecx, esi
00498ec8: ff5004 call dword ptr [eax+04h]
00498ecb: 5f pop edi
00498ecc: 5e pop esi
00498ecd: 5d pop ebp
00498ece: 32c0 xor al, al
00498ed0: 5b pop ebx
00498ed1: c20400 ret 0004
00498ed4: 90 nop
00498ed5: 90 nop
00498ed6: 90 nop
00498ed7: 90 nop
00498ed8: 90 nop
00498ed9: 90 nop
00498eda: 90 nop
00498edb: 90 nop
00498edc: 90 nop
00498edd: 90 nop
00498ede: 90 nop
00498edf: 90 nop
00498ee0: 56 push esi
00498ee1: 57 push edi
00498ee2: 8bf1 mov esi, ecx
00498ee4: 8b4608 mov eax, [esi+08h]
00498ee7: 33ff xor edi, edi
00498ee9: 57 push edi
00498eea: 50 push eax
00498eeb: ff1534645300 call dword ptr [00536434]
00498ef1: 57 push edi
00498ef2: 8bce mov ecx, esi
00498ef4: e867abffff call 00493a60 (VDProject::SetUICallback)
00498ef9: 8b8e2c010000 mov ecx, [esi+12c]
00498eff: 3bcf cmp ecx, edi
00498f01: 7405 jz 00498f08 (VDProjectUI::Detach+28)
00498f03: 8b11 mov edx, [ecx]
00498f05: ff5204 call dword ptr [edx+04h]
00498f08: 89be2c010000 mov [esi+12c], edi
00498f0e: 8b8630010000 mov eax, [esi+130]
00498f14: 3bc7 cmp eax, edi
00498f16: 53 push ebx
00498f17: 8b1db8665300 mov ebx, [005366b8]
00498f1d: 7409 jz 00498f28 (VDProjectUI::Detach+48)
00498f1f: 50 push eax
00498f20: ffd3 call ebx
00498f22: 89be30010000 mov [esi+130], edi
00498f28: 55 push ebp
00498f29: 8bae3c010000 mov ebp, [esi+13c]
00498f2f: 3bef cmp ebp, edi
00498f31: 742a jz 00498f5d (VDProjectUI::Detach+7d)
00498f33: 8b8634010000 mov eax, [esi+134]
00498f39: 3bc7 cmp eax, edi
00498f3b: 7411 jz 00498f4e (VDProjectUI::Detach+6e)
00498f3d: 50 push eax
00498f3e: e84dd8fcff call 00466790 (VDGetIVideoWindow)
00498f43: 8b10 mov edx, [eax] <-- FAULT
00498f45: 83c404 add esp, 04h
00498f48: 57 push edi
00498f49: 8bc8 mov ecx, eax
00498f4b: ff5210 call dword ptr [edx+10h]
00498f4e: 55 push ebp
00498f4f: ffd3 call ebx
00498f51: 89be3c010000 mov [esi+13c], edi
00498f57: 89be44010000 mov [esi+144], edi
00498f5d: 8bae40010000 mov ebp, [esi+140]
00498f63: 3bef cmp ebp, edi
00498f65: 742a jz 00498f91 (VDProjectUI::Detach+b1)
00498f67: 8b8638010000 mov eax, [esi+138]
00498f6d: 3bc7 cmp eax, edi
00498f6f: 7411 jz 00498f82 (VDProjectUI::Detach+a2)
00498f71: 50 push eax
00498f72: e819d8fcff call 00466790 (VDGetIVideoWindow)
00498f77: 8b10 mov edx, [eax]
00498f79: 83c404 add esp, 04h
00498f7c: 57 push edi
00498f7d: 8bc8 mov ecx, eax
00498f7f: ff5210 call dword ptr [edx+10h]
00498f82: 55 push ebp
00498f83: ffd3 call ebx
00498f85: 89be40010000 mov [esi+140], edi
00498f8b: 89be48010000 mov [esi+148], edi
00498f91: 8b8634010000 mov eax, [esi+134]
00498f97: 3bc7 cmp eax, edi
00498f99: 5d pop ebp
00498f9a: 7409 jz 00498fa5 (VDProjectUI::Detach+c5)
00498f9c: 50 push eax
00498f9d: ffd3 call ebx
00498f9f: 89be34010000 mov [esi+134], edi
00498fa5: 8b8638010000 mov eax, [esi+138]
00498fab: 3bc7 cmp eax, edi
00498fad: 7409 jz 00498fb8 (VDProjectUI::Detach+d8)
00498faf: 50 push eax
00498fb0: ffd3 call ebx
00498fb2: 89be38010000 mov [esi+138], edi
00498fb8: 8b8628010000 mov eax, [esi+128]
00498fbe: 3bc7 cmp eax, edi

Built on KOS-MOS on Sun May 28 15:10:00 2006 using compiler version 1200

Windows 5.1 (Windows XP build 2600) [Service Pack 2]

EAX = 00000000
EBX = 77d4daea
ECX = 77d49491
EDX = 7c90eb94
EBP = 000604d0
ESI = 00bb00f8
EDI = 00000000
ESP = 0012d634
EIP = 00498f43
EFLAGS = 00010246
FPUCW = ffff027f
FPUTW = ffffffff

Crash reason: Access Violation

Crash context:
An out-of-bounds memory access (access violation) occurred in module 'VirtualDub'...

...reading address 00000000.

Pointer dumps:

EBX 77d4daea: 001163b8 0300ba00 12ff7ffe 900004c2 90909090 8b55ff8b 758b56ec 3e836608
ECX 77d49491: 900008c2 ff909090 8effffff 9777d70b 8b77d70b cd042444 0004c22b 90909090
EDX 7c90eb90: 90909090 24a48dc3 00000000 0024648d 90909090 24548d90 c32ecd08 9cec8b55
ESI 00bb00f8: 00547bdc 00ac0178 0011035c 00000000 00000000 00000000 00000000 00000000
ESP 0012d630: 00000000 000d0494 00000082 00000000 00bb0200 00bb26e0 004a3515 0011035c
0012d650: 00bb26e0 00000001 0012d684 00532435 00000000 004a3394 0011035c 00bb0200
0012d670: 00000001 00000000 004a3250 0012d6b8 00000000 0012d710 0053241b ffffffff
0012d690: 77d48734 0011035c 00000082 00000000 00000000 004a3250 dcbaabcd 00000000

Thread call stack:
00498f43: VDProjectUI::Detach()
004a3515: VDUIFrame::DetachNow()
004a3394: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d4d05b: USER32!EnumDisplayMonitors [77d40000+cf63+f8]
77d4b4c0: USER32!DefWindowProcW [77d40000+b33c+184]
77d4dabd: USER32!IsWindowInDestroy [77d40000+da74+49]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d494be: USER32!GetWindowLongA [77d40000+945d+61]
7c3556b3: MSVCR71!wcstod [7c340000+154f8+1bb]
7c35575f: MSVCR71!wcstod [7c340000+154f8+267]
77d4b3f9: USER32!DefWindowProcW [77d40000+b33c+bd]
77d4b393: USER32!DefWindowProcW [77d40000+b33c+57]
004a3041: VDUIFrame::DefProc()
0049b9b8: VDProjectUI::MainWndProc()
747309a9: MSCTF!TF_UninitSystem [74720000+10469+540]
77d618f4: USER32!UnhookWinEvent [77d40000+2187d+77]
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
0049b4de: VDProjectUI::WndProc()
004a3304: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
77d4b4c0: USER32!DefWindowProcW [77d40000+b33c+184]
77d4b50c: USER32!DefWindowProcW [77d40000+b33c+1d0]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d494be: USER32!GetWindowLongA [77d40000+945d+61]
77d484fc: USER32!000084fc
77d4ba0e: USER32!IsWindow [77d40000+b933+db]
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d4b3f9: USER32!DefWindowProcW [77d40000+b33c+bd]
5ad73c20: uxtheme!DrawThemeText [5ad70000+3031+bef]
5ad8e300: uxtheme!GetThemeTextMetrics [5ad70000+1de50+4b0]
5ad71ac7: uxtheme!00001ac7
5ad71b3d: uxtheme!00001b3d
77d4bb15: USER32!GetPropW [77d40000+badb+3a]
004a3041: VDUIFrame::DefProc()
0049b9b8: VDProjectUI::MainWndProc()
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
0049b4de: VDProjectUI::WndProc()
004a3304: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
77d4b89b: USER32!GetParent [77d40000+b72f+16c]
77d4b903: USER32!SendMessageW [77d40000+b8ba+49]
5ad9881f: uxtheme!GetThemeTextMetrics [5ad70000+1de50+a9cf]
5ad71ac7: uxtheme!00001ac7
5ad71b3d: uxtheme!00001b3d
77d4bb15: USER32!GetPropW [77d40000+badb+3a]
004a3041: VDUIFrame::DefProc()
0049b9b8: VDProjectUI::MainWndProc()
77d484fc: USER32!000084fc
77d485a4: USER32!000085a4
77d49488: USER32!GetWindowLongA [77d40000+945d+2b]
0049b4de: VDProjectUI::WndProc()
004a3304: VDUIFrame::StaticWndProc()
77d48734: USER32!GetDC [77d40000+86c7+6d]
77d48816: USER32!GetDC [77d40000+86c7+14f]
7c90eae3: ntdll!KiUserCallbackDispatcher [7c900000+ead0+13]
77d489cd: USER32!GetWindowLongW [77d40000+88a6+127]
77d485a4: USER32!000085a4
77d48a10: USER32!DispatchMessageW [77d40000+8a01+f]
77d5e2b9: USER32!CallMsgFilterW [77d40000+1e0a6+213]
77d561c6: USER32!DrawStateW [77d40000+15fd4+1f2]
77d56208: USER32!DialogBoxIndirectParamAorW [77d40000+161d2+36]
77d6205e: USER32!DialogBoxIndirectParamW [77d40000+22043+1b]
00dc2d79: ffdshow!configureEnc [00ce0000+d6a80+c2f9]
7c910992: ntdll!RtlFreeAnsiString [7c900000+10976+1c]
7c85a477: kernel32!OutputDebugStringW [7c800000+5a42d+4a]
00ce5ce9: ffdshow!00005ce9
7752b767: ole32!CreateGenericComposite [774e0000+4a50b+125c]
774ffa51: ole32!OleInitialize [774e0000+1f6da+377]
7c36b864: MSVCR71!fputs [7c340000+2b71f+145]
7c36c958: MSVCR71!wscanf [7c340000+2c8ee+6a]
7c36c9b7: MSVCR71!wscanf [7c340000+2c8ee+c9]
7c36d180: MSVCR71!wscanf [7c340000+2c8ee+892]
7c36d1d1: MSVCR71!wscanf [7c340000+2c8ee+8e3]
7c36c9b7: MSVCR71!wscanf [7c340000+2c8ee+c9]
77f18f29: GDI32!GetMapMode [77f10000+8dfa+12f]
77f191cd: GDI32!SetDIBitsToDevice [77f10000+903c+191]
77f18f29: GDI32!GetMapMode [77f10000+8dfa+12f]
77f191cd: GDI32!SetDIBitsToDevice [77f10000+903c+191]
77f1918f: GDI32!SetDIBitsToDevice [77f10000+903c+153]
77f191ac: GDI32!SetDIBitsToDevice [77f10000+903c+170]
77f1aa66: GDI32!CreateDIBitmap [77f10000+a935+131]
7c9105c8: ntdll!RtlFreeHeap [7c900000+1043d+18b]
7c910551: ntdll!RtlFreeHeap [7c900000+1043d+114]
7c91056d: ntdll!RtlFreeHeap [7c900000+1043d+130]
77f16e56: GDI32!DeleteObject [77f10000+6c2d+229]
77f16f43: GDI32!DeleteDC [77f10000+6e98+ab]
77f16eaa: GDI32!DeleteDC [77f10000+6e98+12]
77d52874: USER32!EnumDisplayDevicesA [77d40000+1217f+6f5]
77d526f7: USER32!EnumDisplayDevicesA [77d40000+1217f+578]
77d526df: USER32!EnumDisplayDevicesA [77d40000+1217f+560]
77d533b0: USER32!CreateIconIndirect [77d40000+12d02+6ae]
77d53377: USER32!CreateIconIndirect [77d40000+12d02+675]
77d53382: USER32!CreateIconIndirect [77d40000+12d02+680]
7c910833: ntdll!RtlAllocateHeap [7c900000+105d4+25f]

-- End of report

[Inventive Software]

For the record, the same issue occurs with 1.6.11, and 1.5.10 ;) I'm not gonna post the crash log, unless it's requested.

[vlada]

cc979
You are probably confused by DVDs. I would suggest you to forget that you ever heard about 720x576. It is 5:4 and your monitor is 4:3. So it's absolutely clear that there will be borders at the sides, because you probably set ffdshow to letterbox your movie to 4:3. So ffdshow behaves absolutely correctly IMHO. If you don't use subtitles the letterboxing is up to a player so there are no black borders at sides.

[Delerue]

Is there any plan to improve the motion estimation in 'framerate doubler'? I love the idea, but today (until build 960 if I'm not wrong) it has a lot of bugs and can't render well a few videos.

Thanks, guys. And congratulations for this amazing job. Keep going. ;)

[haruhiko_yamagata]

maybe the auto y size thing could be displayed on resize tab somewhere in ffdshow
Thanks, it's a good idea.

[multiblitz]

OK, Just a short update on playback of HD-DVD / VC-1 with ffdsshow: I managed to get a graph working with the Sonic-De-Muxer/Video-Decoder and ffdshow raw.

The picture is A M A Z I N G.

I am using a Ruby and a 2.2m*4.2m screen. So, pq matters a lot.

The drawback:
- In graphedit, I can connect ffdshow audio to the Sonic HD De-Muxer, but even with SPDIF passthrough I do not get any sound.
- I would like to use ffdshow as well as video-decoder. I understand in theory it is possible (VC-1). But it does not work yet, any ideas ?


Clearly the picture is so much better that you want nothing else once you saw it. HD-DVD is nice, but with ffdshow it is simply unbelieveable, much better than in a cinema.

[madshi]

- In graphedit, I can connect ffdshow audio to the Sonic HD De-Muxer, but even with SPDIF passthrough I do not get any sound.
With most HD DVDs SPDIF passthrough can't work, because most HD DVDs use E-AC3 (and not AC3), which is a format receivers don't understand via SPDIF.

- I would like to use ffdshow as well as video-decoder. I understand in theory it is possible (VC-1). But it does not work yet, any ideas ?
Yeah, I'd also love to see VC-1 decoding working with ffdshow tryouts. It kind of works with the Sonic video decoder for me, but it's not really fluid that way. Having more decoder choices is always a great thing.

[Keepitsimple]

Is anybody working on enhancements to the x264 decoder? I have to overclock my core2duo to 3Ghz to play 1080p with ffdshow/mpc right now. The problem is mostly that only one core is utilized.

[LoRd_MuldeR]

Is anybody working on enhancements to the x264 decoder? I have to overclock my core2duo to 3Ghz to play 1080p with ffdshow/mpc right now. The problem is mostly that only one core is utilized.

At first there is no such thing as a "x264 decoder". x264 is only an encoder, it is *not* a decoder! ffdshow uses the libavcodec from ffmpeg for decoding H.264 video streams. AFAIK this is the only (working) OpenSource decoder for H.264 available. ffdshow-tryout usually adopts ffmpeg updates pretty fast, so we will have to wait for the ffmpeg team to further improve their H.264 decoder. BTW: Did you try to disable the Inloop Deblocking in order to boost the performance?

[Inventive Software]

Did you also enable "Queueing", so that both cores are utilised (one decodes / renders, the other queues the frames)?

[Keepitsimple]

I dont know what neither "Inloop Deblocking" nor "Queueing" is.

I got 20% less utilization by not forcing rgb as output. However without RGB my movies hang when I move them between screens. :(

[_xxl]

I got 20% less utilization by not forcing rgb as output. However without RGB my movies hang when I move them between screens.
Can you provide a sample?

[_xxl]

Is anybody working on enhancements to the x264 decoder? I have to overclock my core2duo to 3Ghz to play 1080p with ffdshow/mpc right now. The problem is mostly that only one core is utilized.
No multithreaded support for h264 libavcodec decoder.

[foxyshadis]

The second-screen stuttering is an ATI driver problem, it can't deal with splitting a video across two screens. (And has similar problems with GDI windows being placed on top of a VMR9 surface - notably a right-click menu or codec config window.) The 20% overhead for RGB sounds pretty high though.

[Keepitsimple]

it was pretty much every movie. seems to have vanished with haalis new renderer. however it stilll takes a lot longer to switch screens. with rgb its not even a black picture if you move it fast. with nv12 it sort of stops at the edge for 1 second before it moves to the next screen.

[fastplayer]

The volume levels are updated very slowly in the latest 969 build. In previous builds they were updated "real-time"-like. Maybe this has something to do with bug fix rev963...

[LoRd_MuldeR]

The volume levels are updated very slowly in the latest 969 build. In previous builds they were updated "real-time"-like. Maybe this has something to do with bug fix rev963...

I can confirm that with r971 build. Also effects the "Info & Debug" dialog.
You can notice that when you play VBR audio and watch "Current input bitrate"

Maybe those values are just updated less often to save some CPU cycles?

[fastplayer]

AFAIK the bug was only in the audio mixer section. At least I couldn't reproduce it in the volume section...

Edit: No, it wasn't the mixer section. It was just enough to have the audio dialog opened for CPU usage to increase.

[haruhiko_yamagata]

The volume levels are updated very slowly in the latest 969 build. In previous builds they were updated "real-time"-like. Maybe this has something to do with bug fix rev963...
Thank you. I can confirm. The dialog is updated every 500ms, because "Mixer" is still heavy at 100ms. The code judging by CPU monitor may help.

[madshi]

One question to you ffdshow tryouts gurus:

How can I make "WVC1 -> wmv9" work? If I activate "WVC1 -> libavcodec" I can successfully decode VC-1 EVO files (with the latest Haali Media Splitter). But as soon as I choose "WVC1 -> wmv9" the Haali Media Splitter and ffdshow tryouts can't connect, anymore.

FWIW, I have Microsoft's WVC1 decoder (wvc1dmod.dll) downloaded and registered.

Thank you guys!

[KoD]

Does WVC1 -> disabled work ?

[madshi]

Does WVC1 -> disabled work ?
If I set it to "disabled" or "wmv9", Haali Media Splitter and ffdshow cannot connect. There's no difference between "disabled" and "wmv9". Haali Media Splitter and ffdshow tryouts can connect only if I set it to "WVC1" to "libavcodec".

[_xxl]

Works for me.Please install WMP11.
http://img150.imagevenue.com/loc21/th_89542_Capture_122_21lo.jpg (http://img150.imagevenue.com/img.php?image=89542_Capture_122_21lo.jpg)

[madshi]

Thank you - after install WMP11 it works! It's not really fluid, though, although I have a quite fast PC (Merom Dual-Core). Too bad... :scared:

Edit: After muxing to MKV things are fluid - yey!!

[Delerue]

I found a strange thing: if a subtitle open a italic tag ('<i>') but doesn't close it in the same line (dialog), so all subtitles after that remain italic until you write a close italic tag ('</i>'). This thing doesn't happen with VobSub, DirectVobSub and Subtitle Workshop.

EDIT: Forgot to say that the build tested was 971.

[jffulcrum]

Found bug in Resize&Aspect filter.
When i select 2xSai, 2xSuperSai or hq2x methods and leave in "New size" filed vertical size as 0 (automatic), resize actually not performed.
FFDShow-tryout build 967 (also checked 927), MPC latest build by celtic_druid (with BSPlayer also wrong).

[diizzy]

Hi,
First of all I have to admit that I've been a "silent" user for over a year so I'm not really a new member. I also would like the developers of ffdshow/ffdshow-tryouts for their work and time spent on the project, its really appreciated.

Anyhow I have a little bug report/feature request.

Issue: "RAW" interlaced/deinterlaced video doesnt get detected as interlaced using scalers resulting in improper scaling.

How to reproduce: Play a deinterlaced source using an external filter and upscale.

Fix: Instead of letting ffdshow decide for itself (having the Interlaced box gray) make sure to check it.

Software used:

ffdshow_rev906_20070210_clsid.exe
YV12 input and YV12 output
Lanczos used as scaler algorithm

Dscaler5.IVTC.251206.zip
Can be found here
http://www.avsforum.com/avs-vb/showthread.php?p=9266528&&#post9266528

MPC - mplayerc.rev611-3.2kxp.7z
http://ffdshow.faireal.net/
VMR7 (windowed) and WMR9 mixer mode disabled as output

So the flow is the following: Splitter --> Dscaler5 (decoding) --> ffdshow (scaling) --> MPC

I'm also wondering if its possible to include this modified/optimized MPEG audio decoder into ffdshow.
http://otachan.com/in_!mpg123.html

I would love to contribute but my programming skills arent that great yet :-/

//Danne

[mariner]

Hi drevil_xxl,

"Media Player 2 (MPC) + Haali Media Splitter + ffdshow ac3 decoder."

Greetings and thanks for the ffdshow decoders., but was not able to get it to decode DD+. It won't connect.

Thanks for you kind assistance.

[foxyshadis]

Found bug in Resize&Aspect filter.
When i select 2xSai, 2xSuperSai or hq2x methods and leave in "New size" filed vertical size as 0 (automatic), resize actually not performed.
FFDShow-tryout build 967 (also checked 927), MPC latest build by celtic_druid (with BSPlayer also wrong).
How do you leave anything in new size? The entire left side of the resize panel should be greyed out when you choose one of those three methods, because it's always set to 2x. If this isn't actually the "resize if" section, you might have found some GUI bug.

Oh yeah, almost forgot, you have to reload the video after setting it to one of the game resizers or back. They aren't on-the-fly capable at all.

[haruhiko_yamagata]

Found bug in Resize&Aspect filter.
When i select 2xSai, 2xSuperSai or hq2x methods and leave in "New size" filed vertical size as 0 (automatic), resize actually not performed.
FFDShow-tryout build 967 (also checked 927), MPC latest build by celtic_druid (with BSPlayer also wrong).
What is "2xSai, 2xSuperSai or hq2x methods" ?

[haruhiko_yamagata]

Issue: "RAW" interlaced/deinterlaced video doesnt get detected as interlaced using scalers resulting in improper scaling.
Hi,
Does checking "Resize & aspect->Settings->Interlaced" fix your problem?
If your report means ffdshow does not do it automatically, I can reproduce your issue.

[kikker]

Hi, I have the latest 971 build, but I can't seem to switch my presets. I have 4 besides the "default", which is the only one in bold type (active preset?). Ffdshow remembers all my settings and are entered in the registry, but the "active preset" reg entry never changes from "default", so none of the other presets can be applied.

I am running Vista and the ffdshow preset settings are at HKCU\software\GNU. I am also running as a limited user, but this shouldn't matter as all the preset settings are entered just fine.
Ffdshow was installed as an admin, but when I try to run it now as an admin, I get an error "ffdshow.ax module couldn't be found."

Why can't that reg entry "active preset" be changed?

[JarrettH]

aren't those resize modes used in zSNES?

[Syris2k4]

They are methods for smoothing resized images, so yes... things like zSNES, Exult and Pentagram have them to upscale the wonderfully pixilated early 90's bitmaps :D

@haruhiko_yamagata the methods he refers to are found in :
Resize and aspect -> Settings -> Luma Method
Check the drop down list, they are near the bottom... if you didnt find them :)

[HeadBangeR77]

@ kikker
In case it could be of any use, double-clicking on a chosen preset (I've got more than ten of them) makes it bold, and it's loaded as the active one after re-opening ffdshow /player. One-click just changes the preset sort of temporarily, until you close ffdshow /player - should work for you too. Generally speaking, no problems here with rev 971, but I'm running WinXP Pro SP2.

[diizzy]

@ haruhiko_yamagata
Yes, checking interlaced does help.
Is it possible to make this option work automatically?
//Danne

[kikker]

@ kikker
In case it could be of any use, double-clicking on a chosen preset (I've got more than ten of them) makes it bold, and it's loaded as the active one after re-opening ffdshow /player. One-click just changes the preset sort of temporarily, until you close ffdshow /player - should work for you too. Generally speaking, no problems here with rev 971, but I'm running WinXP Pro SP2.

:) I don't know if I hate it or love it that such answers can be so simple. Who would have thought to *double*-click on it?

[HeadBangeR77]

:) I don't know if I hate it or love it that such answers can be so simple. Who would have thought to *double*-click on it?
Wish all the answers concerning this forum were so simple. ;) :D
Does one-click work for you also, in the way I described it above?

[zambelli]

Would be possible to integrate some sort of standard performance test into the build compile process? Commonly used functions such as XviD decoding could be benchmarked, for example. The results might not mean much on their own, but if perf results were available for every new tryout build, then it'd be possible to track regressions in performance and answer the often asked question "Is this build faster than xyz build?"

[Keepitsimple]

Levels in ffdshow

As I have understood it:

The histogram has intensity on the y axis and grayscale on the x-axis.

Posterize determines how many levels the grayscale will have.

Now some questions:

Fullrange makes the picture gets darker. The histogram also compresses to 16-235 levels instead of 0-255. But what is fullrange?

What do the different modes and modify only luminance mean?

How do I know what levels a particular media file uses? Ie should I look at the histogram with fullrange enabled or not?

[foxyshadis]

@ haruhiko_yamagata
Yes, checking interlaced does help.
Is it possible to make this option work automatically?
//Danne

If the solution was as simple as the problem, it would have been done a very long time ago. Interlaced detection is a very hard problem, because there's no good realtime method without a false positive/negative rate of at least 5%.

[foxyshadis]

Levels in ffdshow

As I have understood it:

The histogram has intensity on the y axis and grayscale on the x-axis.
Well, the relative count of pixels having that intensity.

Fullrange makes the picture gets darker. The histogram also compresses to 16-235 levels instead of 0-255. But what is fullrange?

What do the different modes and modify only luminance mean?

How do I know what levels a particular media file uses? Ie should I look at the histogram with fullrange enabled or not?

Fullrange is 0-255, Video range is 16-240. Most video isn't fullrange, in fact it's quite rare to see them unless you create them yourself. However, because of the uselessness of different video drivers, expansion to fullrange is sometimes needed. By checking fullrange you get the real histogram and can expand it to the full levels. Otherwise you get one without the (often empty) extremes, the input will be clamped to (16,240) and the output scaled to (16,240), removing any BWB or WTW content.

It's the output scaling that makes the difference in brightness when you change the fullrange state, if you don't change the output sliders yourself.

[haruhiko_yamagata]

Found bug in Resize&Aspect filter.
When i select 2xSai, 2xSuperSai or hq2x methods and leave in "New size" filed vertical size as 0 (automatic), resize actually not performed.
FFDShow-tryout build 967 (also checked 927), MPC latest build by celtic_druid (with BSPlayer also wrong).
I understood, but I can't reproduce this. It works with vertical size 0 (automatic).

[haruhiko_yamagata]

@ haruhiko_yamagata
Yes, checking interlaced does help.
Is it possible to make this option work automatically?
//Danne
Checking "Interlaced" intermidiate state (half checked) help automatic detection. It works correctly as long as the inputing information is correct.
ffdshow can't detect weave or bob. So it's hard to detect automatically in DVD.

What the default setting should be?
Perhaps ffdshow resizer is often used for DVD upscaling, and DVD is weave in most cases.
If it is bob (truely interlaced), it is recommended to use software de-interlacer before resizing.
So I think the default setting is better to be off (as it is now).

[tekNerd]

ffdshow 971, icl 9

Even if Resize&Aspect is not checked I still can override the Aspect ratio. Is this normal?

About the Audio, delay filter: Is it possible to have something more like MPC's "Audio time shift"? I am not using it that much [actually almost zero], but...just asking.

Anywho, thanks to all for this beautiful piece of software and for all the help that I find in this great forum.

[kikker]

Wish all the answers concerning this forum were so simple. ;) :D
Does one-click work for you also, in the way I described it above?

Yes, it works like you describe.

[cc979]

Would be possible to integrate some sort of standard performance test into the build compile process? Commonly used functions such as XviD decoding could be benchmarked, for example. The results might not mean much on their own, but if perf results were available for every new tryout build, then it'd be possible to track regressions in performance and answer the often asked question "Is this build faster than xyz build?"

do you mean to benchmark a compile after each different compiler and test to see what each module/or setting work best ?

i did benchmark most builds earlier but not that much of late

ffdshow_rev971_20070227.exe -o3 pentium3
null: 9s, kernel: 0s, total: 9s, real: 9s, fps: 71.8, dfps: 70.6
vmr9: 9s, kernel: 2s, total: 12s, real: 13s, fps: 53.9, dfps: 49.0

ffdshow-913-sse1-gcc4.1.2
vmr9: 8s, kernel: 1s, total: 10s, real: 12s, fps: 60.8, dfps: 52.5

ffdshow-20060929-rev655-cc979-sse.exe
null: User: 8s, kernel: 0s, total: 8s, real: 10s, fps: 74.5, dfps: 63.0
vmr9: User: 9s, kernel: 1s, total: 11s, real: 12s, fps: 58.3, dfps: 51.0

cc979 rev584 -o3
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 81.3, dfps: 74.7
vmr9: User: 7s, kernel: 0s, total: 8s, real: 10s, fps: 81.8, dfps: 61.1
ovrl: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 75.8, dfps: 72.7

clsid
ffdshow_rev584_20061123_clsid.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 75.1, dfps: 72.9
vmr9: User: 8s, kernel: 0s, total: 8s, real: 11s, fps: 75.8, dfps: 58.0
ovrl: User: 9s, kernel: 0s, total: 9s, real: 9s, fps: 71.4, dfps: 69.4

clsid
ffdshow_rev494_20061103_clsid.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 75.1, dfps: 74.0
vmr9: User: 8s, kernel: 0s, total: 8s, real: 11s, fps: 77.4, dfps: 58.3
ovrl: User: 8s, kernel: 0s, total: 9s, real: 9s, fps: 72.4, dfps: 69.4

clsid
ffdshow_rev348_20061008_clsid.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 77.7, dfps: 75.8
vmr9: User: 8s, kernel: 0s, total: 8s, real: 10s, fps: 79.0, dfps: 61.1
ovrl: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 73.6, dfps: 71.3

clsid
ffdshow_rev187_20060915_clsid.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 76.4, dfps: 73.7
vmr9: User: 8s, kernel: 0s, total: 8s, real: 10s, fps: 77.8, dfps: 59.8
ovrl: User: 8s, kernel: 0s, total: 9s, real: 9s, fps: 72.4, dfps: 70.3

clsid
ffdshow_rev187_20060915_clsid_generic.exe
null: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 74.5, dfps: 70.1
vmr9: User: 8s, kernel: 0s, total: 8s, real: 11s, fps: 75.0, dfps: 58.1
ovrl: User: 8s, kernel: 0s, total: 9s, real: 9s, fps: 72.7, dfps: 70.1

clsid
ffdshow_rev181_20060913_clsid.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 75.7, dfps: 74.6
vmr9: User: 8s, kernel: 0s, total: 8s, real: 11s, fps: 75.4, dfps: 59.0
ovrl: User: 9s, kernel: 0s, total: 9s, real: 9s, fps: 70.2, dfps: 68.7

videomixer9
ffdshow-tryouts-rev155.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 81.6, dfps: 80.7
vmr9: User: 7s, kernel: 1s, total: 9s, real: 10s, fps: 72.0, dfps: 61.2
User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 78.2, dfps: 75.0

drevil_xxl
ffdshow-tryouts-rev155-AMD-XP-K8.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 79.6, dfps: 78.7
vmr9: User: 8s, kernel: 1s, total: 9s, real: 10s, fps: 68.3, dfps: 59.8
ovrl: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 76.8, dfps: 73.0

videomixer9
ffdshow-tryouts-rev148.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 81.4, dfps: 79.0
vmr9: User: 8s, kernel: 0s, total: 9s, real: 10s, fps: 68.4, dfps: 60.8
ovrl: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 77.2, dfps: 74.6

drevil_xxl
ffdshow-tryouts-rev145-amd-xp.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 79.3, dfps: 76.5
vmr9: User: 8s, kernel: 0s, total: 9s, real: 10s, fps: 67.1, dfps: 59.5
ovrl: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 74.9, dfps: 72.2
...
ffdshow-tryouts-rev145-k8.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 78.5, dfps: 75.7
vmr9: User: 8s, kernel: 1s, total: 9s, real: 10s, fps: 66.2, dfps: 59.9
ovrl: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 75.3, dfps: 72.8
...
ffdshow-tryouts-rev145-p4.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 79.7, dfps: 76.8
vmr9: User: 8s, kernel: 1s, total: 9s, real: 10s, fps: 70.2, dfps: 59.9
ovrl: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 74.5, dfps: 72.7

clsid
ffdshow_rev142_20060906.exe
null: User: 8s, kernel: 0s, total: 8s, real: 8s, fps: 78.8, dfps: 77.7
vmr9: User: 8s, kernel: 1s, total: 9s, real: 10s, fps: 65.5, dfps: 59.8
ovrl: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 75.1, dfps: 72.5

drevil_xxl
ffdshow-tryouts-rev138-sse.exe
null: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 78.7, dfps: 70.7
vmr9: User: 8s, kernel: 0s, total: 9s, real: 10s, fps: 69.4, dfps: 59.7
ovrl: User: 8s, kernel: 0s, total: 8s, real: 9s, fps: 75.3, dfps: 72.0

ffdshow-tryouts-rev138-sse2.exe
null: User: 9s, kernel: 0s, total: 10s, real: 10s, fps: 65.4, dfps: 64.7
vmr9: User: 10s, kernel: 0s, total: 11s, real: 12s, fps: 58.3, dfps: 51.3
ovrl: User: 10s, kernel: 0s, total: 10s, real: 10s, fps: 63.1, dfps: 61.2

there has been changes up/down in different revs

[_xxl]

Decoding speed is provided by libavcodec.
Libavcodec compiled with -msse/sse2 is slower.
Please test generic builds only.

[hdboy]

Using Haali's latest splitter and celtic_druid's latest MPC build and ffdshow tryouts 3/01/07 nightly build.

I get video playing an EVO file, but no audio. Looking at haali's pin info, the audio subtype is

{0000EAC3-0000-0010-8000-00AA00389B71}

which must be E-AC3. I've set ffdshow audio decoder to "prefer" in MPC, and even added the subtype above to the list for ffdshow audio decoder under MEDIATYPE_Audio, but no sound and ffdshow audio is not being invoked. MPC's filter menu shows

Video Renderer
Overlay Mixer
WMVVideo Decoder DMO (WMP11 installed)
Haali Media Splitter (AR)
filename.evo

I've also tried to use Cyberlink audio decoder PDVD6.x, PDVD7, PDVD7 UPNP. No go.

Powerdvd gives me incompatible graphic device (I have a 7600GT non HDCP card and non-HDCP monitor). I'm close with MPC. suggestions? thanks.

[fastplayer]

ffdshow doesn't support E-AC3.

[zambelli]

do you mean to benchmark a compile after each different compiler and test to see what each module/or setting work best ?
Sort of. It doesn't necessarily need to be a comparison of different compilers; it could be just a way to track performance over time too.

One could pick most commonly used components - i.e. libavcodec decoder, resizer, deinterlacer, sharpener - and then benchmark how long it takes to decode, resize, or whatnot a standard piece of test content.

If this was automatically run post-build and results posted with the downloads, then users would have some sort of idea what to expect from particular builds in terms of performance. It would also help the devs spot performance bugs when they creep up due to other checkins.

i did benchmark most builds earlier but not that much of late
That's pretty interesting. What did you measure exactly and how did you measure it?

[cc979]

That's pretty interesting. What did you measure exactly and how did you measure it?

i use haali splitter and his time-codec - with a hd-xvid - using no resize, no filters or anything - default ffdshow settings - and it tells you the fps for null/overlay mixer/vmr9

[_xxl]

Does Haali Media Splitter supports E-AC3?
ATSC A/52b is an extension to A/52a and was approved for the Bluray and HD-DVD standard. It is the same as Enhanced AC3 or EAC3 for short. EAC3 is also supposed to be backwards compatible with AC3.
EAC3 is backwards compatible with AC3?

[cc979]

Does Haali Media Splitter supports E-AC3?

EAC3 is backwards compatible with AC3?

it does now i think, but you need an EAC3 filter eg. windvd's

[_xxl]

http://img.photobucket.com/albums/0903/pakthedude/Screenshots/greenlines.jpg
http://sourceforge.net/tracker/index.php?func=detail&aid=1669045&group_id=53761&atid=471489
Tested:
directvobsub 2.37 + ffdshow-rev899_20070209_gcc_xxl +WMP

[haruhiko_yamagata]

http://sourceforge.net/tracker/index.php?func=detail&aid=1669045&group_id=53761&atid=471489
Tested:
directvobsub 2.37 + ffdshow-rev899_20070209_gcc_xxl +WMP
What is the holizontal size?
Does unchecking YV12 help?

[nurbs]

Does Haali Media Splitter supports E-AC3?

EAC3 is backwards compatible with AC3?

It is not backwards compatible per se, but hardware that supports E-AC3 has to be able to convert it to AC3.

[Peuj]

Hi,

I have 2 questions about the ffdshow WMV codecs:

in the ffdshow video there are:
WMV1/7
WMV2/8
WMV3/9

and in the ffdshow audio there are:
WMA 7
WMA 8/9

Does the WMA 7 and WMA8/9 is them than WMA 1 and WMA 2/3?
If yes then you should update the names (as it is with the video) with something like:
WMA 1/7
WMA 2/3 8/9

If I set libavcodec for WMA8/9 and a play a movie with an audio stream WMA2, then the ffdshow codec is used.
But if I play a movie with an audio stream WMA3 the ffdshow codec is not used.

So maybe it's normal but then the information should be updated too and separate WMA 2/8 and WMA 3/9.

Last question, are there some known interests to use libavcodec for VMW/A. I just use them because I wanted to try the "set all stable formats to libavcodec" but i don't see better performance or quality.

Thanks

[wyrd]

Hi,authors
Thank you for your efforts.

I can't playback(weird color/block noise?) this sample in wmv3(libavcodec).(ff_wmv9 is fine)
http://tirnanog.fate.jp/tmp/sample/wmv3/FullMetal%5bwmv+AAC%5d_remux-006.mkv
As for this sample works fine with libavcodec/ff_wmv9 either.
http://tirnanog.fate.jp/tmp/sample/wmv3/Bebop%5bwmv+AC3%5d_remux-004.mkv
I have tested with mpc/wmp10 and haali/gabest splitter.
Can anyone reproduce?

Regards.

[Peuj]

Hi,authors
Thank you for your efforts.

I can't playback(weird color/block noise?) this sample in wmv3(libavcodec).(ff_wmv9 is fine)
http://tirnanog.fate.jp/tmp/sample/wmv3/FullMetal%5bwmv+AAC%5d_remux-006.mkv
As for this sample works fine with libavcodec/ff_wmv9 either.
http://tirnanog.fate.jp/tmp/sample/wmv3/Bebop%5bwmv+AC3%5d_remux-004.mkv
I have tested with mpc/wmp10 and haali/gabest splitter.
Can anyone reproduce?

Regards.

Hi,

I can reproduce your playback problem (weird color/block noise).
I only have the problem with WMV3 set to libavcodec

[Px]

http://sourceforge.net/tracker/index.php?func=detail&aid=1669045&group_id=53761&atid=471489
Tested:
directvobsub 2.37 + ffdshow-rev899_20070209_gcc_xxl +WMP
Bug, more vobsub bug, than ffdshow. May i guess, that that video have non-standart dimensions, like 638x478 instead of 640x480?

[_xxl]

Bebop[wmv+AC3]_remux-004.mkv

Works for me.
MPC(WMP11)+Haali+latest ffdshow.
FullMetal[wmv+AAC]_remux-006.mkv

Confirmed.
Maybe libavcodec wmv3 support is incomplete.
Opening video decoder: [dmo] DMO video codecs
ERROR: Could not open required DirectShow codec wmvdmod.dll.
IMediaObject ERROR: 00DF62C6 no such class object (0x80004001 : -2147467263)
Failed to create DMO filter
You need to upgrade/install the binary codecs package.
Go to http://www.mplayerhq.hu/dload.html
VDecoder init failed :(
Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
[wmv3 @ 00CC2A50]Profile value 2 is forbidden (and WMV3 Complex Profile is unsupported)
Could not open codec.
VDecoder init failed :(
Cannot find codec matching selected -vo and video format 0x33564D57.
Read DOCS/HTML/en/codecs.html!
==========================================================================
Opening audio decoder: [faad] AAC (MPEG2/4 Advanced Audio Coding)
Opening audio decoder: [faad] AAC (MPEG2/4 Advanced Audio Coding)
FAAD: compressed input bitrate missing, assuming 128kbit/s!
AUDIO: 48000 Hz, 2 ch, s16le, 128.0 kbit/8.33% (ratio: 16000->192000)
ID_AUDIO_BITRATE=128000
ID_AUDIO_RATE=48000
ID_AUDIO_NCH=2
Selected audio codec: [faad] afm: faad (FAAD AAC (MPEG-2/MPEG-4 Audio) decoder)
==========================================================================
AO: [dsound] 48000Hz 2ch s16le (2 bytes per sample)
ID_AUDIO_CODEC=faad
Video: no video
Starting playback...
Exiting... (End of file)

[Peuj]

Works for me.
MPC(WMP11)+Haali+latest ffdshow.

Yes this one works well for me and for wyrd too (I think)

and I use too: MPC(WMP11)+Haali+latest ffdshow

[kikker]

With the recent builds (I'm running 973) and VSfilter *OR* Vobsub and not at the same time, WMP11 just sits there if there are subtitles in the same folder as the AVI. In fact VistaMC just crashes altogether claiming the culprit is either VSfilter or Vobsub.

I was scratching my head over that one for a while until I rememberd the 'subtitles' section in ffdshow. If I check that off, things seem to be normal. Strange thing is that I never had to check that off before. :o

[_xxl]

I was scratching my head over that one for a while until I rememberd the 'subtitles' section in ffdshow. If I check that off, things seem to be normal. Strange thing is that I never had to check that off before.
Any samples?

[zambelli]

and in the ffdshow audio there are:
WMA 7
WMA 8/9
Actually, WMA7, WMA8 and WMA9 are all the same codec = WMA v2. The older v1 was used in products that I don't think even carried the Windows Media name, like MSN Audio or something. There really isn't a v3, unless you count Pro and Lossless as v3, but nobody does that.

[Anima123]

I'm using drevil_xxl's build. Since rev978, there's compatible issue with CrystalPlayer. Cannot configure ffdshow within player and thus the output color space wont be YV12. Besides, the picture is upside down.

The last working build is rev964.

Edit: Just tried with success under Administrator. While I'm working under another restricted account, could that be the problem? How could I fix it?

[Keepitsimple]

Is there a wikipage or something which lists what the different options in ffdshow do?

The number of things you can do is almost overwhelming without a decent documentation.

[foxyshadis]

I'm not very versed on how inno and nsis work, so I'll leave it to you guys. But I found out from akupenguin that the default keyframe distance in FFV1 is way, way too high, and should be reset to 2-10 or so. 2 would probably be fine (4% larger), if you want to know how much loss in involved vs 300 frames:

1 - 7%
2 - 4%
3 - 3%
5 - 2%
10 - 1%

If you could set it to force a lower value if upgrading from an older version, that'd be cool. If anyone really needs it, I doubt it's a major inconvenience to reset it.

[haruhiko_yamagata]

I'm using drevil_xxl's build. Since rev978, there's compatible issue with CrystalPlayer. Cannot configure ffdshow within player and thus the output color space wont be YV12. Besides, the picture is upside down.

The last working build is rev964.

Edit: Just tried with success under Administrator. While I'm working under another restricted account, could that be the problem? How could I fix it?
Do you see ffdshow-icons in the task tray?
What is the executable filename for your crystalplayer?
Does unchecking "Use ffdshow only in:" help?

[cc979]

I'm using drevil_xxl's build. Since rev978, there's compatible issue with CrystalPlayer. Cannot configure ffdshow within player and thus the output color space wont be YV12. Besides, the picture is upside down.

The last working build is rev964.

Edit: Just tried with success under Administrator. While I'm working under another restricted account, could that be the problem? How could I fix it?

i thought crystal-player had its own codecs/filters

[Delerue]

i thought crystal-player had its own codecs/filters

MPEG 1 and 2 are build inside the player, but xVid, DivX and H.264/AVC don't. I don't know about FLV (can't open), and WMV I'm not sure, but I think that it uses external codec.

BTW, this player has an amazing option called 'motion morphing' that is similar to the TrimensionDNM (from WinDVD), but unfortunately still has some bugs. :(

----------

Revision 981
srt subtitle:fix sticky italic

Oh, yeah, baby. Thanks!
;)

[Dark_Angel_PT]

Can anyone tell me if the branch (beta 1) is being updated with bugs found and fixed in the trunk? Like the srt italics one?

[foxyshadis]

Not until Beta 2.

Thanks drevil, checking it out. =D

[Inventive Software]

Regarding your keyframes in FFV1 foxy, isn't that cause it doesn't really have delta-frames implemented as efficiently as, say, MSU?

[JarrettH]

what's in store for rev1000?:D

[Inventive Software]

Hopefully beta2, it's been long enough... ;)

[foxyshadis]

Well, akupenguin explained to me that it doesn't use delta frames at all in the classic sense. It only keeps context (dictionaries) around for future frames, sort of like rar solid archiving. He had a patch for delta frames that sounds cool and majorly effective, but it was never finished.

I was extremely surprised, actually, because I thought it was simply impossible to get that level of performance - the most efficient lossless video codec I've used - without advanced inter prediction. But then again, it only gives marginal gains to MSU as well, about 4% on my test case. (ending up about the same as ffv1 with 3 kf dist, and w/o it's a bit larger than ffv1 pure intra) Maybe MSU doesn't use deltas either, just context.

It's probably only really effective with true duplicate frames or 90% duplicates.

[akupenguin]

Maybe MSU doesn't use deltas either, just context.
MSU has a "Motion Estimation" tab. They wouldn't call it that if it didn't use conventional delta frames.

[foxyshadis]

This was brought to my attention by Fransiskus Febrianto in an email: http://sourceforge.net/tracker/index.php?func=detail&aid=1673817&group_id=173941&atid=867360

A strtrim() function doesn't exist in straight C/C++, sadly, but it exists in several alternative string libraries or the linux kernel, if anyone wants to pull a known good one out.

[WRFan]

ffdshow sets wrong merit for mpeg1 audio, I hate this decoder. I am using Vista x64 and there are no x64 merit changers available. Had to change the merits manually, both for MS Audio Mpeg decoder and ffdshow audio decoder. Why the hell are they doing this? Nobody would use this decoder unless absolutely necessary. I only installed it, because I couldn't find a x64 version of the intervideo decoders I usually use. I don't need it to decode mpeg1, because MS already provides a decoder for this purpose.

There is not a single x64 dx decoder for dvd available, so one would thing the ffdshow developers would invest their time in adding this function. Instead, they develop an mpeg1 decoder? which nobody needs! and then forcing people to use it?! how ugly is that?!!

To change the merits, go to

HKEY_CLASSES_ROOT\CLSID\{083863F1-70DE-11D0-BD40-00A0C911CE86}\Instance\{0F40E1E5-4F79-4988-B1A9-CC98794E6B55}

and

HKEY_CLASSES_ROOT\CLSID\{083863F1-70DE-11D0-BD40-00A0C911CE86}\Instance\{4A2286E0-7BEF-11CE-9BD9-0000E202599C}

the first one is the ffdshow crappy decoder, the second is the good old MS one. the merit starts (or rather ends, it's the other way around for some reason) with the fifth hex byte, set 00,00,60,00 for ffdshow and 00,00,80,00 for MS audio. or rather, you know what? set the ffdshow to 0x200001, that's just 1 above the do_not_use value. 'cause this decoder is as crappy as the mainconcept decoders that can't decode anything, but set the highest merit possible.

If someone knows of a x64 dx dvd decoder, please tell me. Can't even watch my favorite movies on x64 version of WMP, what a shame

[_xxl]

the first one is the ffdshow crappy decoder, the second is the good old MS one. the merit starts (or rather ends, it's the other way around for some reason) with the fifth hex byte, set 00,00,60,00 for ffdshow and 00,00,80,00 for MS audio. or rather, you know what? set the ffdshow to 0x200001, that's just 1 above the do_not_use value. 'cause this decoder is as crappy as the mainconcept decoders that can't decode anything, but set the highest merit possible.

It seems to me that you know everything better.

[Leak]

ffdshow sets wrong merit for mpeg1 audio, I hate this decoder.
Whoa there - did ffdshow's audio decoder run over your dog or something? :confused:

Instead of flaming about how ffdshow "sets the wrong merit" you could just have opened it's audio config dialog, gone to the "Codecs" page and just disabled the much hated "MP1/MP2" decoder or chosen to use libmad instead of mp3lib if it's quality is that bad.

Please do some research next time instead of making yourself look silly here...

[WRFan]

Instead of flaming about how ffdshow "sets the wrong merit" you could just have opened it's audio config dialog, gone to the "Codecs" page and just disabled the much hated "MP1/MP2" decoder or chosen to use libmad instead of mp3lib if it's quality is that bad.


I can't disable the mp2 decoder, because that's the only reason I installed this thing. the problem is that it's coupled with mp1 decoder for some reason, so you can't disable one without disabling the other one at the same time.

and yes, I tried the other mpeg2 decoder too, the libmad thingy, so what? ffdshow still decodes mpeg1, unless you change the merit. it's not even so much about ffdshow, but I've witnessed too many decoders in my life which are completely useless and still set high merits. those bundled with soundforge, the mainconcept crap, are a good example. please, just let the user decide! the developers always think they know better than the consumers.

[Leak]

In other things... it looks like I finally have succeeded to make an updated AviSynth filter that allows AviSynth some limited access to frames other than the current one (by buffering a specified amount of frames ahead and back) *and* makes it possible to actually drop or add frames (as long as it does so uniformly across the whole clip).

I.e. you can use stuff like TIVTC (which was my main motivation), Decomb, temporal filters and such on-the-fly... :D

I've also added an option to actually honor the interlace flags set by the decoder and apply 3:2 pulldown to the video before handing the frames to AviSynth - that helps a lot with smoothness if the flags on the video are horrible and you IVTC it yourself instead.

Of course, since my current code is mostly frankensteined together I'll need to go over it again to make a submittable patch and clean up some other stuff, and I still have to add the new options to the config dialog.

Anyway, if the patch is ready to go, who/where do I send it to? Or should I post it to the patches tracker on Sourceforge?

Also - is there a way to detect that the end of the stream has been hit? I'd need to deliver all samples still in the buffer at that point, as currently those silently get dropped - then again, dropping ~5 frames at the end of an episode or a movie probably is peanuts if everything else was properly IVTCed... :)

[Leak]

I can't disable the mp2 decoder, because that's the only reason I installed this thing. the problem is that it's coupled with mp1 decoder for some reason, so you can't disable one without disabling the other one at the same time.
Well, how about just asking to uncouple the MP1 and MP2 decoders on the config page then?

Also, the tone of your first post above was just rude - what was that meant to accomplish?

[fastplayer]

Also, the tone of your first post above was just rude - what was that meant to accomplish?
Tone? Those were just the joys of being an early adopter. 5 years too early. At least... :D

[Bathrone]

Ok, testing build revision 993, drevil compile, I always after install select the run audio and video config windows.

There was a crash as I moved onto the info and debug section of the video config app , and vista 32bit caught the following:

Product
Windows host process (Rundll32)

Problem
Stopped working

Date
5/03/2007 6:39 PM

Status
More information available

Problem signature
Problem Event Name: APPCRASH
Application Name: rundll32.exe
Application Version: 6.0.6000.16386
Application Timestamp: 4549b0e1
Fault Module Name: ffdshow.ax
Fault Module Version: 1.0.3.989
Fault Module Timestamp: 45eb1715
Exception Code: c0000005
Exception Offset: 000f0660
OS Version: 6.0.6000.2.0.0.256.1
Locale ID: 3081
Additional Information 1: cae0
Additional Information 2: c6b143c661515d69d4d8c4d358f23b56
Additional Information 3: 70ba
Additional Information 4: 86a876815086556d3329ef640b2e521b

Extra information about the problem
Bucket ID: 390435836

[_xxl]

Exception Code: c0000005
https://sourceforge.net/tracker/index.php?func=detail&aid=1670119&group_id=173941&atid=867360

[foxyshadis]

Leak, that sounds amazing. I gave it an hour and sort of gave up, I'm so glad you have more stamina. ;) Attaching a patch to the tracker and making a post linking to it here would probably be the best way.

[Leak]

Leak, that sounds amazing. I gave it an hour and sort of gave up, I'm so glad you have more stamina. ;) Attaching a patch to the tracker and making a post linking to it here would probably be the best way.
Roger that. :)

Gimme a day or two or so for the aforementioned clean-ups and interface changes, though.

And then there's that dentist appointment I'm off to in a few minutes... *ouch*

[haruhiko_yamagata]

Ok, testing build revision 993, drevil compile, I always after install select the run audio and video config windows.

Please test rev 994 or later, if it is fixed.

[clsid]

Subtitle suggestion:

Currently a user must choose between "Search in" and "File". I prefer (heuristic) search, but occasionally I manually need to select the correct file (from "subtitle sources"). Problem is that ffdshow remembers that setting. So next time I use it with a different video, it will use the wrong subs.

My suggestion is to remove the radio buttons. ffdshow should always use the Search method. A user can override the selected sub when needed by using the File field. That sub should then be used during the active session only.

[fastplayer]

Problem is that ffdshow remembers that setting.
Good idea!
It drove me nuts one time until I "figured out" to set ffdshow back to search for subtitles...

[molitar]

Well I am trying to use a software called uICE to send keyboard messages to the window to activate OSD. But for the life of me I can't get it to work at all. OSD works fine if I manually type in the keys but when I try to use the app uICE to send keys to it it does not register.

uICE can be downloaded from here... http://www.mediatexx.com/ it's only $20 vs $100 for Proximis Girder so I really would like some help on figuring out how to make it work with ffdshow so I can use this one. It also works 100x better with my serial port X10 remote while Proximis Girder will keep messing up the remote commands.

[500 Kip]

I get the following error when trying to launch the "Audio Decoder Configuration" from the start menu.
http://www.rmruk.com/pdf/ffdshow_audio.png
I'm using "ffdshow_rev1004_20070306_xxl.exe"

[Peuj]

I get the following error when trying to launch the "Audio Decoder Configuration" from the start menu.
http://www.rmruk.com/pdf/ffdshow_audio.png
I'm using "ffdshow_rev1004_20070306_xxl.exe"

I don't reproduce the problem with the ffdshow_rev1004_20070306_xxl.exe

[haruhiko_yamagata]

I get the following error when trying to launch the "Audio Decoder Configuration" from the start menu.
http://www.rmruk.com/pdf/ffdshow_audio.png
I'm using "ffdshow_rev1004_20070306_xxl.exe"
Are there any body who can reproduce this?
If not, the file (your local copy) may be broken.

[Episode]

Sorry, I cannot reproduce it either. Must be broken file or incomplete installation.

[Leak]

Gimme a day or two or so for the aforementioned clean-ups and interface changes, though.
Okay... it's probably gonna take a wee bit longer yet, as I hadn't taken into account that static DVD menus don't like simply being buffered for a certain number of frames because they'll never get displayed due to there being no notification that the remaining frames need to be flushed.

That's why I had to make a small modification to tvideocodeclibmpeg2.cpp/.h as well to signal the end of an MPEG2 sequence inside the picture's fieldtype flags. If that flag is encountered, the AviSynth filter will now flush all buffered frames and then reset the AviSynth environment, which is a bit ugly but also about the only way to make sure nothing is left in AviSynth's cache...

Oh yeah - code for applying 3:2 pulldown according to the TFF/BFF flags was also neccessary in turn to fix badly flagged DVDs with TIVTC.

So what's left is extending the UI to make the filter configurable, making a patch, testing it against a recent SVN checkout and submitting it, but I'll leave that for tomorrow. *yawn*

Oh yeah, another thing - while working with ffdshow as an MPEG2 decoder I noticed a total absence of DVD subtitles; am I right to assume that decoding/displaying of those is not implemented yet, or did I forget to throw a switch somewhere in the config dialogs?

np: Boards Of Canada - Aquarius (Version 3) (Peel Session)

[MacAddict]

Leak, thanks for your efforts in the cleanup. For me, watching a DVD with subtitles just requires me to right click in the window and select the subtitle from the sub-menu. Is this what you were looking for? Now that I think about it that might be MPC's sub engine :-\

[HeadBangeR77]

Leak, thanks for your efforts in the cleanup. For me, watching a DVD with subtitles just requires me to right click in the window and select the subtitle from the sub-menu. Is this what you were looking for? Now that I think about it that might be MPC's sub engine :-\

I can only second the first sentence. :)

As to MPEG-2 subtitles, there are none ... :( MPC's internal engine renders them in VMR9 and VMR7 renderless modes, which I don't use (VMR7 windowed as a rule). A fellow like me is derived of all possibilities of having DVD subtitles - MPC's splitter indicates their presence, but they cannot be rendered - neither while using MPC's MPEG-2 decoder (works better) nor ffdshow's one (works a bit worse, though has improved much recently).

If ffdshow's MPEG-2 decoding improved, then I could use VMR9 renderless + HQ RGB32. Internal ffdshow's support would be nice, though. ;)

@ clsid
Thanks a bunch for the latest 1006 ICL9 build! :)

@ h_yamagata
I could praise your name for working on ffdshow's subtitles (and albain's, whoever he is) ;)

[Keepitsimple]

whats wrong with ffdshows mpeg2 decoding?

[HeadBangeR77]

whats wrong with ffdshows mpeg2 decoding?
By simple watching, if you don't touch the keyboard, absolutely nothing. Artefacts (usually some macroblocks) appear while seeking/skipping + you've got to wait a while (from 1 sec to eternity ;)). I haven't got such problems while using the built-in MPC's decoding (ironically, both come from the same source i.e. ffmpeg).

[haruhiko_yamagata]

By simple watching, if you don't touch the keyboard, absolutely nothing. Artefacts (usually some macroblocks) appear while seeking/skipping + you've got to wait a while (from 1 sec to eternity ;)). I haven't got such problems while using the built-in MPC's decoding (ironically, both come from the same source i.e. ffmpeg).
As for DVD subtitles, I have not tested much.
But the artifacts have been fixed at rev 750 (libmpeg2).
Is it still noticeable?

Oh, yes, libavcodec mpeg2 decoder has such issue, please use libmpeg2.

[_xxl]

haven't got such problems while using the built-in MPC's decoding ironically, both come from the same source i.e. ffmpeg
MPC uses libmpeg2 not ffmpeg.

[madshi]

Hey guys,

I'm trying to get an instance of the ffdShow Audio Processor in my little Delphi program. But I'm desperately failing. This returns "class not registered":

const ffdShowAudioProcessor : TGuid = '{B86F6BEE-E7C0-4D03-8D52-5B4430CF6C88}';

CoCreateInstance(ffdShowAudioProcessor, nil, CLSCTX_ALL, IID_IBaseFilter, ffdShow);
The very same code works for every other DShow filter I've tried - but not for ffdShow. This code works:

CoCreateInstance(ffdShowAudioProcessor, nil, CLSCTX_ALL, IID_IUnknown, ffdShow);
However, trying to get an IBaseFilter interface (by using QueryInterface) again fails.

Can anybody help, please?

[HeadBangeR77]

As for DVD subtitles, I have not tested much.
But the artifacts have been fixed at rev 750 (libmpeg2).
Is it still noticeable?
I know it has really improved much somewhere after ICL9 743 (which still had many artefacts), as I wrote above. I thought everything was ok, yet after 10 minutes of seeking and jumping forwards and backwards I noticed some macroblocks (occasionally). Seeking wasn't perfect btw. (tried many DVDs, and my both optical drives + my notebook).

Oh, yes, libavcodec mpeg2 decoder has such issue, please use libmpeg2.
I've been using libmpeg2 all the time back from the beginning, so it's libmpeg2 (ffdshow) vs. libmpeg2 (MPC) duel. ;)

[DeepBeepMeep]

It seems there is a compatibility problem with Haali Filter and Ffdshow tryouts. The Video is accelerated at 29.97 fps instead of staying at 23.976 and Audio / Video gets out sync
when an EVO with VC1 is played.

It occurs if Ffdshow is added in the DirectShow chain that is:
- if Ffdshow is the VC1 decoder with either choice of codec
- if WMVVideo is the VC1 decoder and is followed by Ffdshow in the DirectShow chain

If WMVVideo is used alone there isn't any audio / video sync problem.

My assumption is that Telecine flags such as "Repeat field" are ignored by the the Ffshow filter.

[Bathrone]

Please test rev 994 or later, if it is fixed.

Thanks mate, confirmed fixed

[Leak]

As for DVD subtitles, I have not tested much.
Well, I'm not getting *any* subtitles using ffdshow; neither normal ones nor mouseover menu highlights - so does anyone know if there is a decoder/filter to handle them in there or not?

[Leak]

So what's left is extending the UI to make the filter configurable, making a patch, testing it against a recent SVN checkout and submitting it, but I'll leave that for tomorrow. *yawn*
Well, it's tomorrow, it's late at night *again* :D and the patch is finally submitted (http://sourceforge.net/tracker/index.php?func=detail&aid=1676882&group_id=173941&atid=867362)... go ahead and go nuts with it. :)

(Sorry, I don't have an installer for it, as I haven't looked into building one and I don't think we really need yet another build variant...)

np: Boards Of Canada - Poppy Seed (So Soon)

[vlada]

Hello,
if somebody's working on DVD subtitles, I would really appreciate if you could add support for 4:3 (letterbox) subtitles. I mentioned the problem with subtitles on a 4:3 monitor/TV in this thread (http://forum.doom9.org/showthread.php?t=110830). I'm still looking for a solution on this issue.

[_xxl]

It seems there is a compatibility problem with Haali Filter and Ffdshow tryouts. The Video is accelerated at 29.97 fps instead of staying at 23.976 and Audio / Video gets out sync when an EVO with VC1 is played.
It occurs if Ffdshow is added in the DirectShow chain that is:
- if Ffdshow is the VC1 decoder with either choice of codec
- if WMVVideo is the VC1 decoder and is followed by Ffdshow in the DirectShow chain
If WMVVideo is used alone there isn't any audio / video sync problem.
Confirmed.

[igor1st]

Well, it's tomorrow, it's late at night *again* :D and the patch is finally submitted[/url]... go ahead and go nuts with it. :)

Tested. Works great for me.

[Leak]

Tested. Works great for me.
Yay. :D

I guess I'll give building an installer a try when I get home; for kicks and easier testing... :)

Okay, making an installer turned out a bit easier than I feared so here is one:

ffdshow rev. 1007 + AviSynth patch installer (http://leak.no-ip.org/AviSynth/ffdshow_AviSynth_rev1007_20070309.exe) (pure MSVC8+GCC 3.4.4 compile)

[Leak]

*crickets chirping*

So, um... any other comments on the above? :)

[Rectal Prolapse]

Wow Leak - I was wondering why I kept getting stuttering video whenever I attempted to use Tdeint() in my AVISynth scripts in FFDShow - I guess your patches fixes this? What needs to be configured to enable this extra frame buffering?

Thanks!

[Leak]

Wow Leak - I was wondering why I kept getting stuttering video whenever I attempted to use Tdeint() in my AVISynth scripts in FFDShow - I guess your patches fixes this?
Quite possibly. The original problem was that no matter which frame AviSynth requested from ffdshow, it would always be handed the same old frame that had just been decoded - which is what the changes in my patch are about.

Here's the documentation I posted alongside the patch at SourceForge:

Here's my patch for ffdshow's AviSynth filter that implements buffering to allow AviSynth access to more than just the current frame. It also allows an AviSynth script to adjust the number of frames, as long as this happens evenly distributed across the whole video.

I've added 4 new controls to the AviSynth page:

* Buffer back: number of previous frames to keep around, i.e. the number of frames AviSynth can look back.

* Buffer current: a top limit for the number of frames that will be buffered for a single output frame; the actual value for this is calculated by dividing the number of frames going in by the number of frames going out - but if your video is a million frames and you request only one, without a limit one million frames would be buffered, which would probably just crash the player... :)

* Buffer ahead: number of frames to buffer ahead before feeding anything to AviSynth, i.e. the number of frames AviSynth can look ahead.

* Apply pulldown: applies a 3:2 pulldown to the video according to the interlace flags that the frames coming in have. Should be activated for IVTC, bobbing and other stuff that drops or adds frames. Deactivate it if you want progressive frames straight from a DVD, unless it was flagged incorrectly.

If you get jerky video, that most of the time means the buffering values are to small.

For example, the following script needs to buffer 1 frame back and 10 ahead:

TFM(order=1)
TDecimate(mode=1,hybrid=1)

The default 10/10/10 is probably a good compromise; and keeping more back buffers only uses more memory but doesn't slow things down. Also, you should probably use something like SetMemoryMax(16) in your script to limit the amount of memory AviSynth uses.

So in your case, where Tdeint probably checks just one frame back and one forward you could just use a setting of 1/10/1 (no need to tweak the middle value unless you're dropping over 90% of all fields, which is a bit adventurous for playback to say the least), but it should work just the same with 10/10/10 - it'll only take a bit longer to start playing as 11 frames have to be buffered first.

Also, you should probably use ffdshow's libmpeg2 for DVD decoding or at least turn AviSynth off while you're working your way through the DVDs menus, as without my changes to libmpeg2 static single-frame menus won't get displayed most of the time, as the AviSynth filter then can't know that the video had ended before enough frames could be buffered.

(Of course, that'll currently lose you the support for subtitles, so you might want to set up a hotkey for the AviSynth filter in ffdshow...)

np: Boards Of Canada - Hey Saturday Sun (The Campfire Headphase)

[haruhiko_yamagata]

@leak
Thank you for the patch.
I have some issue with it.
When I try to play avi files (any files, libavcodec or Xvid), I have blackscreen depending on video renderers. I used MPC on WindowsXp sp2 for testing.
OK with old renderer, VMR9/VMR9 renderless, VMR7 renderless
NG with VMR7/overlaymixer

The script(one line): FadeIn(30)

It is OK when I stated playing with the check (AviSynth) off and enabled after starting.
Some DirectShow filters do not switch from pause to playing untill the first frame is delivered, thus we can't receive the second frame. If it helps.

Please supply a sample script that show the benefit.

[Rectal Prolapse]

Thanks for the description leak! I guess it would be very tricky getting this to work with DScaler5 without some fiddling with hotkeys and presets. Hmmmm. :)

[Leak]

@leak
Thank you for the patch.
I have some issue with it.
When I try to play avi files (any files, libavcodec or Xvid), I have blackscreen depending on video renderers. I used MPC on WindowsXp sp2 for testing.
OK with old renderer, VMR9/VMR9 renderless, VMR7 renderless
NG with VMR7/overlaymixer

The script(one line): FadeIn(30)

It is OK when I stated playing with the check (AviSynth) off and enabled after starting.
Some DirectShow filters do not switch from pause to playing untill the first frame is delivered, thus we can't receive the second frame. If it helps.
*goes to test this*

Ummm... strangely enough, it works for me here, using a script of "FadeIn(240)". At least, it does when playing a DVD - what kind of video were you using?

I'm running Windows XP SP2 with an ATI Radeon X1900 XT here.

Wait, scratch that - it works when playing a DVD, but not when I play an AVI file.

I had only tested it with VMR9. I guess just passing the first frame received through as-is should fix that... I gotta go shopping now, but I'll take a stab at it when I get home.

On a side note, I do have noticed that when I'm using the Overlay Mixer MPC doesn't seem to switch from 4:3 to 16:9 and back when playing a DVD, thus cutting off the image at the right side, but that happens whether the AviSynth filter is turned on or off - so I don't think it has anything to do with my patch.

EDIT: The same thing happens with ffdshow_rev1008_20070308_clsid.exe, so it's definitely not caused by my code...

To reproduce: I run MPC, set it to "Zoom 100%" and Overlay Mixer, then I play the first US volume of Ghost In The Shell: Stand Alone Complex. The DVD starts, the 4:3 FBI notice comes up, fills the window, then I hit "Title Menu", the content is now 16:9, but the window is still 4:3, cutting off part of the image's right side - which seems to be totally missing, as it doesn't appear when I make the video smaller. Checking "Keep Aspect Ratio" just squishes the image vertically, but doesn't help otherwise.

Everything works as it should with VMR7/9, though.

Please supply a sample script that show the benefit.
Do you mean the benefit of my patch?

The TFM/TDecimate combination in my last post would be such a script... as would be using just about any temporal filter, or any filter that drops/adds frames like Bob, SelectEvery and the like - but not Trim, as that would need real random access in the video stream.

Is that what you meant?

Just use

SetMemoryMax(16) #optional
TFM(order=1)
TDecimate(mode=1,hybrid=1)

in conjunction with your favourite telecined DVD and watch it un-telecined as it was meant to be seen... of course, having a display that's capable of displaying 48Hz or 72Hz helps a lot.

[Leak]

Thanks for the description leak! I guess it would be very tricky getting this to work with DScaler5 without some fiddling with hotkeys and presets. Hmmmm. :)
Yeah, well - the same goes for about every other option in ffdshow that needs to be set according to the type of video you want to watch... :)

np: Boards Of Canada - Twoism (Twoism)

[Hans Ohlo]

i have a strange problem. maybe this is vista related in how they now handle media internally in wmp11 and vista media center. i have videos (mkv: h264, ac-3, avi:xvid, ac-3) where the ac-3 stream changes from 2.0 to 5.1 after a couple of milliseconds at the beginning (around 300ms max.). in media player classic (using only external dshow filters as wmp11 and vme) and zoomplayer 5 everything plays fine using ffdshow to decode all the streams (xvid, h264, ac-3) and haali splitter. but in wmp11 and vme either the stream won't start playing or the sound vanishes right at the start. when i try to look at the infos in ffdshow while playing these files nothing is shown either. if i use ac3filter to decode the ac-3 streams this problem does not appear.

maybe anyone has an idea what goes wrong here?

[khagaroth]

I'm translating ffdshow to Czech, using the integrated translation mode. If I try to translate the tooltip for "Accept embedded subtitles" (by rightclicking on an empty area of the dialog) I get a data execution prevention error message from Windows DEP. This is the only error information it displays, probably not very useful:
EventType : BEX P1 : rundll32.exe P2 : 5.1.2600.2180 P3 : 41107dbc
P4 : unknown P5 : 0.0.0.0 P6 : 00000000 P7 : e877d394 P8 : c0000005 P9 : 00000008

[Leak]

I had only tested it with VMR9. I guess just passing the first frame received through as-is should fix that... I gotta go shopping now, but I'll take a stab at it when I get home.
Yeah, that was the ticket - I now just more or less pass the first frame through (that is, I'm requesting a far away frame from AviSynth with just one frame buffered; that will give about the same result as with the original, unpatched implementation) and that makes the graph tick even with the Overlay Renderer.

I've also made the filter reset AviSynth whenever it's disabled as it doesn't make much sense to work on frames that have been long gone when the filter is reactivated.

The patch can be found here again (http://sourceforge.net/tracker/index.php?func=detail&aid=1676882&group_id=173941&atid=867362) (along with some documentation about what the new options do and what the improvement in general is)
Updated installer that is based on rev. 1017 (http://leak.no-ip.org/AviSynth/ffdshow_AviSynth_rev1017_20070311.exe)

np: Boards of Canada - Hi Scores (Hi Scores)

[Eragon4ever]

I'm translating ffdshow to Czech, using the integrated translation mode. If I try to translate the tooltip for "Accept embedded subtitles" (by rightclicking on an empty area of the dialog) I get a data execution prevention error message from Windows DEP. This is the only error information it displays, probably not very useful:
EventType : BEX P1 : rundll32.exe P2 : 5.1.2600.2180 P3 : 41107dbc
P4 : unknown P5 : 0.0.0.0 P6 : 00000000 P7 : e877d394 P8 : c0000005 P9 : 00000008

I can confirm this. The JIT debuger of VS2005 tells me:
"Eine Ausnahme (erste Chance) bei 0x90909090 in rundll32.exe: 0xC0000005: Zugriffsverletzung beim Lesen an Position 0x90909090."
Which means something like:
"An exeption (first chance) at 0x90909090 in rundll32.exe: 0xC0000005: access violation during reading at position 0x90909090."

As I don't know how to use it I can't give you more information.

[Rectal Prolapse]

I am having problems with FLAC multichannel audio playback with FFDShow. Is there a known problem with this?

I have a FLAC 5.1 channel track inside an .MKA container (and also in an MKV container with video), and both give me the same audio glitches. Encoded at 16 bits. The original WAV file works fine when muxed into the containers.

FFDShow-tryouts version 964.

Thanks!

[Leak]

I've updated the build with my patch to SVN revision 1017 (http://leak.no-ip.org/AviSynth/ffdshow_AviSynth_rev1017_20070311.exe) as I noticed that the OSD tended to crash when you toggled a filter while a video was playing...

(Silly me was assuming that changes to the subtitle code would not affect me since I don't use them - but of course the OSD uses the same code... :D)

np: Boards Of Canada - Amo Bishop Roden (In A Beautiful Place Out In The Country)

[Frans]

I already posted this in ffdshow tryout page, but I thought maybe I should post it in here too. When I play an MKV file with embedded vobsub, the vobsub isn't being showed by ffdshow. If I use directvobsub, then everything is fine. Is it a bug or ffdshow just doesn't support embedded vobsub? can anyone confirm this?
And since lately there has been improvement on the subtitle part in ffdshow, is it possible to have ffdshow override the font size used in ssa sub? It would be better if the subtitle style and the font can be override too.
How about the status of whitespace-only line not treated as blank (the one being posted in ffdshow tryout page)? Is something like this easy or hard to fix?
This is my first time posting in this forum and I would like to say thanks to everyone who are working to improve the ffdshow. Keep up the good work! :)

[vlada]

The only application I know where I can build and try different DVD-Video graphs is MV2Player. Normally I use DScaler as video decoder and it looks like this:

http://img185.imageshack.us/img185/9743/dscalervk1.png

Everything works O.K.

Then I tried to build a DVD graph with ffdshow:

http://img95.imageshack.us/img95/8290/ffdshowrf7.png

With this graph I have no subtitles. FFDShow sees them (I can select embedded in tray icon menu), but the subtitles are not displayed.

Another strange thing is, that if I use ffdshow as postprocessing filter for DVD in MPC, letterboxing doesn't work. Why? What is so special on DVD playback, that there are so many difficulties to make it work?

[haruhiko_yamagata]

I know it has really improved much somewhere after ICL9 743 (which still had many artefacts), as I wrote above. I thought everything was ok, yet after 10 minutes of seeking and jumping forwards and backwards I noticed some macroblocks (occasionally). Seeking wasn't perfect btw. (tried many DVDs, and my both optical drives + my notebook).

Confirmed.
The DVD subtitle issue have been fixed at rev 1018.

[haruhiko_yamagata]

I already posted this in ffdshow tryout page, but I thought maybe I should post it in here too. When I play an MKV file with embedded vobsub, the vobsub isn't being showed by ffdshow. If I use directvobsub, then everything is fine. Is it a bug or ffdshow just doesn't support embedded vobsub? can anyone confirm this?
And since lately there has been improvement on the subtitle part in ffdshow, is it possible to have ffdshow override the font size used in ssa sub? It would be better if the subtitle style and the font can be override too.
How about the status of whitespace-only line not treated as blank (the one being posted in ffdshow tryout page)? Is something like this easy or hard to fix?
This is my first time posting in this forum and I would like to say thanks to everyone who are working to improve the ffdshow. Keep up the good work! :)
Thank you for your report, embedded vobsub issue have been fixed.
The srt whitespace-only line issue is not very easy because some of such lines must be rendered as a blank line, I'll fix it later (maybe after beta2). Basically whitespace-only line should not be used for a separater line, but there is no guide line for srt as far as I know.

[Leak]

Confirmed.
The DVD subtitle issue have been fixed at rev 1018.
Trying this with my Ghost In The Shell: Stand Alone Complex DVD, I finally get subtitles again :D, but a release build crashes in the DVD's main menu as soon as I click on "PLAY".

Building the release build with debugging info, I found that it crashed in src\subtitles\tspuimage.cpp, line 438. Disabling optimization for that file showed that the check in the first line of TspuImageSimd<_mm>:: ownprint isn't enough.

Changing the check at the method's beginning from

if (!plane[0].stride) return;

to the more thorough

if (!plane[0].stride || !plane[0].c || !plane[0].r) return;

fixes the crash for me. Somehow, both pointers in plane were null, but the stride was still 64...

BTW - have you tried my updated AviSynth patch?

np: Boards Of Canada - Zoetrope (In A Beautiful Place Out In The Country)

[Blinx123]

Hi guys,
I have a question. Is it possible to use whatever Version of FFDSHOW directly with PowerDVD Ultra 7.2 or WinDVD 8? I don´t want to install yet another program like Zoomplayer.

[Leak]

I guess I should have my builds hosted on SourceForge... :D

MSVC/GCC build based on rev. 1018, with my AviSynth patch applied and the DVD decoder crash fixed (http://leak.no-ip.org/AviSynth/ffdshow_AviSynth_rev1018_20070311.exe)

That (and AviSynth and TIVTC) is basically everything you need to do on-the-fly inverse-telecine for DVDs. :)

[Episode]

@Blinx123, or you could try Media Player Classic. It takes only about 4 megabytes disc space, has a very lightweight user interface and doesn't need installation since it's just one single file. PowerDVD and WinDVD won't work correctly (if at all) with ffdshow :-/

[Blinx123]

Ok,so PowerDVD and WinDVD won´t work on the fly with FFDSHOW. Maybe I´ll get Zoomplayer then.

Isn´t Media Player Classic outdated? I can´t play DVDs with it,can I?

[HeadBangeR77]

Confirmed.
The DVD subtitle issue have been fixed at rev 1018.
Thanks very much :) - gonna have a look at that new revision soon.

:thanks:

[Leak]

Isn´t Media Player Classic outdated? I can´t play DVDs with it,can I?
Huh? :confused:

Playing DVDs (and the build-in MPEG2 decoder) is one of MPC's main features...

As of the last revision (modulo that bug I discovered and fixed since), ffdshow is also very much capable of decoding DVDs and DVD subtitles again...

Just try it out - it's not as if MPC were a big download or anything.

[vlada]

With rev 1018 DVD subtitles work, but without letterboxing :-(

Another problem is, that if I resize or move video window, the movie gets cropped from right side. Unfortunately I don't know how to take a screenshot of overlay window and if don't use overlay, I get just a black screen.

This problem only appears with 16:9 movies, 4:3 is O.K.

[Px]

I'm translating ffdshow to Czech, using the integrated translation mode. If I try to translate the tooltip for "Accept embedded subtitles" (by rightclicking on an empty area of the dialog) I get a data execution prevention error message from Windows DEP. This is the only error information it displays, probably not very useful:
EventType : BEX P1 : rundll32.exe P2 : 5.1.2600.2180 P3 : 41107dbc
P4 : unknown P5 : 0.0.0.0 P6 : 00000000 P7 : e877d394 P8 : c0000005 P9 : 00000008
More information you can find in Dr.Watson log (Start - Run - drwtsn32)

[Delerue]

Something old that I always forgot to tell you: If I use 'resize', DVD subtitles and buttons highlight can't be align correctly. Tested with 1019 built by 'xxl'.

[WorBry]

Having archived more than 500GB of deinterlaced/denoised home video in FFDShow-HuffYuv (YV12) format, I've been looking of late at alternative lossless codecs with a higher compression ratio. FFV1 certainly fits the bill. However, whilst the encodes open OK in VDub, if I try and play-back in other media players (TCMP, MPC etc) they play for a few seconds and then disintegrate into a mosaic of coloured blocks before either freezing or repeating the same pattern. I'm currently using the latest clsid build (Rev 1008, 20070308) with the following FFV1 settings - AC, Large Content, KF interval 10 (default) - but the same playback problem results no matter what settings I use. Searching the forum I see that this is/was a known problem, at least with older (2003) builds:

http://forum.doom9.org/showthread.php?p=400588#post400588

Could anyone point me to a recent (FFV1 stable) FFDShow build that does not suffer this problem or advise a remedy (other than use another LC).

Many thanks.

[Leak]

Something old that I always forgot to tell you: If I use 'resize', DVD subtitles and buttons highlight can't be align correctly. Tested with 1019 built by 'xxl'.
You might want to drag the subtitle filter above the resize filter in the filter list in this case so the subtitles get etched into the image before resizing.

[Delerue]

You might want to drag the subtitle filter above the resize filter in the filter list in this case so the subtitles get etched into the image before resizing.

Hmmm... I thought MPC was handling the subtitles. Thanks, man. ;)

[haruhiko_yamagata]

Trying this with my Ghost In The Shell: Stand Alone Complex DVD, I finally get subtitles again :D, but a release build crashes in the DVD's main menu as soon as I click on "PLAY".

Building the release build with debugging info, I found that it crashed in src\subtitles\tspuimage.cpp, line 438. Disabling optimization for that file showed that the check in the first line of TspuImageSimd<_mm>:: ownprint isn't enough.

Changing the check at the method's beginning from

if (!plane[0].stride) return;

to the more thorough

if (!plane[0].stride || !plane[0].c || !plane[0].r) return;

fixes the crash for me. Somehow, both pointers in plane were null, but the stride was still 64...

BTW - have you tried my updated AviSynth patch?

np: Boards Of Canada - Zoetrope (In A Beautiful Place Out In The Country)
Thank you for the DVD subtitle fix. I can reproduce the problem and your code fixes it. I'll commit your fix soon.

As for AviSynth, avi file starts palying.
I have a flash on seek.
frames 1 2 3 4 5 6 7 8 9
When I seek from 7 to 2, the frame is like 2-8(7?)-3.

[Leak]

Thank you for the DVD subtitle fix. I can reproduce the problem and your code fixes it. I'll commit your fix soon.
I was wondering though why the stride was not set to zero in this case - there's probably some other bug in there somewhere...

But of course checking if the two pointers are zero before using them is good practice anyway.
As for AviSynth, avi file starts palying.
I have a flash on seek.
frames 1 2 3 4 5 6 7 8 9
When I seek from 7 to 2, the frame is like 2-8(7?)-3.
I'll have a look into it when I get home - probably something to do with the first frame getting passed through as is. Which script and which settings did you use?

By the way - would it be possible to give me write access to the SVN repository? I'm pretty sure there'll be corrections or optimizations to make if this patch is checked in, and I was thinking of making an internal ffdshow filter out of TIVTC with a "proper" config dialog...

[B.F.]

About bugs.
WMVA(the same as WVC1) not supported.
URL http://samples.mplayerhq.hu/V-codecs/WMVA/wmva-smpl-bdsm.wmv
Postprocessing in Mpeg 1&2 files doesn't work if libmpeg2 support enabled (in libavcodec all OK)

[haruhiko_yamagata]

I'll have a look into it when I get home - probably something to do with the first frame getting passed through as is. Which script and which settings did you use?
The script.
SetMemoryMax(16) #optional
TFM(order=1)
TDecimate(mode=1,hybrid=1)

The inputing file is progressive (just for testing), DX50, AVI.

By the way - would it be possible to give me write access to the SVN repository? I'm pretty sure there'll be corrections or optimizations to make if this patch is checked in, and I was thinking of making an internal ffdshow filter out of TIVTC with a "proper" config dialog...
Please wait, at least after beta2. Your work is great, advanced and complicated, needs more testing.

[_xxl]

WMVA(the same as WVC1) not supported
WMVA is crashing using wmv9lib.WMVA is not supported.
Postprocessing in Mpeg 1&2 files doesn't work if libmpeg2 support enabled (in libavcodec all OK)
I'll fix it.

[Leak]

Please wait, at least after beta2. Your work is great, advanced and complicated, needs more testing.
Okay, until then I'll just continue putting up my own builds now and then for testing. I'm also thinking of making my version act exactly like the old version (i.e. no buffering at all) when the buffers are set to 0/1/0 - I guess that should ease the transition.

Is there a timeframe for a second beta on the horizon, if I may ask?

[clsid]

Having archived more than 500GB of deinterlaced/denoised home video in FFDShow-HuffYuv (YV12) format, I've been looking of late at alternative lossless codecs with a higher compression ratio. FFV1 certainly fits the bill. However, whilst the encodes open OK in VDub, if I try and play-back in other media players (TCMP, MPC etc) they play for a few seconds and then disintegrate into a mosaic of coloured blocks before either freezing or repeating the same pattern. I'm currently using the latest clsid build (Rev 1008, 20070308) with the following FFV1 settings - AC, Large Content, KF interval 10 (default) - but the same playback problem results no matter what settings I use. Searching the forum I see that this is/was a known problem, at least with older (2003) builds:

http://forum.doom9.org/showthread.php?p=400588#post400588

Could anyone point me to a recent (FFV1 stable) FFDShow build that does not suffer this problem or advise a remedy (other than use another LC).

Many thanks.
Could you upload a sample file that doesn't play correctly?

I re-encoded a small clip with default settings and it played back ok.

[haruhiko_yamagata]

Is there a timeframe for a second beta on the horizon, if I may ask?
I'm messed up with subtitle and some other bugs, but as soon as possible :D .
Recent changes have to be tested a little more.

[Hans Ohlo]

i saw on the ffmpeg devel list that some patches for the ac-3 decoder were made. are these allready included in the latest builds?

maybe these fix my problem i described...

[_xxl]

ffdshow doesn't use ffmpeg to decode AC3 only to encode.

[CiNcH]

Hope I am allowed to post bugs here...

I am using ffdshow tryouts beta 1 (from clsid) as a raw video post processor in a DVB filter graph (using CyberLink PDVD 7 as MPEG-2 decoder).
Problem now is, that when I switch to a channel with different resolution (e.g. from channel with 544 x 576 to channel 720 x 576) the picture is suddenly displayed twice, one by one (at channel change, graph is not being rebuilt, so resolution change is on-the-fly). Only rebuilding the filter graph helps.
Problem only occurs when using Overlay Mixer, no problems with VMR.

Probably it is just a configuration issue...

[Leak]

Probably it is just a configuration issue...
Well - have you tried a more recent build than beta 1? There have been quite a few changes in between those, after all...

[CiNcH]

Have tried yours (rev1018) but the filter somehow does not even register (some error in regsvr32).

[CiNcH]

Ok, problem solved with 1021.

Now I have the problem, that with anamorphic material, picture seems zoomed and with wrong aspect ratio (flat heads...), but again, only with Overlay Mixer.

[Inventive Software]

With the amount of bug fixes from beta1, would the devs like to put forward a beta2 build?

[sillKotscha]

With the amount of bug fixes from beta1, would the devs like to put forward a beta2 build?

already answered a few posts above ;)

I'm messed up with subtitle and some other bugs, but as soon as possible :D .
Recent changes have to be tested a little more.

[Atak_Snajpera]

FFDshow still uses only one core for decoding H264. 1080p movies with bitrate at 15mbit are killing my C2D@2.8Ghz

[Leak]

FFDshow still uses only one core for decoding H264. 1080p movies with bitrate at 15mbit are killing my C2D@2.8Ghz
Yeah, well - libavcodec (http://en.wikipedia.org/wiki/Libavcodec) still uses only one core for it. That's what ffdshow uses - you'd need to bring that up with the libavcodec developers (aka the mplayer team)...

np: Swod - Walz 57 (Gehen)

[Fizick]

Leak,
Extented Avisynth support is very interesting!
But I can not install any of your builds :(
Setup stop at 90 percent of line with message window:
Internal error: Expression error 'Runtime Error (at 93:412): RegSvr32 failed with exit code 0x3'.

I try use filemon - regmon to catch. may be it is with last lines:
regsvr32 C:\Program Files\ffdshow\ffdshow.ax.2.manifest FILE NOT FOUND
regsvr32 C:\Program Files\ffdshow\ffdshow.ax.2.config FILE NOT FOUND

AthlonXP, WinXP Home Russian SP2, MS VC2005 express, etc

generic clsid and xxl builds are installed OK. But seems they do not use VC80.
(do not install msvcr80.dll)

[WorBry]

Having archived more than 500GB of deinterlaced/denoised home video in FFDShow-HuffYuv (YV12) format, I've been looking of late at alternative lossless codecs with a higher compression ratio. FFV1 certainly fits the bill. However, whilst the encodes open OK in VDub, if I try and play-back in other media players (TCMP, MPC etc) they play for a few seconds and then disintegrate into a mosaic of coloured blocks before either freezing or repeating the same pattern. I'm currently using the latest clsid build (Rev 1008, 20070308) with the following FFV1 settings - AC, Large Content, KF interval 10 (default) - but the same playback problem results no matter what settings I use. Searching the forum I see that this is/was a known problem, at least with older (2003) builds:

http://forum.doom9.org/showthread.php?p=400588#post400588

Could anyone point me to a recent (FFV1 stable) FFDShow build that does not suffer this problem or advise a remedy (other than use another LC).

Many thanks.



Could you upload a sample file that doesn't play correctly?

I re-encoded a small clip with default settings and it played back ok.

Sure, but as I mentioned the 'problem' appears a few secs (4-5) into playback, so the samples are necessarily around 40-50MB

Here is the HuffYuv-YV12 source:

http://rapidshare.com/files/20713071/Sample_HuffYuv-YV12_source.avi

and here is the FFV1 (AC, Large Context, KF=10) encode:

http://rapidshare.com/files/20716143/Sample_FFV1_AC_Large_Context_KF10.avi

[clsid]

I can confirm the FFV1 bug.

[foxyshadis]

FFV1 bug was filed (http://sourceforge.net/tracker/index.php?func=detail&aid=1679220&group_id=173941&atid=867360) btw. So far it only happens with AC encodes.

[Leak]

generic clsid and xxl builds are installed OK. But seems they do not use VC80.
(do not install msvcr80.dll)
I'm afraid VC80 is all I have, and I'm using the same InnoSetup script everyone else seems to be using. I'll try installing it at work tomorrow and see what happens.

But I'm always open for hints on how to fix this...

(@Haruhiko: didn't you have a similar problem recently?)

[Fizick]

where is the Innosetup script?
I do not see it at
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/

[Leak]

where is the Innosetup script?
I do not see it at
http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/
Ah yes, that one's a bit non-obvious - it's here. (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/bin/distrib/InnoSetup/)

[sillKotscha]

I have noticed that libmad won't decode the sound of this clip (http://www.wikiupload.com/download_page.php?id=101744) (down right you'll find the download link)... mp3lib does work...

as a side note - even though this is a normal *.mpg file Haali's splitter is used as mpg demultiplexer - why??

I have other *.mpg files and as usual my Elecard MPEG Demultiplexer is used.

EDIT:

ffdshow is still (after your various fantastic bug hunting work) unable to decode the sound of that clip (http://www.wikiupload.com/download_page.php?id=101790)...

if I select the Quicktime components to decode the file everything is ok but if I select the DShow framework in MPC to decode the clip the sound is totally fu**ed up...

here you see the sound properties of the file:

http://img389.imageshack.us/img389/2452/001ta4.jpg (http://imageshack.us)

I use ffdshow_rev1021_20070312_xxl.exe and MPC's internal MP4/MOV splitter - with Haali's splitter I hear no sound at all and with Quicktime's movie parser I can hear the sound as it should

[haruhiko_yamagata]

I'm afraid VC80 is all I have, and I'm using the same InnoSetup script everyone else seems to be using. I'll try installing it at work tomorrow and see what happens.

But I'm always open for hints on how to fix this...

(@Haruhiko: didn't you have a similar problem recently?)You use MSVC8 SP1, don't you? In that case, you'll have to replace \trunk\bin\distrib\msvcr80.dll, \trunk\bin\distrib\InnoSetup\msvc80\ffdshow.ax.manifest and maybe microsoft.vc80.crt.manifest.

[Px]

FFDshow still uses only one core for decoding H264. 1080p movies with bitrate at 15mbit are killing my C2D@2.8Ghz
It's strange, because on same CPU I have problems only with streams with average bitrate 24 Mbps....
One another thing - ffdshow can use both cores, for that I use "tweak" with process affinity, I look for core, on which video player running, and set affinity to another core, then switch back, and finally, allows player to use both cores.

[cc979]

Ok, problem solved with 1021.

Now I have the problem, that with anamorphic material, picture seems zoomed and with wrong aspect ratio (flat heads...), but again, only with Overlay Mixer.

overlay anamorphic .avi bug is old, vmr9 or haali will play them correctly

[_xxl]

FFDshow still uses only one core for decoding H264. 1080p movies with bitrate at 15mbit are killing my C2D@2.8Ghz
True.ffmpeg decoder can't handle H264 1080p.
It's strange, because on same CPU I have problems only with streams with average bitrate 24 Mbps....
One another thing - ffdshow can use both cores, for that I use "tweak" with process affinity, I look for core, on which video player running, and set affinity to another core, then switch back, and finally, allows player to use both cores.
You are wrong.Both cores but no multithreaded.No speed improvement.
http://forum.doom9.org/showthread.php?p=922249#post922249
http://forum.doom9.org/showthread.php?p=923944#post923944
http://forum.doom9.org/showthread.php?p=923950#post923950

[_xxl]

Haali's splitter I hear no sound at all and with Quicktime's movie parser I can hear the sound as it should
True.Haali can't handle them.
I have noticed that libmad won't decode the sound of this clip (down right you'll find the download link)... mp3lib does work...

Please provide a decent link.

[sillKotscha]

Please provide a decent link.

all there - see my post above...

just open the provided link and open your eyes for "click here for download" and as I said you will find that invitation down right ;)

EDIT:

True.Haali can't handle them.

ok, but even MPC's internal MP4/MOV splitter is unable to connect to ffdshow to let ffdshow decode the audio of that clip... thus isn't ffdshow the culprit as it should (at least it would be nice) handle that sort of audio...

http://img389.imageshack.us/img389/2452/001ta4.jpg (http://imageshack.us)

[Leak]

You use MSVC8 SP1, don't you? In that case, you'll have to replace \trunk\bin\distrib\msvcr80.dll, \trunk\bin\distrib\InnoSetup\msvc80\ffdshow.ax.manifest and maybe microsoft.vc80.crt.manifest.
Ugh... now that you mention it... >_<

Okay, I'll fix that when I get home...

But, ummm... is there anyone here not using SP1 for VC80?

EDIT: Come to think of it, there's something not right with the manifests in SVN anyway - the msvcr80.dll at trunk/bin/distrib/ (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/bin/distrib/) that gets included in the installer is version 8.00.50727.42, but the manifest (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout/trunk/bin/distrib/microsoft.vc80.crt.manifest?revision=677&view=markup) for it specifies version 8.0.50608.0...

I guess I'll have to replace the runtime DLL in the distrib folder with the one I'm using and update the manifest, right?

[igor1st]

I have noticed that libmad won't decode the sound of this clip (http://www.wikiupload.com/download_page.php?id=101744) (down right you'll find the download link)... mp3lib does work...
No problem here. libmad works.

[sillKotscha]

ok, sorry... someone PM'ed me that the webhosting service I have uploaded that clip to is out of service right now - so here is a second attempt providing you with the clip in question...

Steve Balmer won't talk to me (http://www.savefile.com/files/550371) if I select libmad to decode the audio of that clip...

EDIT:

No problem here. libmad works.

which stream splitter do you use?? ... In my case Haali's splitter gets used as stream splitter - I've un-registered Haali's splitter concerning MPEG-PS support and now MPEG-I stream splitter (windows native MPEG-I splitter) is used and now the clip does play using libmad...

so, I nailed it down :) must be Haali's MPEG implementation which does caused my trouble... sorry for bothering the ffdshow devs

/ Sill

[Peuj]

No problem here. libmad works.

Yes libmad works for me too.

[Peuj]

ffdshow is still (after your various fantastic bug hunting work) unable to decode the sound of that clip (http://www.wikiupload.com/download_page.php?id=101790)...

if I select the Quicktime components to decode the file everything is ok but if I select the DShow framework in MPC to decode the clip the sound is totally fu**ed up...

I confirm that one.

[LotharZ]

I'm having problems with subtitles, If I select yellow as a colors, appears some defects.

Running ffdshow 1023 + MPC latest

http://img133.imagevenue.com/loc61/th_78568_snapshot20070313103415_122_61lo.jpg (http://img133.imagevenue.com/img.php?image=78568_snapshot20070313103415_122_61lo.jpg) http://img174.imagevenue.com/loc196/th_78885_Clip_122_196lo.jpg (http://img174.imagevenue.com/img.php?image=78885_Clip_122_196lo.jpg)

[haruhiko_yamagata]

I'm having problems with subtitles, If I select yellow as a colors, appears some defects.

Running ffdshow 1023 + MPC latest

Yes, I have been trying to fix this issue. I have many other issues related to subtitles.
But now, pehaps I should prioritize some other bugs that should be fixed before beta2. Please wait.

[CiNcH]

overlay anamorphic .avi bug is old, vmr9 or haali will play them correctly

With tryouts beta 1 and Overlay, anamorphic source material is displayed correctly.

I am not replaying from an AVI container, but from DVB stream.

[Px]

You are wrong.Both cores but no multithreaded.No speed improvement.
When I last time tested it (
http://forum.doom9.org/showthread.php?p=920783#post920783) I've got speed improvement ;)

[_xxl]

When I last time tested it (
http://forum.doom9.org/showthread.php?p=920783#post920783) I've got speed improvement ;)
Use timecodec.
No speed improvement for ffmpeg's h264 decoder on AMD X2.

[haruhiko_yamagata]

With tryouts beta 1 and Overlay, anamorphic source material is displayed correctly.

I am not replaying from an AVI container, but from DVB stream.
I don't have DVB. Is there any way to reproduce your problem?

[sillKotscha]

I don't have DVB.

here you go (http://www.savefile.com/files/550664)... DVB-S stream captured as PVA

EDIT: PVA DS Filter...

- RadLight PVA DirectShow filter v1.0.0.0 (http://www.free-codecs.com/download/RadLight_PVA_DirectShow_filter.htm)

- or use MPC's internal PVA filter

[Leak]

I don't have DVB. Is there any way to reproduce your problem?
Errrr... I had the same problem with the DVD I used for testing when it tried to switch from 4:3 to 16:9 (and failed), so I guess it's the same problem - but it only happens when using the Overlay Mixer.

The image gets stretched horizontally and is cut off at the right side.

[clsid]

I vaguely remember Overlay Mixer not supporting on-the-fly size changes.


@sillKotscha, submit your file to Haali, so that he can fix his splitter. He only just recently added MPEG PS support.

[Leak]

I vaguely remember Overlay Mixer not supporting on-the-fly size changes.
Well, the tooltip on the "Allow output format changes during playback > Connect to compatible filters only" checkbox lists Overlay Mixer as the very first compatible filter... :confused:

[ring0]

The following encoders only work with 25fps input: MPEG4, MPEG 1 and MPEG 2.

They seem to work with any frame rate which can be represented as 25*2^n, where n is an integer (even negative). So, it works for 6.25, 12.5, 25, 50, 100 etc.

[haruhiko_yamagata]

Sure, but as I mentioned the 'problem' appears a few secs (4-5) into playback, so the samples are necessarily around 40-50MB

Here is the HuffYuv-YV12 source:

http://rapidshare.com/files/20713071/Sample_HuffYuv-YV12_source.avi

and here is the FFV1 (AC, Large Context, KF=10) encode:

http://rapidshare.com/files/20716143/Sample_FFV1_AC_Large_Context_KF10.avi
Thank you for the sample.
I found that the bug begins at rev 2411 of the original ffdshow.
Please uncheck "Quality control"->"Drop frame on delay" for the time being. Of course, more radical fix is needed.

[haruhiko_yamagata]

With tryouts beta 1 and Overlay, anamorphic source material is displayed correctly.

I am not replaying from an AVI container, but from DVB stream.
Confirmed, on testing DVD playback.
Thank you, sillKotscha and Leak.

Well, the tooltip on the "Allow output format changes during playback > Connect to compatible filters only" checkbox lists Overlay Mixer as the very first compatible filter...
I'm wondering if I should remove overlaymixer from that list.

[haruhiko_yamagata]

Ugh... now that you mention it... >_<

Okay, I'll fix that when I get home...

But, ummm... is there anyone here not using SP1 for VC80?

I dislike the complicated run time problem, so I don't use SP1.
If you want to distribute ffdshow using SP1, you'll need a proper testing environment that msvcr80 (or .NET frame work) is not installed as shared assembly. I installed WindowsXp on a new partition for that purpose. Windows 2000 and Xp differs completely in that regard.

[clsid]

Can someone give a clear and precise description of the Overlay bug? Then I can update the opening post.

Haruhiko, can you list some of the bugs you know of that are in the subtitle code? Then I can add those to the list as well.

@all, are there any known bugs that are not yet in my opening post?

[CiNcH]

The image gets stretched horizontally and is cut off at the right side.

Sounds pretty much like the same problem.


I vaguely remember Overlay Mixer not supporting on-the-fly size changes.

It is working with tryouts beta 1 and Overlay Mixer.

But the problem with beta 1 is (as I posted earlier) that on-the-fly resolution changes do not work properly with Overlay Mixer now. I solved this problem by performing a graph rebuild at resolution change.
Another way to solve this problem is to use a later daily build. There you do not need to send down the information any longer but again have the problem with wrong anamorphic displaying.

[CiNcH]

Can someone give a clear and precise description of the Overlay bug?


Anamorphic material is just not displayed correctly when using the Overlay Mixer.
Like Leak said "The image gets stretched horizontally and is cut off at the right side."
I am just using ffdshow as a raw video post processor, so no decoder is involved and the problem also occurs when no filter (deinterlacing, sharpening or whatever) or anything else is used.

[haruhiko_yamagata]

It is working with tryouts beta 1 and Overlay Mixer.

But the problem with beta 1 is (as I posted earlier) that on-the-fly resolution changes do not work properly with Overlay Mixer now. I solved this problem by sending the connection format down the filter graph.
Another way to solve this problem is to use a later daily build. There you do not need to send down the information any longer but again have the problem with wrong anamorphic displaying.
On-the-fly resolution changes do not work with overlay mixer. It's the spec of the overlaymixer. Are there any ffdshow that can do that?
On-the-fly AR change should work with overlay mixer, while current ffdshow does not work correctly.

[CiNcH]

On-the-fly resolution changes do not work with overlay mixer. It's the spec of the overlaymixer. Are there any ffdshow that can do that?

As I said, it works with latest daily build if I am not completely mistaken. There does not seem to be a graph rebuilt, so it is really on-the-fly.

[haruhiko_yamagata]

Haruhiko, can you list some of the bugs you know of that are in the subtitle code? Then I can add those to the list as well.

As for srt subtitle and OSD, the color problem should be fixed.

Font->Charset->Default should use the native language of the country (version of the Windows language). For Japanese version of Windows, it should use shift-Jis, but it uses ANSI. This issue is true with UNICODE build as long as the subtitle file is ANSI (or shift-Jis).

As for SSA/ASS/ASS2, it's in the very early stage of the development, not implemented rather than very buggy.

[Delerue]

Anamorphic material is just not displayed correctly when using the Overlay Mixer.
Like Leak said "The image gets stretched horizontally and is cut off at the right side."
I am just using ffdshow as a raw video post processor, so no decoder is involved and the problem also occurs when no filter (deinterlacing, sharpening or whatever) or anything else is used.

Yeah. It seems to be a MPC (or maybe non-WMP) bug, not a FFDShow bug, because even using Microsoft codecs, some WMV anamorphic videos (like Step Into Liquid (http://download.microsoft.com/download/e/a/d/eadb9b42-728b-42b0-bfdf-b472fa2a2464/Step_into_Liquid_1080.exe)) can't be resized correctly. But if you open it with WMP, everything is fine even with overlay.

[chros]

I'm using ffdshow to convert dts to aac through avisynth:
graph: mkv (Haali's splitter) -> ffdshow (on output only 16 bit allowed)
script: DirectShowSource("audio.grf",video=false,Audio=true)

When I try to encode this in megui or behappy it gives me an error (see below). Strange, that I can playback the avs script in MPC ...
If I switch ffdshow to AC3filter in the graph, the encoding works well ...

The megui log:
Log for job job32

Error:
MeGUI.AviSynthException: DirectShowSource: Couldn't open GRF file.audio.grf:
This objectsurface is not supported
(audio.avs, line 5)
at MeGUI.AviSynthClip..ctor(String func, String arg, AviSynthColorspace forceColorspace, AviSynthScriptEnvironment env)
at MeGUI.AviSynthAudioEncoder.encode()
----------------------------------------------------------------------------------------------------------
The current job contains errors. Skipping chained jobs

[_xxl]

As for SSA/ASS/ASS2, it's in the very early stage of the development, not implemented rather than very buggy.

Implementing this is important because vsfilter is no longer updated and it has some bugs.Maybe after beta2 if time is available.

[haruhiko_yamagata]

As I said, it works with latest daily build if I am not completely mistaken. There does not seem to be a graph rebuilt, so it is really on-the-fly.
It also works with beta 1, but only when sending the connection format downstream.
Only when new resolusion is smaller than before.
Milan's 20051129 does not try to resize if new resolusion is bigger.

but only when sending the connection format downstream.What is this?

[Leak]

I dislike the complicated run time problem, so I don't use SP1.
If you want to distribute ffdshow using SP1, you'll need a proper testing environment that msvcr80 (or .NET frame work) is not installed as shared assembly. I installed WindowsXp on a new partition for that purpose. Windows 2000 and Xp differs completely in that regard.
Yeah, I was thinking of installing XP in a VirtualBox instance for testing...

Oh well, the good thing about VM images is that they're highly reuseable... :)

[CiNcH]

What is this?

Actually a mistake. Because what solved the problem was automatic graph rebuilding at resolution change... sorry.
Sending connection format downstream solved another problem, not ffdshow related...


Yeah. It seems to be a MPC (or maybe non-WMP) bug not a FFDShow bug

Using DVBViewer. Fact is, with beta 1 it works, with latest daily build it doesn't ;) .

[Fizick]

Here is ffdshow_rev1022_avisynth2_20070312_igor1st.exe
with Leak Avisynth contribution (compilied VC7 ?):

http://www.mytempdir.com/1252011

It works well! I try DeGrainMedian filter (it use 3 frames).
Bug: first 5 frames are the same as very first (i.e. duplicates).

What is 10,10,10 buffers in settings?

[igor1st]

@all, are there any known bugs that are not yet in my opening post?
OSD item "Encoder info" doesn't work anymore.

[Leak]

It works well! I try DeGrainMedian filter (it use 3 frames).
Bug: first 5 frames are the same as very first (i.e. duplicates).
That's strange - with default settings the plugin should buffer the first 11 frames before even requesting the frame 0 from AviSynth, so AviSynth and DeGrainMedian should be able to request frames 0-10 right away...

Whenever AviSynth requests a frame, the nearest one that is buffered is used.

Wait a second - the very first frame is produced by buffering a single input frame followed by requesting frame 10-million-and-a-bit to avoid filters re-using anything they cached while processing that first frame. Maybe DeGrainMedian doesn't like this? (i.e. I'm requesting frames 10810800, 1, 2, 3, etc.)

This was the easiest solution for the problem of having to pass the first frame through so the filter graph starts to run. If I passed the first frame through as-is I'd have to resize it, change the colourspace and whatnot myself... sure, I could blow away AviSynth and re-initialize it after fetching the first frame, but it's already bad enough doing it once after every seek...

What is 10,10,10 buffers in settings?
Those settings specify how many buffers should be kept - while ffdshow is running, there's always a "current" frame. Buffering happens by taking the frames that come into the filter until there's enough frames buffered, and only then does anything get passed further up the filter chain and the DirectShow graph.

The first value specifies how many buffers should be kept before the "current" frame.

The second value specifies a limit for how many frames should be buffered for producing the "current" frame - i.e. SelectEven halves the number of frames, so the filter will buffer 2 input frames to produce one output frame; this is calculated from the number of frames the resulting AviSynth clip has but I needed some upper limit so doing a "Trim(0,1)" won't allocate buffers for the whole 10 million frames or so the fake input clip is long.

The third value specifies the number of frames to keep buffering ahead.

So with settings of x/y/z AviSynth filters can request frames from n-x to n+z when working on frame n, while the total number of frames AviSynth returns can be up to 1/y of the number of frames ffdshow hands over to AviSynth.

This all assumes that adding or removing frames happens linearly, of course.

In other news, I've managed to reproduce the installation problem (d'oh!) and fixed it. I'll post an updated build in a bit.

np: One Second Bridge - Alt. Ending (One Second Bridge)

[Leak]

Okay, here's a new build based on rev. 1028 (http://leak.no-ip.org/AviSynth/ffdshow_AviSynth_rev1028_20070313.exe) that should work even if the Visual C 8.0 redistributables aren't installed on your PC.

Could someone who had problems installing my earlier builds give it a test please? (I know it worked on a fresh Windows XP SP2 install, but I'd like to be sure...)

np: Console - Tunnelvision (Pan Or Ama)

[Liisachan]

Implementing this is important because vsfilter is no longer updated and it has some bugs.Maybe after beta2 if time is available.

Maybe mplayer's code or libASS (http://sourceforge.net/projects/libass/) is reusable for this?

[WorBry]

Thank you for the sample.
I found that the bug begins at rev 2411 of the original ffdshow.
Please uncheck "Quality control"->"Drop frame on delay" for the time being. Of course, more radical fix is needed.

Yes, the 'temporary remedy' appears to work. So, if I understand the situation correctly, the problem is entirely related to the FFDShow decoder per se and has nothing to do with the FFV1 encoding part. In other words, am I safe in committing my HuffYuv-YV12 archives to FFV1 (AC, Large Context KF10) without the possibility of changes to the FFV1 coder that might result in decoding incompatibilities with future FFDShow builds in which this bug has been fixed?

[haruhiko_yamagata]

Yes, the 'temporary remedy' appears to work. So, if I understand the situation correctly, the problem is entirely related to the FFDShow decoder per se and has nothing to do with the FFV1 encoding part. In other words, am I safe in committing my HuffYuv-YV12 archives to FFV1 (AC, Large Context KF10) without the possibility of changes to the FFV1 coder that might result in decoding incompatibilities with future FFDShow builds in which this bug has been fixed?
Probably no, because your computer (and most computers) is not fast enough to decode FFV1. FFV1 is not as fast as HuffYuv.

The problem has been fixed at rev 1030.

[haruhiko_yamagata]

Maybe mplayer's code or libASS (http://sourceforge.net/projects/libass/) is reusable for this?
Thank you, I did not know libass. I'll look at it when I have time.
Can mplayer render SSA? ffdshow uses code from mplayer, which is not fully implemented.

[Milvus]

libASS is part of the mplayer project, so yes, mplayer can render SSA/ASS. Not as complete as VSFilter, but still quite nice.

Another link you should find very interesting : The ASA wiki (http://asa.diac24.net/) : Lot of informations about subtitling libraries, like this Compatibility Chart (http://asa.diac24.net/Compatibility).

[haruhiko_yamagata]

libASS is part of the mplayer project, so yes, mplayer can render SSA/ASS. Not as complete as VSFilter, but still quite nice.

Another link you should find very interesting : The ASA wiki (http://asa.diac24.net/) : Lot of informations about subtitling libraries, like this Compatibility Chart (http://asa.diac24.net/Compatibility).
Thank you for the links. I'll read later.

Do you really think mplayer is better than current ffdshow in handling SSA? I tested MPUI 20070310 for embedded SSA in mkv (mewmew-vorbis-ssa.mkv (http://www.matroska.org/samples/mewmew/index.html)), but I think ffdshow is better. Yes, I know it's meaningless to argue this, both of them are quite incomplete. Or is it the issue with Windows port or my installation?

[Episode]

On my OSX build of mplayer I need to use -ass -embeddedfonts -fontconfig flags to get subtitles rendered properly. Perhaps similar flags will work on Windows-version as well.

[haruhiko_yamagata]

On my OSX build of mplayer I need to use -ass -embeddedfonts -fontconfig flags to get subtitles rendered properly. Perhaps similar flags will work on Windows-version as well.
Adding these config, mplayer crashes. Sorry for OT, but how to make it play on Windows Xp?

[Episode]

@haruhiko_yamagata, how about trying it without the -fontconfig flag? I just tried mplayer 1.0rc1 on my Windows machine and "mplayer -ass -embeddedfonts filename" seems to work quite well.

[Eragon4ever]

On Windows MPlayer displays ASS with "-sid 0 -ass" however it crashes if the subtitle track is SSA or SRT.

[haruhiko_yamagata]

Thank you. ASS plays.
Code have already been imported from mplayer, though it is not updated. Main problem seems to be the porting.

[Peuj]

clsid2: Removed WavPack support. It was far to incomplete to be usable. Use CoreWavPack DS filter instead.

just for information: the latest CoreWavPack filter can be find here http://coreforge.org/frs/?group_id=28

[JnZ]

Could you please repair "proper" multi-channel decoding for OGG Vorbis files? (Due to Vorbis I speciffication). Official vorbis decoder decodes proper channels, but FFDShow decodes bat channel order.
Quote: Vorbis I uses a single mapping type (0), with implicit PCM channel mappings.

EDIT: OR FFDshow use right setting?

[_xxl]

Removed WavPack support. It was far to incomplete to be usable. Use CoreWavPack DS filter instead.
Libavcodec wavpack was removed.

[Fizick]

Leak,
your new build installed well:
1. Windows2000 without msvc8
2. WinXP with MS VC2005 express (NET, etc)

But bug still exist:
Number of extra very first frames= (buffer ahead)+1

Try script:

showframenumber()

and load decoded ffdshow video in virtualDub.

first frames returns number 1071079 (not 0).

[Px]

Use timecodec.
I prefer to watch video in video players ;), and I saw speed improvement (on 45 Mbps frames), but not too much, yes....

[_xxl]

but not too much, yes....
Nothing because libavcodec h264 decoder is not multithreaded.

[foxyshadis]

I prefer to watch video in video players ;), and I saw speed improvement (on 45 Mbps frames), but not too much, yes....

Any improvement you saw probably came from queuing keeping short cpu peaks from killing smoothness. Queuing can't thread the decoder, but it does keep the decoder working while waiting for the last frame to be displayed, which it won't without queuing.

[Leak]

But bug still exist:
Number of extra very first frames= (buffer ahead)+1

Try script:

showframenumber()

and load decoded ffdshow video in virtualDub.

first frames returns number 1071079 (not 0).
Please see my answer to this here (http://sourceforge.net/tracker/index.php?func=detail&aid=1676882&group_id=173941&atid=867362)...

[Px]

Any improvement you saw probably came from queuing keeping short cpu peaks from killing smoothness. Queuing can't thread the decoder, but it does keep the decoder working while waiting for the last frame to be displayed, which it won't without queuing.
I agree with you :)

[interfan]

Hi,

I have a problem to which I can't find a solution. I'm lost :(

After getting a new PC, I've installed XP SP2 and the latest ffdshow from www.free-codecs.com.
I then have noticed a problem, so I've reinstalled it, again, after d/l the latest version from the same source.
My current(which is 2nd) version is the FFDShow 2007-02-09 (rev. 899) SSE version.

But the problem persists.

In some sites the embedded video plays in strange colors(green, purple and other such colors). What bothers me most is the P2P based streaming videos like PPSTREAM.

I've updated my ATI driver to the latest as well- it didn't helped either.

I even don't know whether it's ffdshow-related problem, as I've not tried to play streaming videos before installing FFDSHOW. And, furthmore, except FFDSHOW, the only thing I install is real and quick-time alternatives.


I don't want to install and reinstall too much, especially as long as I'm enjoying a clean and fast system(you how things go in the beginning) and I like FFDSHOW very much, especially as it has given to my old P3 some another 2 years of usability :)


I'll appreciate help.

[clsid]

Those embedded videos are likely Windows Media files. By default ffdshow doesn't decode such files. Your problem can be solved by disabling DXVA for the Windows Media codecs. You should be able to find an option to disable it somewhere in your graphics card control panel.

[_xxl]

In some sites the embedded video plays in strange colors(green, purple and other such colors). What bothers me most is the P2P based streaming videos like PPSTREAM.

Is ffdshow decoding this?

[zambelli]

Can somebody else confirm these bugs:


AVIS decoding is not working at all in DirectShow mode - the filter fails to connect. In VfW mode it just decodes black.
When YV12 decoding is enabled only in VfW mode and YV12 video is played through DirectShow, the image gets mirrored
When YV12 decoding is enabled in both DShow and VfW, YV12 video is always converted to YUY2 when rendered to VMR (i.e. GraphEdit playback).

I've verified them with both beta1 and a clsid build from 3-14-2007. I'm fairly certain about the 1st one, but the 2nd & 3rd might be video hardware related.

[Episode]

I can confirm these bugs.

[Px]

In some sites the embedded video plays in strange colors(green, purple and other such colors). What bothers me most is the P2P based streaming videos like PPSTREAM.
That is not ffdshow problem, update your Windows Media Player at least to 10, better to 11 at http://windowsupdate.microsoft.com,
section "Optional"

[interfan]

That is not ffdshow problem, update your Windows Media Player at least to 10, better to 11 at http://windowsupdate.microsoft.com,
section "Optional"

I hate Microsofts' media players- they are so heavy...

But I'll try your suggestion, as I have some feelings that might help. I didn't understand clsids' suggestion anyway, because I don't even know what DXVA is.

I'll report whether it fixes the problem.

[clsid]

Installing just "Windows Media Format Runtime 11" instead of the whole Windows Media Player might be enough.

DXVA is related to hardware acceleration. Google it.

[LigH]

DXVA = DirectX Video Accelleration: The graphics chipset should support a little part of the MPEG decoding algorithm... Should. Not all models of a graphic card family provide it. Sometimes applications (like software DVD players) fail to use it even if the chip provides it. On top of it, Microsoft waited with the full support for Vista.

http://en.wikipedia.org/wiki/DXVA

[interfan]

Installing just "Windows Media Format Runtime 11" instead of the whole Windows Media Player might be enough.

How can I install it? it's a separated software to install or it's installed as a component in the WMP11 installation?

----

Never mind- I see these are just free codecs. I was hoping to avoid installing too much codecs, but don't really have an option...

[Px]

Installing just "Windows Media Format Runtime 11" instead of the whole Windows Media Player might be enough.
Sometimes, there are strange problems with standalone runtime, which doesn't appear when full pack with player installed, so I prefer to recommend player update, nor runtime. Also, I must mention, that in some cases you need to additionally install WMV9 VCM (last time - game "You are empty")

[Px]

How can I install it? it's a separated software to install or it's installed as a component in the WMP11 installation?
As a component
Never mind- I see these are just free codecs. I was hoping to avoid installing too much codecs, but don't really have an option...
You have these codecs already installed, but they outdated, and because of that, produces color artifacts on video...

[interfan]

It's working!!! :)

Those Windows Media Format Runtime 11 really helped and it's definitely the better way(who the hell need WMP?!) for me.

I'm thankful as I've gone through 3 forums before this one and nobody could help me there.

One last question- I like controlling my system and I dislike having garbages which I can't erase purely. Will I be able to uninstall "Windows Media Format Runtime 11"? If yes- how?

[zambelli]

I can confirm these bugs.
Thanks!

Can anybody from the ffdshow team look into this? I can file an official bug through Sourceforge too if necessary.

[clsid]

I'll add them to the buglist. See opening post. Thanks for submitting them.

[foxyshadis]

MPC build discussion moved to new thread (http://forum.doom9.org/showthread.php?t=123537).
DVD navigator bug moved to new thread (http://forum.doom9.org/showthread.php?t=123538).

[Blinx123]

Has anybody here used ffdshow with the Lord of the Rings DVDs? How does it look in the fact of detail and beauty?

Could anybody post some Comparision Pictures between the normal and the scaled/post processed version?

[foxyshadis]

That might be a more appropriate question for Jeremy's thread (realtime postprocessing with ffdshow and avisynth (http://forum.doom9.org/showthread.php?t=115727)) than here.

[skystrife]

Again, this has probably been asked several times before, but are there any plans to compile a 64-bit version of ffdshow-tryouts? The only 64-bit build out there right now is the celtic_druid one, and IIRC it's fairly old.

[_xxl]

Now ffdshow can't be compiled for 64-bit.Also libavcodec.This is not a priority.

[Dr Pizza]

Again, this has probably been asked several times before, but are there any plans to compile a 64-bit version of ffdshow-tryouts? The only 64-bit build out there right now is the celtic_druid one, and IIRC it's fairly old.

I wish I could at least find the source to the celtic druid 64-bit build. It's apparently a build of 2546 (which is the most recent ffdshow source on sf.net), but it must have been modified, because 2546 won't build as an x64 binary (indeed, there aren't even project settings for it). If I could get that working then patching with updates from ffdshow-tryouts might be possible, to produce an up-to-date 64-bit build. But unfortunately, I cant.

The big stumbling block is the assembler portions. C and C++ I can do. At a pinch, SSE intrinsics are possible. Assembler, however, I can't do a damn thing with. If yasm won't build it, I can't fix it.

I wish x64 were regarded as more important than it currently is.

[Episode]

Is there really a need for x64 build of ffdshow? I mean, it won't be any faster than normal 32-bit build and x64 is compatible with normal builds.

[haruhiko_yamagata]

AVIS decoding is not working at all in DirectShow mode - the filter fails to connect. In VfW mode it just decodes black.


AVIS decoding in DirectShow mode works for me.
In vfw mode, ffdshow fails to connect.:confused:

[Dr Pizza]

Is there really a need for x64 build of ffdshow? I mean, it won't be any faster than normal 32-bit build and x64 is compatible with normal builds.
64-bit software can't use 32-bit codecs (so if you're using 64-bit WMP, for example, you need 64-bit codecs).
And it probably should be a bit faster, too, due to the extra architectural registers (as long as the compiler does a decent job, at least).

[lilhobo]

why cant i using ffdshow for RV10 in sony vegas???

or do i need to remove realplayer first??

[skystrife]

Is there really a need for x64 build of ffdshow? I mean, it won't be any faster than normal 32-bit build and x64 is compatible with normal builds.

There are several x64 players out there that won't use 32-bit codecs. If you take a x64 build of Vista, your Media Center, Movie Maker, and DVD Maker are all 64-bit, so the current 32-bit ffdshow won't work with those applications.

I'm just really hoping that this will happen, because I love ffdshow for it's immense flexibility, and I'd like to be able to use an updated 64-bit version of it (celtic_druid's build has some issues with H.264 decoding- artifacts galore).

[Liisachan]

why cant i using ffdshow for RV10 in sony vegas???

or do i need to remove realplayer first??

EDIT
Sorry, I posted an idiotic reply mistaking rv10 with rv40 ^^;
*deletes everything*

Ok, the only I could say is,,, the config dialog says "Real Video (RV10, RV20, incomplete)"--so the support is imcomplete...

[clsid]

Does Sony Vegas use VFW? In that case make sure that you have enabled RV10 in the ffdshow VFW decoder.

[haruhiko_yamagata]

AVIS decoding in DirectShow mode works for me.
In vfw mode, ffdshow fails to connect.:confused:
In vfw mode it works now. My setting was broken (AVI Decompressor was disabled).
@zambelli: could you upload samples (with the script and original file)?

[KoD]

embeeding asm in C programs on x64 architecture machines is different than on x86. Explanations in MSDN. Not to mention 64 bit in Linux is different than 64 bit in Windows (int is also 64 bit while on Windows no), so don't ever expect for any 64 bit code developed on a Linux machine (like libavcodec is) to be easily ported to 64 bits on Windows, unless it was designed to be. I doubt that Linux developers really care about porting it to Windows though, not to mention fast bit operations may not be ported unless rewritten specifically.

[HowlerX]

Was wondering why I couldn't get h.264 mov trailer to decode with megui. Took me almost two hours to figure out that I had left "Use ffdshow only in:" active but did not have megui in the list. :(

Suggestion: Please add megui.exe to list of programs allowed.

... I should really be sleeping...

[Dr Pizza]

embeeding asm in C programs on x64 architecture machines is different than on x86. Explanations in MSDN.
Right, no inline asm in x64. But that's probably not a problem, because VC++'s inline asm syntax was different from gcc/g++'s anyway, so I wouldn't expect much or any inline asm to be used at all; nasm or yasm syntax is far closer to portable.

Not to mention 64 bit in Linux is different than 64 bit in Windows (int is also 64 bit while on Windows no), so don't ever expect for any 64 bit code developed on a Linux machine (like libavcodec is) to be easily ported to 64 bits on Windows, unless it was designed to be.
I think most of this kind of code uses typedefs anyway so the different 64-bit models shouldn't make a difference. Stuff like the H.264 decoder already has 64-bit asm with Windows-specific code paths, it just doesn't work properly.

[haruhiko_yamagata]

Was wondering why I couldn't get h.264 mov trailer to decode with megui. Took me almost two hours to figure out that I had left "Use ffdshow only in:" active but did not have megui in the list. :(

Suggestion: Please add megui.exe to list of programs allowed.

... I should really be sleeping...
Thank you. Added at rev 1048. Please open the dialog (pressing edit button) and reset the list.

[ONjA]

Hi! Some long lived bug forced me to register.
ffdshow version: at least any starting from this year, icl9 or generic (clsid build)
Sample: http://chukos.narod.ru/sa2.avi 1.2Mb (DIV3)
Player: mpc/crystal
Setup: Postprocessing on, > fast spp deblock <
How to: While playng video open and close ffdshow's settings window several times. You'll probably see access violation...

Can't determine any dependency on file's properties, but I have quite a lot of such files. Some similars in type do not exploit that.
Normal (not fast) SPP works fine.

[haruhiko_yamagata]

Hi! Some long lived bug forced me to register.
ffdshow version: at least any starting from this year, icl9 or generic (clsid build)
Sample: http://chukos.narod.ru/sa2.avi 1.2Mb (DIV3)
Player: mpc/crystal
Setup: Postprocessing on, > fast spp deblock <
How to: While playng video open and close ffdshow's settings window several times. You'll probably see access violation...

Can't determine any dependency on file's properties, but I have quite a lot of such files. Some similars in type do not exploit that.
Normal (not fast) SPP works fine.
Thank you, but I can't reproduce. Any aditional information such as CPU, OS, extra filter... are wellcome. Which page to open?

[foxyshadis]

There are several x64 players out there that won't use 32-bit codecs. If you take a x64 build of Vista, your Media Center, Movie Maker, and DVD Maker are all 64-bit, so the current 32-bit ffdshow won't work with those applications.

I'm just really hoping that this will happen, because I love ffdshow for it's immense flexibility, and I'd like to be able to use an updated 64-bit version of it (celtic_druid's build has some issues with H.264 decoding- artifacts galore).

It'll happen as soon as a developer or willing volunteer gets x64. Until then it's untestable even if it was tweaked to compile; how could we fix any inevitable crashes?

[Dr Pizza]

If someone could provide me with a buildable x64 source tree and suitable files to test it, I could certainly have a look.

[ONjA]

Thank you, but I can't reproduce. Any aditional information such as CPU, OS, extra filter... are wellcome. Which page to open?

Verified on 3 computers:
AXP and Duron (old, no sse, generic) - w2ksp4, no additional codecs\splitters\packs.
AXP under clean ws2003.

Just invoke ffdshow via player's control, i.e. mpc - filters -> ffdshow video decoder, then close, then invoke once more. Or the same via tray control.

[Px]

It'll happen as soon as a developer or willing volunteer gets x64.
I can test such builds in my second os, XP x64

[Leak]

Okay, here's another updated version of my AviSynth filter:

The current installer based on revision 1048 (and previous versions) can be found here (http://leak.no-ip.org/AviSynth/ffdshow/)
Patch is here (https://sourceforge.net/tracker/index.php?func=detail&aid=1676882&group_id=173941&atid=867362)

Changes are that setting 0/1/0 and disabling "Apply Pulldown" should prevent any buffering ahead, just like the old filter. I've also added some code to send an onStop event to the filters when using the DirectShow encoder and encoding finishes, as I need to do some cleaning up if encoding should work more than once in a row.

Note: if you're using ffvfw (the Video for Windows wrapper around ffdshow) for encoding with the AviSynth filter turned on and it is set to buffer ahead you *will* get repeated frames at the beginning of your encode, as VfW *always* expects a compressed frame to come out when it hands an uncompressed frame to a codec - and there's nothing I can do about that.

Encoding using the DirectShow "ffdshow video encoder" works fine, though.

[multiblitz]

I guess this is a bug:

If I use ffdshow (rev 1020) to playback VC-1 (HD-DVD) (using Haali's splitter, AnyDVD and ZP), I get a white picture (pure white screen) if I don't resize the picture. With resize (lanzos), I have a nice picture. But I don't want to resize (1920*1080 is already there...)

BIggest Bug so far: If ffdshow is used in the video-part (does not matter if it is only postprocessing-raw-filter after the original MS VC-1 Codec or directly the ffdshow-VC-1-decoder): Instead of correct speed of 23.976 fps, it recognizes 29.999 fps. So, the video is much faster than the audio which indeed is the case...without ffdshow in the video-chain, with pure MS VC-1, it works, eventhough SPDIF is still somewhat jaggy (core duo 6400 at 3.6 ghz).

[cc979]

with rev1050 and before
can't get ffdshow to be used with virtualdub (us sure it work before) it falls back to opening with xvid

is there something up with the vfw part?

cheers

[clsid]

It works fine here. If you have both installed it is to be expected that Xvid is used instead of ffdshow. VFW doesn't have a merit system afaik.

[skystrife]

I can test such builds in my second os, XP x64

I know nothing about compiling and such, but I would certainly be interested in beta-testing such software. =)

[cc979]

VFW doesn't have a merit system afaik.

i did'nt know about that, cheers

[akupenguin]

Is there really a need for x64 build of ffdshow? I mean, it won't be any faster than normal 32-bit build and x64 is compatible with normal builds.

It would be faster if you didn't have to disable asm. A 64bit build of ffh264 is 15% faster than a 32bit build on linux, entirely due to the extra registers.

Stuff like the H.264 decoder already has 64-bit asm with Windows-specific code paths, it just doesn't work properly.
There is a little 64bit asm (though mostly we just let gcc compile the same asm for 32 and 64bit). But there are no windows-specific codepaths in libavcodec.

[bob0r]

ffdshow_rev1050_20070318_clsid.exe
ffdshow_rev1050_20070319_clsid_sse_icl9.exe

Which build should i put on x264.nl?
And why ICL9? (does it still only benefit on Intel and not AMD?)

[fastplayer]

And why ICL9? (does it still only benefit on Intel and not AMD?)
:readfaq:
ffdshow tryouts FAQ (http://ffdshow-tryout.sourceforge.net/html/en/faq.htm)

[Leak]

:readfaq:
ffdshow tryouts FAQ (http://ffdshow-tryout.sourceforge.net/html/en/faq.htm)
That still doesn't say anything about whether the executable has Intels dain bramaged "if it isn't an Intel CPU perform much worse/don't run" check patched or not...

[foxyshadis]

It's been patched. Current versions of icl just crash when you use AMD, they don't even bother to use the slow route anymore. (At one point VM9 was going to patch them to crash on Intel instead of AMD... ;p)

[haruhiko_yamagata]

Verified on 3 computers:
AXP and Duron (old, no sse, generic) - w2ksp4, no additional codecs\splitters\packs.
AXP under clean ws2003.

Just invoke ffdshow via player's control, i.e. mpc - filters -> ffdshow video decoder, then close, then invoke once more. Or the same via tray control.
Thank you, now I can reproduce the problem using another computer.

[Dr Pizza]

It would be faster if you didn't have to disable asm. A 64bit build of ffh264 is 15% faster than a 32bit build on linux, entirely due to the extra registers.


There is a little 64bit asm (though mostly we just let gcc compile the same asm for 32 and 64bit). But there are no windows-specific codepaths in libavcodec.

tbh, if there were generic C paths that would do me for the moment. Making sure that the C parts are 64-bit LLP64 clean (god what a stupid model! I sorta understand why MS did it, but it's still annoying; they should have bitten the bullet and gone LP64) would be a good start.

[clsid]

It's been patched. Current versions of icl just crash when you use AMD, they don't even bother to use the slow route anymore. (At one point VM9 was going to patch them to crash on Intel instead of AMD... ;p)
I have patched the genuineintel check in the ICL libraries, so SIMD instructions should work on AMD processors too. Are you sure it crashes on AMD?

[foxyshadis]

I mean, current unpatched icl. Sorry.

[bob0r]

So my question is answered with Beta 2? :thanks:

[ExtraEye]

But of course. Beta 2 is the answer to all questions.

[bob0r]

http://x264.nl/slap.gif

[fastplayer]

@clsid:
Could you add "beta2" to the filename like in the first beta (ffdshow_beta1_20061211_clsid.exe)?

Edit: Thanks! This should help separate it from the nightlies.

[HeadBangeR77]

Hello,
We've been discussing some colour-spaces issues in the ColorMatrix thread, and KillerZero seems to have discovered an interesting thing, that might affect ffdshow decoding.

http://forum.doom9.org/showthread.php?p=972785#post972785

I can't give you more info, since I can't reproduce his findings.

and ...

:thanks:

VERY MUCH for the new beta2! :)

PS. I don't know how the counter works, yet it shows 0 downloads, while at least two people have already downloaded it (me and a friend of mine)...

[akupenguin]

tbh, if there were generic C paths that would do me for the moment. Making sure that the C parts are 64-bit LLP64 clean (god what a stupid model! I sorta understand why MS did it, but it's still annoying; they should have bitten the bullet and gone LP64) would be a good start.

The C parts of libavcodec are definitely LLP64 clean: The only codec that uses (long) at all is the faad wrapper.

[Avish]

I have patched the genuineintel check in the ICL libraries, so SIMD instructions should work on AMD processors too. Are you sure it crashes on AMD? No. It does not crash. I've AMD 64 X2 & Im using ur latest build [ICL9 one] without any problems so far.

[Leak]

No. It does not crash. I've AMD 64 X2 & Im using ur latest build [ICL9 one] without any problems so far.
If it crashed on that CPU something would be really wrong - as all newer AMD CPUs include full SSE2 support (http://en.wikipedia.org/wiki/SSE2#CPUs_supporting_SSE2)...

[foxyshadis]

PS. I don't know how the counter works, yet it shows 0 downloads, while at least two people have already downloaded it (me and a friend of mine)...

It only updates every few hours. They do that on purpose so that all of the early adopters get to feel special, like they're the very first to grab it. That's my theory anyway. ;)

The rest of the colorspace issues raised in that thread are too murky to give me any idea of where the fault might lie. And I can't usually tell the difference between 601 and 709 - mild colorblindness - so I'd never notice anyway.

Want to include leak's patch now? It as good a time as ever.

[HeadBangeR77]

I've been using ICL9 builds for months and it never crashed on me, though I don't have an SSE2-capable CPU (Athlon XP-M).

@ foxyshadis:
Good theory. And you know what, I felt somehow "special" downloading the beta2 and seeing "0" on the counter. :D

As to colours issue with HD - yes, I know it's very vague, and I wasn't able to reproduce it so far. Could be graphic card drivers dependent, instead of ffdshow.

[fastplayer]

PS. I don't know how the counter works, yet it shows 0 downloads, while at least two people have already downloaded it (me and a friend of mine)...
Then the sourceforge servers would do nothing else besides generating pages after every single download where only the counter changed. Unnecessary traffic and CPU load IMO...

[Leak]

PS. I don't know how the counter works, yet it shows 0 downloads, while at least two people have already downloaded it (me and a friend of mine)...
You can click on that "0" to get to the download statistics page for that file - there you'll find the up-to-date download count. ;)

[cc979]

did some quick tests with ffdshow using gcc.4.2.0 pre-release

cc-gcc-4.0.4, and haali splitter v1.7.100.18
tested using elephant dream_hd xvid/ac3 (8mbs stream mostly)

rev1058 - gcc 4.0.4 and gcc 4.2.0 on ffmpeg,mplayer,libmpeg2 - using mmx only
null: 8s, kernel: 0s, total: 8s, real: 8s, fps: 79.9, dfps: 75.3
vmr9: 8s, kernel: 1s, total: 9s, real: 10s, fps: 68.7, dfps: 59.9
ovrl: 8s, kernel: 0s, total: 8s, real: 8s, fps: 76.9, dfps: 73.8

rev1058 - all gcc 4.0.4 - mmx only
null: 8s, kernel: 0s, total: 8s, real: 8s, fps: 77.1, dfps: 74.5
vmr9: 8s, kernel: 0s, total: 9s, real: 11s, fps: 67.2, dfps: 57.8
ovrl: 8s, kernel: 0s, total: 9s, real: 9s, fps: 72.7, dfps: 70.8

rev1041
using o4 mmx
null: 9s, kernel: 0s, total: 9s, real: 9s, fps: 71.2, dfps: 67.7
vmr9: 9s, kernel: 0s, total: 10s, real: 11s, fps: 64.4, dfps: 54.9
ovrl: 9s, kernel: 0s, total: 9s, real: 9s, fps: 69.5, dfps: 66.9

using o4 sse1
null: 9s, kernel: 0s, total: 9s, real: 9s, fps: 70.9, dfps: 68.5
vmr9: 9s, kernel: 0s, total: 10s, real: 11s, fps: 63.3, dfps: 54.7
ovrl: 9s, kernel: 0s, total: 9s, real: 9s, fps: 69.8, dfps: 67.0

looks like a improvement me

[_xxl]

@albain
http://img144.imagevenue.com/loc427/th_85485_Capture_122_427lo.jpg (http://img144.imagevenue.com/img.php?image=85485_Capture_122_427lo.jpg)
You have by mistake commited some errors.

[KoD]

cc979, you do realize that's a very short sample and that you'd have to repeat the measurement at least three times in each case to assure that it's consistent.

[zambelli]

In vfw mode it works now. My setting was broken (AVI Decompressor was disabled).
@zambelli: could you upload samples (with the script and original file)?
This is a very strange bug. I generated two .avs scripts, both with same dimensions, pixel type and framerate. The only difference between them is that one reads from ImageSource(), while the other generates ColorBars(). The scripts play exactly the same in DirectShow. Yet, when I create AVIS files from them, those play differently in DirectShow.

When I drop them into GraphEdit:
ColorBars AVIS uses the DirectShow ffdshow decoder and plays back correctly. ImageSource AVIS uses the VfW Ffdshow decoder (AVI Decompressor is inserted) and throws a "Graph can't play" error on playback.

You can download samples from here: http://www.citizeninsomniac.com/download/AVIS.7z

Confirmed on tryouts build 1029. Avisynth version is 2.57.

When YV12 decoding is enabled only in VfW mode and YV12 video is played through DirectShow, the image gets mirrored
This apparently only happens when Ffdshow VfW is set to handle YV12 and renders to VMR7. It forces conversion to YUY2 and mirrors the image. If VMR9 is selected, then it forces a conversion to RGB32 but displays it correctly.

I also noticed different behavior for YV12 AVI playback versus YV12 Avisynth playback. The former always forced RGB32 conversion, while the latter seemed to prefer YUY2 ouput.
When YV12 decoding is enabled in both DShow and VfW, YV12 video is always converted to YUY2 when rendered to VMR (i.e. GraphEdit playback).
Most likely related to the behavior above... The major difference between this one and the VfW behavior is that this one always renders to YUY2, with both VMR7 and VMR9.

Interestingly enough, this YV12 issue only seems to affect rendering. When I connected Ffdshow decoder to the input of the WMV9 encoder in GraphEdit, it passed YV12 to the encoder without any problem.

[haruhiko_yamagata]

This is a very strange bug. I generated two .avs scripts, both with same dimensions, pixel type and framerate. The only difference between them is that one reads from ImageSource(), while the other generates ColorBars(). The scripts play exactly the same in DirectShow. Yet, when I create AVIS files from them, those play differently in DirectShow.

When I drop them into GraphEdit:
ColorBars AVIS uses the DirectShow ffdshow decoder and plays back correctly. ImageSource AVIS uses the VfW Ffdshow decoder (AVI Decompressor is inserted) and throws a "Graph can't play" error on playback.

You can download samples from here: http://www.citizeninsomniac.com/download/AVIS.7z

Confirmed on tryouts build 1029. Avisynth version is 2.57.

It seems to be the "path". Double clicking the AVI file launches MPC and plays, but druging a file to MPC's window does not work unless the working directory is configured. I'll look into it further.

[KillerZero]

I used latest beta (rev 1054) and it renders YV12 in MPC if set to VMR7 (windowed). No conversion to YUY2.
When set to VMR7 renderless it output YUY2, but if I disable all output options except YV12 in output tab, it outputs YV12. Maybe just format selection is wrong?

BTW is conversion of interlaced YV12 --> YUY2 correct? It seems little worse than when playing with mplayer -vf ilpack,scale (I have to use scale, because mplayer crashes with YUY2 output for some reason)

[SealTooGreat]

With:
ffdshow_rev1050_20070319_clsid_sse_icl9.exe
ffdshow_beta2a_20070322_clsid.exe
ffdshow_beta2_20070319_clsid.exe
.... using MPC+Haali splitter, I can't watch my anamorphic encodes(it doesn't stretch to correct AR) and also there's some green line in right border.
Using same ffdshow video presets I don't have that issue with:
ffdshow_beta1_20061211_clsid.exe
Any idea why?!

[fastplayer]

@Devs:
Should the download links (in changelog and tryouts homepage) be updated to the new beta2a build?

[KillerZero]

Would be possible to make ffdshow specify DXVA_NominalRange_Wide
and if specified in stream DXVA_VideoTransferMatrix and DXVA_VideoPrimaries?
I think it should solve problems with WMR modes and that Rec601/Rec709 issue, so SD MPEG2 files with Rec709 colorimetry would play correctly without upsampling to HD.
http://msdn2.microsoft.com/en-us/library/ms796493.aspx

[Leak]

To Haruhiko or whoever has a better understanding of MPEG2 decoding:

If I'm using ffdshow to decode DVDs I've noticed that the only events getting fired from the DVD Navigator are BeginFlush/EndFlush, but the video decoder pin doesn't handle those, so I don't think libmpeg2 or anything else is actually waiting for a keyframe after seeking - that might explain why there's still quite some macroblocking going on when seeking around on a DVD.

Incidentally, that also prevents my filter from just dropping the remaining buffered frames and makes it process them, which is quite wasteful - should we perhaps add an onFlush() method (like the existing onStop and onSeek) to TimgFilter(s) that gets called on BeginFlush(), or should we at least call onSeek() in that case?

[foxyshadis]

I swear this thread generates more side threads than anything else in the forum. :p New hardware discussion. (http://forum.doom9.org/showthread.php?t=123809)

KZ, so there actually is a DXVA colorprimary flag? Is it part of VideoInfoHeader2?

Both RFEs have been filed: Use stream information with: ffdshow conversion to RGB (http://sourceforge.net/tracker/index.php?func=detail&aid=1686502&group_id=173941&atid=867363) and passing YUV along to video card (http://sourceforge.net/tracker/index.php?func=detail&aid=1686510&group_id=173941&atid=867363).

[Dr Pizza]

The C parts of libavcodec are definitely LLP64 clean: The only codec that uses (long) at all is the faad wrapper.
I'll take a look at the current source, maybe the situation isn't as grim as it was before.

[clsid]

"make 64BIT=yes" should select the appropriate blocks of assembly code. Note that I haven't been able to test it, but unless I forgot something or screwed up somewhere it should work.

The makefile should also be modified with different GCC parameters to set the correct target architecture. That hasn't been done yet.

[haruhiko_yamagata]

With:
ffdshow_rev1050_20070319_clsid_sse_icl9.exe
ffdshow_beta2a_20070322_clsid.exe
ffdshow_beta2_20070319_clsid.exe
.... using MPC+Haali splitter, I can't watch my anamorphic encodes(it doesn't stretch to correct AR) and also there's some green line in right border.
Using same ffdshow video presets I don't have that issue with:
ffdshow_beta1_20061211_clsid.exe
Any idea why?!It works for me. Which video renderer do you use?
Could you upload the sample?
What is the OS and version of the directShow?
What is the video card?

[haruhiko_yamagata]

@Devs:
Should the download links (in changelog and tryouts homepage) be updated to the new beta2a build?
Yes, please do so.

[haruhiko_yamagata]

To Haruhiko or whoever has a better understanding of MPEG2 decoding:

If I'm using ffdshow to decode DVDs I've noticed that the only events getting fired from the DVD Navigator are BeginFlush/EndFlush, but the video decoder pin doesn't handle those, so I don't think libmpeg2 or anything else is actually waiting for a keyframe after seeking - that might explain why there's still quite some macroblocking going on when seeking around on a DVD.

Incidentally, that also prevents my filter from just dropping the remaining buffered frames and makes it process them, which is quite wasteful - should we perhaps add an onFlush() method (like the existing onStop and onSeek) to TimgFilter(s) that gets called on BeginFlush(), or should we at least call onSeek() in that case?TvideoCodecLibmpeg2:: onSeek and TimgFilterAvisynth:: onSeek are called when we seek on mpeg2 ps/ts/vob and almost any files except DVD:p . It's a bug of the DVD decoding.
I don't think onFlush is necessary as long as DVD can be handled properly.

[Leak]

I don't think onFlush is necessary as long as DVD can be handled properly.
So by "handled properly" you mean onSeek being fired on the next incoming sample when the pin gets a BeginFlush? At least that would be my idea of working around this.

[haruhiko_yamagata]

So by "handled properly" you mean onSeek being fired on the next incoming sample when the pin gets a BeginFlush? At least that would be my idea of working around this.
I don't know. It's not that easy, I guess.
We have to study the code of MPC or DScaler.
I'm not willing to implement full DVD decoding feature.

[albain]

Sorry for the compilation errors : I messed up with the merge (modifications have been made on the files I have been working on, and the merge tool made some weird things...)

It is repaired now : revision 1064

Just one precision : those modifications imply that you have the last version of Windows SDK : v6.0, that includes headers for Vista.
Maybe it will still compile with older SDKs though...

[cc979]

with rev1064

the baseclasses are un-compilable with gcc
tried updating to vista's sdk and converting the libs ... to many errors

like
d:/msys/1.0/dx/Include/objbase.h:1003: error: '__in' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1004: error: '__in' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1005: error: '__in' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1006: error: '__deref_out' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1006: error: initializer expression list treated as compound expression

its beyond me, any clues anyone?

[KillerZero]

[QUOTE=foxyshadis;974086]KZ, so there actually is a DXVA colorprimary flag?
obviously yes.

Both RFEs have been filed: Use stream information with: ffdshow conversion to RGB (http://sourceforge.net/tracker/index.php?func=detail&aid=1686502&group_id=173941&atid=867363) and passing YUV along to video card (http://sourceforge.net/tracker/index.php?func=detail&aid=1686510&group_id=173941&atid=867363).

You filled it wrong.
1) ffdshow always use Rec601 when does SW RGB conversion.
2) more important than DXVA_VideoPrimaries (which set color of RGB values) is DXVA_VideoTransferMatrix (which set which matrix should be used for YV12-->RGB conversion)
Both these values should be set.

[Inventive Software]

@Devs: Can I suggest a separate branch for a Vista so as not to break compilation, so to speak? Vista's screwy already without helping to break other operating systems. ;)

[albain]

with rev1064

the baseclasses are un-compilable with gcc
tried updating to vista's sdk and converting the libs ... to many errors

like
d:/msys/1.0/dx/Include/objbase.h:1003: error: '__in' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1004: error: '__in' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1005: error: '__in' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1006: error: '__deref_out' was not declared in this scope
d:/msys/1.0/dx/Include/objbase.h:1006: error: initializer expression list treated as compound expression

its beyond me, any clues anyone?

Does it compile correctly under MSVC ?

I am using the base classes from ffdshow with the last windows sdk and I didn't have this problem (using MSVC)

I didn't try with gcc : gcc does not work in Vista for now

Concerning the idea of opening a separate branch for Vista, I hope that this could be avoided, because it will split the work of developers.

Things can be done to avoid compilation problems, by using preprocessor macros basing on the OS version. However the last SDK (the one for Vista) should be working under XP

[cc979]

@albain
i don't use MSVC or vista, before i tried updating the mingw headers & libs it did'nt work either

as mingw as its own win32 headers it clashes/or is incompatible with new the windows sdk v6 - mingw headers might get updated later

not sure what to do now - as it a bit over my head

is it possible to alter your edits to keep gcc compatiblity?

or maybe edit it to work with latest dx sdk(feb 2007) but not platform sdk as its that whats breaking gcc - the latest dx sdk is ok with gcc as i used it rev1058 no problem

cheers

[albain]

OK I will try to think about a workaround : maybe by copying the needed content of those headers into a dedicated ffdshow header file.

Maybe your solution is to put your mingw headers directories before windows sdk ones.

Before doing anything, could you try this :
copy functiondiscoverykeys.h and Mmdeviceapi.h to your mingw include directory and remove any references to Windows SDK and compile ffdshow ?

If it does not work, my bet is to extract the used interfaces in my modifications

[cc979]

just tried include them 2 headers but it errors with

d:/msys/1.0/dx/Include/functiondiscoverykeys.h:11: error: expected constructor, destructor, or type conversion before '(' token

d:/msys/1.0/dx/Include/Mmdeviceapi.h:126: error: 'PROPERTYKEY' does not name a type

d:/msys/1.0/dx/Include/Mmdeviceapi.h:198: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:198: error: expected ',' or '...' before 'pwstrDeviceId'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:204: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:204: error: expected ',' or '...' before 'pwstrDeviceId'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:208: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:208: error: expected ',' or '...' before 'pwstrDeviceId'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:212: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:212: error: expected ',' or '...' before 'flow'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:220: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:220: error: expected ',' or '...' before 'pwstrDeviceId'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:343: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:343: error: expected ',' or '...' before '&' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:353: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:353: error: expected ',' or '...' before 'stgmAccess'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:359: error: '__deref_out' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:359: error: expected ',' or '...' before '*' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:363: error: '__out' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:363: error: expected ',' or '...' before '*' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:476: error: '__out' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:476: error: expected ',' or '...' before '*' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:480: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:480: error: expected ',' or '...' before 'nDevice'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:573: error: '__out' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:573: error: expected ',' or '...' before '*' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:654: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:654: error: expected ',' or '...' before 'dataFlow'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:662: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:662: error: expected ',' or '...' before 'dataFlow'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:670: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:670: error: expected ',' or '...' before 'pwstrId'
d:/msys/1.0/dx/Include/Mmdeviceapi.h:676: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:676: error: expected ',' or '...' before '*' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:680: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:680: error: expected ',' or '...' before '*' token
d:/msys/1.0/dx/Include/Mmdeviceapi.h:803: error: '__in' has not been declared
d:/msys/1.0/dx/Include/Mmdeviceapi.h:803: error: expected ',' or '...' before '&' token
make[1]: *** [baseclasses_all.o] Error 1
make[1]: Leaving directory `/home/User/svn/ffdshow-tryout/ffdshow-tryout/src/baseclasses'
make: *** [lib] Error 2


maybe you could try - if not defined them to headers compile for xp or gcc, or something - but looks like it could be tricky problem

cheers

[albain]

Okay, I know why : there are more than those 2 headers, but if I create a new header with only the needed interfaces and constants, it should work

[cc979]

would there be copyright issues distributing ms headers - if remember milan_cutka told me something about it on sourceforge somewhere

[clsid]

The baseclasses in ffdshow are also Microsoft code. So it shouldn't be a problem.

Albain, please create the custom header. Then I could postpone my hardware upgrade a little while till after the Intel price drops :)

[cc979]

i think i got the baseclasses compiled by compiling it on its own but when try build the rest

i get this error
../subtitles/Tstream.h: In member function 'virtual long int TstreamFile::length() const':
../subtitles/Tstream.h:97: error: 'filelength' was not declared in this scope
TbitrateDlg.cpp: In member function 'virtual void TbitrateDlg::init()':
TbitrateDlg.cpp:54: error: 'filelength' was not declared in this scope
make[1]: *** [ffdshow_dialog_all.o] Error 1
make[1]: Leaving directory `/home/User/svn/ffdshow-tryout/ffdshow-tryout/src/dialog'
make: *** [lib] Error 2


not sure what this one is ?

[albain]

I am on it

Sorry to make you stuck guys ;-)

[foxyshadis]

2) more important than DXVA_VideoPrimaries (which set color of RGB values) is DXVA_VideoTransferMatrix (which set which matrix should be used for YV12-->RGB conversion)
Both these values should be set.

Since most of the values map directly to a VUI SEI, I figured supporting any given field would be about as hard as supporting all at once, so most important doesn't really matter.

[_xxl]

Compiling...
ToutputAudioSettings.cpp
src\settings\filters\ToutputAudioSettings.cpp(109) : error C2664: 'IMMDevice::GetId' : cannot convert parameter 1 from 'char_t ** ' to 'LPWSTR * '
Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
src\settings\filters\ToutputAudioSettings.cpp(144) : error C2664: 'IMMDevice::GetId' : cannot convert parameter 1 from 'char_t ** ' to 'LPWSTR * '
Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
src\settings\filters\ToutputAudioSettings.cpp(158) : error C2665: 'strncpy' : none of the 2 overloads can convert parameter 2 from type 'LPWSTR'
C:\Program Files\microsoft visual studio .net 2003\vc7\include\string.h(135): could be 'char *strncpy(char *,const char *,size_t)'
h:\beta\trunk\src\char_t.h(39): or 'wchar_t *strncpy(wchar_t *,const wchar_t *,size_t)'
while trying to match the argument list '(char_t [255], LPWSTR, int)'
src\settings\filters\ToutputAudioSettings.cpp(167) : error C2665: 'strncpy' : none of the 2 overloads can convert parameter 2 from type 'LPWSTR'
C:\Program Files\microsoft visual studio .net 2003\vc7\include\string.h(135): could be 'char *strncpy(char *,const char *,size_t)'
h:\beta\trunk\src\char_t.h(39): or 'wchar_t *strncpy(wchar_t *,const wchar_t *,size_t)'
while trying to match the argument list '(char_t [255], LPWSTR, int)'
Twindow.cpp
src\dialog\Twindow.cpp(146) : error C2065: 'MIIM_STRING' : undeclared identifier
TffdshowDecAudio.cpp
src\TffdshowDecAudio.cpp(101) : warning C4355: 'this' : used in base member initializer list
src\TffdshowDecAudio.cpp(108) : warning C4355: 'this' : used in base member initializer list
src\TffdshowDecAudio.cpp(512) : error C2664: 'IMMDeviceEnumerator::GetDevice' : cannot convert parameter 1 from 'char_t [255]' to 'LPCWSTR'
Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
src\TffdshowDecAudio.cpp(531) : error C2664: 'IFilterGraph::AddFilter' : cannot convert parameter 2 from 'const char [24]' to 'LPCWSTR'
Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
Generating Code...
I get this errors when compiling with msvc2003 + Microsoft SDKs 6, also #define POINTER_64 __ptr64 // <<---- ORROR !!!

typedef void *PVOID;

typedef void * POINTER_64 PVOID64; this issue.

[albain]

Ok, I have made the new header file, and it compiles correctly.
I will do some more tests before committing the changes...

Concerning your compilation errors on MSVS 2003, I am checking this right now. Did you compile it using unicode or ansi ?

[Blight]

Can you please add "zplayer.exe" to the list of allowed players in the installer?
(it isn't listed in the "beta2a" clsid release)

[haruhiko_yamagata]

Can you please add "zplayer.exe" to the list of allowed players in the installer?
(it isn't listed in the "beta2a" clsid release)
It is included in beta2a. The installer load the current settings, so I guess it's your settings. Did you edit the list?

[haruhiko_yamagata]

So by "handled properly" you mean onSeek being fired on the next incoming sample when the pin gets a BeginFlush? At least that would be my idea of working around this.
IPin::NewSegment holds the critical session m_csReceive, while IPin::BeginFlush and IPin::EndFlush can not (must not) hold it. This is a big difference. If onSeenk is called from EndFlush, it can't be stable.

[_xxl]

Did you compile it using unicode or ansi ?
Ansi.

[_xxl]

Compiling...
stdafx.cpp
C:\Program Files\Microsoft SDKs\Windows\v6.0\Include\propkey.h(35) : error C2146: syntax error : missing ';' before identifier 'PKEY_Audio_ChannelCount'
C:\Program Files\Microsoft SDKs\Windows\v6.0\Include\propkey.h(35) : error C2501: 'PKEY_Audio_ChannelCount' : missing storage-class or type specifiers
...
\Include\propkey.h(358) : fatal error C1003: error count exceeds 100; stopping compilation

Build log was saved at "file://h:\beta\trunk\release_unicode\BuildLog.htm"
ffdshow - 102 error(s), 0 warning(s)
Now I get new errors.
MSVC 2003 + SDK 6 + unicode/ansi.

[albain]

I have just committed the changes : there is a new header file called IffMMDevice.h + corrections brought

It compiles correctly on both unicode and ansi without using windows sdk 6.0 headers

Again, sorry for that

If it still doesn't work, I will revert all the changes

[haruhiko_yamagata]

@albain
I'm trying to compile using MSVC8 and "Microsoft Platform SDK for Windows Server 2003 R2".
A few symbols are not defined.
PROPERTYKEY
__RPC__in
__RPC__out
If it still doesn't work, I will revert all the changes
I can wait :D .

[albain]

And what about this one ? (revision 1067)

Thanks

PS : I have renamed my v6 SDK directory and used the old one to confirm the fix. By this way I use the old SDK

[haruhiko_yamagata]

One error left.
REFPROPVARIANT is not defined.
c:\ffdshoww10\trunk\src\iffmmdevice.h(345) : error C2061: error : identifier 'REFPROPVARIANT'

[albain]

Finally ! I have compiled it with my old sdk and it is okay

[Leak]

IPin::NewSegment holds the critical session m_csReceive, while IPin::BeginFlush and IPin::EndFlush can not (must not) hold it. This is a big difference. If onSeenk is called from EndFlush, it can't be stable.
I was more thinking of just setting a flag in Tfilters so onSeek() gets called on the filter chain right before the next frame is handed down to the filters (in Tfilters' process, IIRC) - that should work almost as well, without the synchronization issues.

np: One Second Bridge - Take Me To The Moon (One Second Bridge)

[haruhiko_yamagata]

Finally ! I have compiled it with my old sdk and it is okay
Linker error "_PKEY_Device_FriendlyName". Maybe I should install v6 SDK. That does not matter as long as the output ax is compatible.

// EDIT GCC still have many errors.

[cc979]

@albain something still broke with rev1068

gcc -c -DRELEASE -mno-cygwin -mdll -fno-rtti -mthreads -pipe -D_WINGDI_ -DUCLIBCPP -D_GLIBCPP_HAVE_MBSTATE_T -D_WIN32_IE=0x0500 -mmmx -w -DNDEBUG -UDEBUG -DFFDEBUG=0 -I. -I.. -Iuclibc++ -Ibaseclasses -I../baseclasses -IimgFilters -I../imgFilters -Implayer -I../mplayer -Isettings -I../settings -Isettings/filters -I../settings/filters -Icodecs -I../codecs -Isubtitles -I../subtitles -Iconvert -I../convert -Idialog -I../dialog -IaudioFilters -I../audioFilters -Icygwin -I../cygwin -Iffmpeg -I../ffmpeg -Iacm -I../acm -Ixiph -I../xiph -Ifilters -I../filters -Imuxers -I../muxers -I/dx/Include -L/dx/MingLib -ldx9 -O2 -march=pentium-mmx -mtune=i686 -fomit-frame-pointer -finline-functions -finline -frename-registers -fweb -funit-at-a-time -MMD -o baseclasses_all.o baseclasses_all.cpp
../IffMmdevice.h:464: error: '__in' has not been declared
../IffMmdevice.h:464: error: expected ',' or '...' before 'pwstrDeviceId'
../IffMmdevice.h:470: error: '__in' has not been declared
../IffMmdevice.h:470: error: expected ',' or '...' before 'pwstrDeviceId'
../IffMmdevice.h:474: error: '__in' has not been declared
../IffMmdevice.h:474: error: expected ',' or '...' before 'pwstrDeviceId'
../IffMmdevice.h:478: error: '__in' has not been declared
../IffMmdevice.h:478: error: expected ',' or '...' before 'flow'
../IffMmdevice.h:486: error: '__in' has not been declared
../IffMmdevice.h:486: error: expected ',' or '...' before 'pwstrDeviceId'
../IffMmdevice.h:609: error: '__in' has not been declared
../IffMmdevice.h:609: error: expected ',' or '...' before '&' token
../IffMmdevice.h:619: error: '__in' has not been declared
../IffMmdevice.h:619: error: expected ',' or '...' before 'stgmAccess'
../IffMmdevice.h:625: error: '__deref_out' has not been declared
../IffMmdevice.h:625: error: expected ',' or '...' before '*' token
../IffMmdevice.h:629: error: '__out' has not been declared
../IffMmdevice.h:629: error: expected ',' or '...' before '*' token
../IffMmdevice.h:742: error: '__out' has not been declared
../IffMmdevice.h:742: error: expected ',' or '...' before '*' token
../IffMmdevice.h:746: error: '__in' has not been declared
../IffMmdevice.h:746: error: expected ',' or '...' before 'nDevice'
../IffMmdevice.h:839: error: '__out' has not been declared
../IffMmdevice.h:839: error: expected ',' or '...' before '*' token
../IffMmdevice.h:920: error: '__in' has not been declared
../IffMmdevice.h:920: error: expected ',' or '...' before 'dataFlow'
../IffMmdevice.h:928: error: '__in' has not been declared
../IffMmdevice.h:928: error: expected ',' or '...' before 'dataFlow'
../IffMmdevice.h:936: error: '__in' has not been declared
../IffMmdevice.h:936: error: expected ',' or '...' before 'pwstrId'
../IffMmdevice.h:942: error: '__in' has not been declared
../IffMmdevice.h:942: error: expected ',' or '...' before '*' token
../IffMmdevice.h:946: error: '__in' has not been declared
../IffMmdevice.h:946: error: expected ',' or '...' before '*' token
../IffMmdevice.h:1069: error: '__in' has not been declared
../IffMmdevice.h:1069: error: expected ',' or '...' before '&' token
make: *** [baseclasses_all.o] Error 1

must be a difference between msvc and gcc

cheers

[haruhiko_yamagata]

I was more thinking of just setting a flag in Tfilters so onSeek() gets called on the filter chain right before the next frame is handed down to the filters (in Tfilters' process, IIRC) - that should work almost as well, without the synchronization issues.

np: One Second Bridge - Take Me To The Moon (One Second Bridge)
EndFlush may be called while ffdshow is decoding the last sample of the current segment. In that case,I'm afraid the filter will buffer the last frame of the current segment as the first frame of the coming segment.

[haruhiko_yamagata]

If GCC cannot build Vista application, conditional compilation is the choice. It's not good idea to revert your changes, IMO.

#ifdef __GNUC__

#else

#endif

[haruhiko_yamagata]

This SDK is designed for use with Windows Vista (which includes Framework 3.0). The Windows SDK for Vista also supports creating applications for Windows XP, Windows Server® 2003 SP1, and Windows Server 2003 R2.

This release of the Windows SDK is compatible with Microsoft® Visual Studio® 2005 and the Visual Studio 2005 extensions for .NET Framework 3.0 (WCF & WPF), November 2006 CTP.
Is the object compatible with Windows 2000? We may consider dropping Windows 9x, but I still want to support Windows 2000.

Is MSVC7 compatible with this SDK?

None the less, supporting Vista is important too. Conditional compilation should be something like...

#ifdef __GNUC__
#undef VISTA_SPDIF
#endif

#ifdef VISTA_SPDIF

#else

#endif

[albain]

Ok, I will add those conditional compilation macros

[foxyshadis]

Vista SDK is compatible with 2000 (and some still for 98), as long as you don't use XP-specific APIs, but Microsoft no longer does more than minimal testing with EOL'd OSes, so they don't get mentioned. That was my experience with simpler apps, haven't compiled ffdshow in a week or two.

[albain]

I hope that a better integration of Vista sdk with gcc is possible.

Anyway, I have just committed changes with conditional compilation of Vista features

As proposed, the preprocessor macro VISTA_SPDIF has to be defined in order to compile those features.
It is always disabled for gcc.

Revision 1069

Cheers

[cc979]

is there a lot of extra stuff on sdk v6, compared to dx sdk feb07 (as it does support dx10)

ps. ffdshow builds fine now with gcc

cheers

[cc979]

just been testing the .ts clip from
http://forum.doom9.org/showpost.php?p=974692&postcount=539

it something expands correcting in vlc but in mpc with ffdshow/vmr9 it expands vertical, weird?

[clsid]

Compiled with MSVC2003 and Platform SDK 5.2 (2003 R2):

VISTA_SPDIF defined:

ffdshow error LNK2019: unresolved external symbol _PKEY_Device_FriendlyName referenced in function "protected: class std::hash_map<class DwString<wchar_t>,class DwString<wchar_t>,struct std::hash<class DwString<wchar_t> >,struct std::equal_to<class DwString<wchar_t> >,class std::allocator<class DwString<wchar_t> > > __thiscall ToutputAudioSettings::getDevices(void)" (?getDevices@ToutputAudioSettings@@IAE?AV?$hash_map@V?$DwString@_W@@V1@U?$hash@V?$DwString@_W@@@std@@U?$equal_to@V?$DwString@_W@@@3@V?$allocator@V?$DwString@_W@@@3@@std@@XZ)


VISTA_SPDIF not defined:

compiles succesfully!

[haruhiko_yamagata]

just been testing the .ts clip from
http://forum.doom9.org/showpost.php?p=974692&postcount=539

it something expands correcting in vlc but in mpc with ffdshow/vmr9 it expands vertical, weird?
My result with MPC's internal source filter is,
vmr7/9 beta1 OK,beta2a OK,ffdshow-20051129 OK,DScaler NG,MPC NG
overlay beta1 NG,beta2a NG,ffdshow-20051129 NG,DScaler NG,MPC NG

ffdshow used libmpeg2. MPC means MPC's internal decoder.
NG means no AR change during playback.
Do you mean vertically streched during playback?
What is the video card?

[cc979]

i have a nvidia 6800 card btw

mpc setting repeat for ever
on first play there is no AR change, then after the repeat it on the change it stretchs vertical

i have taken snapshots using mpc they saving with the corrected AR changes - so maybe it is mpc that has the problem

[cc979]

using windows media play with the same clip, on AR change the window remains the same but the video shrinks vertical(with black borders) to the correct AR

[cc979]

the screenshots and printscreens, so show whats happening
http://www.mytempdir.com/1267661

[albain]

Compiled with MSVC2003 and Platform SDK 5.2 (2003 R2):

VISTA_SPDIF defined:



VISTA_SPDIF not defined:

compiles succesfully!

At least there is one config with which it is working :-)

[Leak]

EndFlush may be called while ffdshow is decoding the last sample of the current segment. In that case,I'm afraid the filter will buffer the last frame of the current segment as the first frame of the coming segment.
Maybe so, but I'd rather have 1 sample being needlessly shown than 10 or so of them. Anyway, I'll try playing around with this a bit and see what happens; I could also just drop the first frame after a flush since seeking is inexact anyway *and* the graph is obviously already running.

(Note that I'd of course only do all of this if the previous filter in the DirectShow graph was the DVD Navigator - it works fine with everything else already...)

np: Alias & Tarsier - 5 Year Eve (Neotropic Remix) (Plane That Draws A White Line)

[albain]

is there a lot of extra stuff on sdk v6, compared to dx sdk feb07 (as it does support dx10)

ps. ffdshow builds fine now with gcc

cheers

directx sdk no more includes direct show, it is now located within windows sdk


What's new in v6 (I have listed only the interesting stuff)
- core Audio APIs (the new audio stack from Vista)
- new directshow:
1/The new APIs and interfaces replacing VMR9 : EVR (Enhanced Video Renderer), supposed to be better than VMR9
2/DirectX Video Acceleration V2 (DXVA 2) improves hardware acceleration and better direct3D interaction
3/Support for WDDM capture devices (not useful for ffdshow)
4/Better MPEG1 layer 2 decoder
5/Better DVD playback
6/Audio end point devices

[Dr Pizza]

Hrrrm. Is there something special I have to do to make it work?

I'm building in VC++ 2005, using the Vista Platform SDK, Release, Unicode, x64. I had to make a couple of minor changes to one file to comment out some inline assembler, which isn't supported in x64 (I realize it'll probably break if that codepath is hit, but I'm not using the code in question anyway), and I had to change an operator new overload to use "size_t" instead of "unsigned", but with that done, it's all building and compiling apparently correctly. The .ax registers properly, and the properties pages work fine. However, it won't join any filter graphs. It's definitely doing something (because I can see ffdshow getting loaded to generate thumbnails in Explorer) but it won't join my filter graph in my 64-bit media playing app.

If I try to construct a graph manually in graphedt, ffdshow refuses to join the graph, claiming "No combination of intermediate filters could be found to make the connection. (Return code: 0x80040217)". I'm guessing that something's broken, but I'm not really sure what.

[cc979]

directx sdk no more includes direct show, it is now located within windows sdk


What's new in v6 (I have listed only the interesting stuff)
- core Audio APIs (the new audio stack from Vista)
- new directshow:
1/The new APIs and interfaces replacing VMR9 : EVR (Enhanced Video Renderer), supposed to be better than VMR9
2/DirectX Video Acceleration V2 (DXVA 2) improves hardware acceleration and better direct3D interaction
3/Support for WDDM capture devices (not useful for ffdshow)
4/Better MPEG1 layer 2 decoder
5/Better DVD playback
6/Audio end point devices

theres a few things then, maybe i will upgrade next year when its more stable - or maybe sooner if i'm forced to because dx10 and games

[SealTooGreat]

With:
ffdshow_rev1050_20070319_clsid_sse_icl9.exe
ffdshow_beta2a_20070322_clsid.exe
ffdshow_beta2_20070319_clsid.exe
.... using MPC+Haali splitter, I can't watch my anamorphic encodes(it doesn't stretch to correct AR) and also there's some green line in right border.
Using same ffdshow video presets I don't have that issue with:
ffdshow_beta1_20061211_clsid.exe
Any idea why?!
It works for me. Which video renderer do you use?
Could you upload the sample?
What is the OS and version of the directShow?
What is the video card?
1. Haali renderer(7-march-2007)
2. sample(6mb) (http://depositfiles.com/files/724695)
3. Windows XP Professional SP2. Don't know which software DirectShow version you've meant.
4. Nvidia GeForce 7300 GS.

[Blight]

haruhiko_yamagata:
The previous version I had installed was build 584 (clsid), which was a clean install.

I then tried installing beta2a and was presented with a list that did not contain zplayer.exe. I'm afraid that if Zoom Player users install this version it will no longer work.

I'm not sure that loading old settings for the "work with application" value is a good idea for future compatibility.

[Leak]

I'm not sure that loading old settings for the "work with application" value is a good idea for future compatibility.
Well, just blowing it's old settings away probably isn't a good idea either...

[Blight]

leak:
audio/decoder/processing settings should stay as that makes the upgrade have less of an impact in case new default-on settings may cause unforseen issues. Loading the application whitelist setting from a previous installation means that any new whitelisted application introduced in future ffdshow versions will not work ... which makes no sense.

[Dr Pizza]

Ooops, I was being a dumbass. I hadn't noticed that the libavcodec hadn't built properly. The Win32 project excludes a number of source files (that won't work in VC++, I guess), and so builds the DLL properly. The x64 project doesn't exclude them, so fails to produce a DLL. Excluding the same set of source files allowed the x64 project to emit a file, resulting in a functional ffdshow.

[Leak]

Loading the application whitelist setting from a previous installation means that any new whitelisted application introduced in future ffdshow versions will not work ... which makes no sense.
Well, what about entries the user has made him/herself? Dropping those won't make the user very happy either...

[clsid]

Maybe it would be better to disable the whitelist feature by default. At least the first few months. So that a more comprehensive list of compatible apps can be gathered.

[skystrife]

Ooops, I was being a dumbass. I hadn't noticed that the libavcodec hadn't built properly. The Win32 project excludes a number of source files (that won't work in VC++, I guess), and so builds the DLL properly. The x64 project doesn't exclude them, so fails to produce a DLL. Excluding the same set of source files allowed the x64 project to emit a file, resulting in a functional ffdshow.

Do you think you could make an installer so some of us can start beta testing it?

[_xxl]

I hadn't noticed that the libavcodec hadn't built properly. The Win32 project excludes a number of source files (that won't work in VC++, I guess), and so builds the DLL properly.
Libavcodec shouldn't be compiled by MSVC.MS can't compile GCC asm syntax.Only GCC can.
Maybe it would be better to disable the whitelist feature by default.
We should add a link to ffdshow-tryouts page in Setup.

[haruhiko_yamagata]

haruhiko_yamagata:
The previous version I had installed was build 584 (clsid), which was a clean install.

I then tried installing beta2a and was presented with a list that did not contain zplayer.exe. I'm afraid that if Zoom Player users install this version it will no longer work.
zplayer.exe is at the botom of long scroll list(alphabetical order).
Please tell me what applications are listed there.

I'm not sure that loading old settings for the "work with application" value is a good idea for future compatibility.Yes, better implementation is required. Not just over-writing, add new entries with user's permision.

[haruhiko_yamagata]

Maybe it would be better to disable the whitelist feature by default. At least the first few months. So that a more comprehensive list of compatible apps can be gathered.
Gathering apps exe name and checking compatibility is very very hard work. Unless we set it by default, we won't have any reports.

[wiak]

Is there really a need for x64 build of ffdshow? I mean, it won't be any faster than normal 32-bit build and x64 is compatible with normal builds.

YES there is need
Vista Media Center x64 has problems with 32bit codecs, so it needs a x64 version

[_xxl]

I don't know how much faster x64 will be.Because asm code is written for x86 32bit.New x64 asm code is required.

[Dr Pizza]

Do you think you could make an installer so some of us can start beta testing it?

I've no idea how to make the installer, and someone ought to sanity-check the changes I had to make.

[Leak]

I've no idea how to make the installer, and someone ought to sanity-check the changes I had to make.
Install InnoSetup, then open the bin\distrib\InnoSetup\ffdshow_installer.iss file with it and compile it.

Dunno if that works out-of-the-box for x64, though.

[multiblitz]

I guess this is a bug:

If I use ffdshow (rev 1020) to playback VC-1 (HD-DVD) (using Haali's splitter, AnyDVD and ZP), I get a white picture (pure white screen) if I don't resize the picture. With resize (lanzos), I have a nice picture. But I don't want to resize (1920*1080 is already there...)

BIggest Bug so far: If ffdshow is used in the video-part (does not matter if it is only postprocessing-raw-filter after the original MS VC-1 Codec or directly the ffdshow-VC-1-decoder): Instead of correct speed of 23.976 fps, it recognizes 29.999 fps. So, the video is much faster than the audio which indeed is the case...without ffdshow in the video-chain, with pure MS VC-1, it works, eventhough SPDIF is still somewhat jaggy (core duo 6400 at 3.6 ghz).

Any ideas how to solve this ? Has anyone actually looked into this at all ?

[Blight]

haruhiko_yamagata:
doh, it is listed, it's just that the display list crops the bottom item if there's no room to display a line (this is a weird behavior for a list type display in windows), so it looked to me as if there were only 9 files listed.

leak:
the ratio of users manually adding programs to the white list compared to future white-list additions not being included and stuff not working is pretty high. most users don't change any setting and just click next next next during installation.

[DeepBeepMeep]

FFdshow is still incompatible with EVO file containing VC1 parsed with Haali filter (video is accelerated and audio / video gets out of sync). Is there any place to fix this issue in the next beta? It is a bit sad to be able to use ffdshow here.
Thanks for your support.

[_xxl]

Is this a ffdshow issue?

[multiblitz]

Is this a ffdshow issue?

Just read above, it's the same bug I have now reported twice: BIggest Bug so far: If ffdshow is used in the video-part (does not matter if it is only postprocessing-raw-filter after the original MS VC-1 Codec or directly the ffdshow-VC-1-decoder): Instead of correct speed of 23.976 fps, it recognizes 29.999 fps.

So, even if you don't use fdshow for VC-1-decoding, simply if it is in the filter-chain, you have this bug.

[DeepBeepMeep]

Is this a ffdshow issue?


It seems to be an ffdshow issue. I have exchanged a few messages with Haali who didn't find any issue with his filter. What seems to indicate that this a ffdshow bug is that like multiblitz mentioned it, if you don't use ffdshow as a VC1 decoder but only as a postprocessor filter you get the issue while the same VC1 decoder alone without ffdshow works fine.

My theory is that the EVO file contains some telecine flags such as 'repeat top' which are supposed to increase every other frame by 50%. This is supposed to enable a 'soft' 3:2 pulldown if a movie is not watched on an interlaced display.

Therefore if these flags are taken into account you get 23.976 frames per second. If these flags are ignored / lost on the way, every frame lasts the same duration and at the end more frames fits in the same second and we have 29.97 frames per second.

Basically what calculates time stamps should take into account these flags or ffdshow should forward the flags (AM_VIDEO_FLAG_FIELD1FIRST / AM_VIDEO_FLAG_REPEAT_FIELD?) to the next filter in the chain.

Just my 2 cents...

[haruhiko_yamagata]

just been testing the .ts clip from
http://forum.doom9.org/showpost.php?p=974692&postcount=539

it something expands correcting in vlc but in mpc with ffdshow/vmr9 it expands vertical, weird?
Anybody who can reproduce this, please help.
I think this kind of problem is really important but I can't reproduce.
Is this environment specific?
If so, what does it depend?

[fastplayer]

Anybody who can reproduce this, please help.
I think this kind of problem is really important but I can't reproduce.
Is this environment specific?
If so, what does it depend?
System default renderer:
Sample expands correctly (horizontally) when using MPC+ffdshow's libmpeg2. When using MPC's internal libmpeg, then AR stays the same.
VMR9:
With MPC+ffdshow, the AR changes (slightly bigger black bars at top and bottom of video), but the window doesn't change its size. If I manually seek to the beginning of the sample, then window size changes (vertically, much bigger black bars). In both cases AR is wrong.
When using MPC's internal decoder, the same happens as with VMR7 (AR stays the same throughout the sample).

[Dr Pizza]

Install InnoSetup, then open the bin\distrib\InnoSetup\ffdshow_installer.iss file with it and compile it.

Dunno if that works out-of-the-box for x64, though.

OK, I'm trying that.

I'm not sure I like the look of the install script, though. It looks like it wants to include app manifests and some other detritus, which doesn't seem right; one, because the manifests are all hardcoded to use x86 dependencies (which isn't right for x64, obviously), two, because they should all be embedded anyway (VC8 embeds manifests by default).

I'm also not compiling anything with gcc; according to the setup script this can cause problems due to the VC8 runtime without the VC8 redist installed (i.e. using private assemblies). I'm not bothered about that, to be honest; it doesn't seem particularly onerous to require people to install the redist. There is a small gotcha; the redist on MS's site hasn't been updated for VC8 SP1 (or at least, wasn't when I last checked).

To address that, I will try embedding the VC8 SP1 redist into the installer, though I don't know how to make sure it does stuff in the right order; it needs to run the redist install before it tries to regsvr32 the .ax.

[haruhiko_yamagata]

OK, I'm trying that.
...
two, because they should all be embedded anyway (VC8 embeds manifests by default).

It's embedded, none the less, Windows Xp32 requires external manifest to get Xp visual style work. Vista does not have this bug. I don't know about Xp64.
I'm also not compiling anything with gcc; according to the setup script this can cause problems due to the VC8 runtime without the VC8 redist installed (i.e. using private assemblies). Why do you think so? The installer has been tested and works. What's wrong with using private assemlbies?

If you want to install VC8 SP1 redist, you have to uninstall VC8 sp- redist (when I tested before, sp1beta). I don't think it's a good idea.
I'm not bothered about that, to be honest; it doesn't seem particularly onerous to require people to install the redist.Maybe true with x64 builds, x64 users are experts.

//EDIT
As for x64, ffavisynth.dll may require shared assembly or private assembly for it's own directory. In win32, it's compiled by MSVC7, but in x64, MSVC7 can't compile.

[fastplayer]

It's embedded, none the less, Windows Xp32 requires external manifest to get Xp visual style work.
They can be external or internally embedded as a resource.
I don't know about Xp64.
Changing the line processorArchitecture="X86" to processorArchitecture="*" worked when I was .NETting a while ago...
If you want to install VC8 SP1 redist, you have to uninstall VC8 sp- redist (when I tested before, sp1beta).
Can somebody upload the VC8 SP1 redist? I can't find it on microsoft.com...

[haruhiko_yamagata]

They can be external or internally embedded as a resource.
Well, just install a build using MSVC8, remove ffdshow.ax.manifext and launch the dialog, you'll find Xp visual style is gone even though ffdshow.ax have a proper embedded manifest as a resource. I think it's a bug of Xp. The manifest loading CRT is processed properly in Xp without external manifest. So Xp requires both embedded and external manifest.

As for the AR problem, I found a similar problem using VMR9 renderless-mixer mode(MPC).

[fastplayer]

Where can I get a VC8 build of ffdshow? Your builds are not available anymore on sourceforge...

Edit: Found one on http://ffdshow.wyrdic.net/.
Edit: Regarding the VS problems. Please check out 6th post here:
http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=144703&SiteID=1

[clsid]

I will make some adjustments to the install script so that it can be used to install 64-bit binaries.

Compiling libavcodec with MSVC will probably produce a partially broken binary. Certainly if some asm code must be disabled. There are also two ToDo's that need to be fixed in ffv1.c

[Dr Pizza]

It's embedded, none the less, Windows Xp32 requires external manifest to get Xp visual style work. Vista does not have this bug. I don't know about Xp64.
I don't think that should be the case; I'm sure I've seen themed programs that use only an embedded manifest, just as long as the manifest is correct. The manifests

Why do you think so? The installer has been tested and works. What's wrong with using private assemlbies?
They're hard to patch for e.g. security updates.

If you want to install VC8 SP1 redist, you have to uninstall VC8 sp- redist (when I tested before, sp1beta). I don't think it's a good idea.
Is that actually true, and if it is true, is that actually a problem? None of my VC8 programs have broken since installing VC8SP1, and my SxS directory seems to have both RTM and SP1 runtime libraries, so I think they can coexist fine.

As for x64, ffavisynth.dll may require shared assembly or private assembly for it's own directory. In win32, it's compiled by MSVC7, but in x64, MSVC7 can't compile.
How would I test that?

The only thing I've found that's obviously broken in x64 is the OSD; the display is completely corrupt. This isn't new, though; the old ffdshow x64 build has the same problem, so whatever's broken has been broken for many months.

[Dr Pizza]

Where can I get a VC8 build of ffdshow? Your builds are not available anymore on sourceforge...

Edit: Found one on http://ffdshow.wyrdic.net/.
Edit: Regarding the VS problems. Please check out 6th post here:
http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=144703&SiteID=1

Ah, I only ever build unicode programs (multibyte just needs to go away and die), so I guess I've always had the right manifests embedded correctly.

[Dr Pizza]

I will make some adjustments to the install script so that it can be used to install 64-bit binaries.

Compiling libavcodec with MSVC will probably produce a partially broken binary. Certainly if some asm code must be disabled. There are also two ToDo's that need to be fixed in ffv1.c

The TODOs are commented out, because CONFIG_ENCODERS isn't defined in the project.

Both TODOs relate to C99 VLAs; there are other places in libavcodec that uses them, only they're *not* commented out. I don't think fixing them should be too hard; finding them might be trickier.

[fastplayer]

AFAIK, Haruhiko's builds were all ANSI. That could explain the manifest issue (see link above or here (http://www.codeguru.com/forum/showthread.php?t=386815&highlight=manifest))
It's all so much easier in a managed environment though :D

[clsid]

The TODOs are commented out, because CONFIG_ENCODERS isn't defined in the project.

Both TODOs relate to C99 VLAs; there are other places in libavcodec that uses them, only they're *not* commented out. I don't think fixing them should be too hard; finding them might be trickier.
CONFIG_ENCODERS gets defined in config.h


I have updated the install script to support 64-bit installation.

[Dr Pizza]

CONFIG_ENCODERS gets defined in config.h
Weird. Maybe the IDE is getting confused, 'cos it's adamant that it's not defined. Not the first time IntelliSense loses touch with reality, of course. Anyway, I think I have put together some code that accommodates the lack of VLAs in VC8. Though I'm not 100% sure, as I totally hate C array/pointer declaration syntax.

[Dr Pizza]

Do VFW files really need to go in system32 and not the app dir?

[Dr Pizza]

Can somebody upload the VC8 SP1 redist? I can't find it on microsoft.com...

x64:
http://www.file.am/d/file/vcredist_x64.exe/
x86:
http://www.file.am/d/file/vcredist_x86.exe/

[fastplayer]

x64:
http://www.file.am/d/file/vcredist_x64.exe/
x86:
http://www.file.am/d/file/vcredist_x86.exe/
:thanks:

[clsid]

Do VFW files really need to go in system32 and not the app dir?
It is allowed to put it anywhere you like. However some applications might assume it is always located in system32. This could lead to some problems. But putting it in the appdir would make the runtime issue a bit less complicated.

In the registry section you need to replace "ff_vfw.dll" with "{code:GetShortName|ExpandConstant('{app}\ff_vfw.dll'))}"

[Leak]

Okay, here's yet another updated version of my patch:

Download (http://leak.no-ip.org/AviSynth/ffdshow/) (MSVC build)
Patch (https://sourceforge.net/tracker/index.php?func=detail&aid=1676882&group_id=173941&atid=867362) (against rev. 1074)

I've added an AviSynth info item to ffdshow's OSD, which shows the number of back and ahead buffers used since the last reset (i.e. stop, seek, size change or similar); that should help a bit with finding the correct settings. It also shows the input frame number that's considered "current" for the generated frame and the frame numbers that were requested by the AviSynth script to produce it.

If the current settings are too small it doesn't really matter, unless the filter doesn't work right if it doesn't get the frame it really requested. In that case, just up the buffers a bit and see what happens. Also, seeking around in the stream and/or letting it run for a bit might help finding edge cases where the script needs more frames.

(Note that those values will go up when the buffers are flushed at the end of an MPEG2 stream, but that's because there is nothing left to be buffered and can be safely ignored.)

The values can also be logged to a CSV file by checking "Save to" in ffdshow's OSD options.

The buffer defaults were changed to 0/10/0 and Apply Pulldown is turned off by default for backwards compatibility, but of course at those settings it also doesn't really support (pseudo-)random access - but the OSD info should make tweaking the buffer settings a lot easier... :)

I've also added some code to reset AviSynth on a BeginFlush on the DirectShow video input pin to work around a DVD Navigator bug where seeking doesn't cause an onSeek event, and without knowing that the stream was seeked the frames that were buffered before the seek get processed and played after the seek at full tilt, which looks weird and eats extra CPU time.

@haruhiko:
I've added an onFlush method in Tfilters and Tfilter, since that is the easiest and probably cleanest way to handle this (at least concerning the amount of code that needs to be changed and/or checked), although like you said doing anything in them other than flagging that there was a flush is probably a bad idea due to race conditions that may occur then. The flush itself is then handled when process() is next called.

np: Christ - Vernor Vinge (Blue Shift Emissions)

[Leak]

Oh yeah, I noticed that (at least) the outline strength and radius sliders on the OSD font configuration page always get reset to 0 as opposed to reflecting the current settings.

[marcellus]

Hi
I use ffdshow for to encode to mpeg2 (the great libavcodec encoder). I live in a PAL country so I usually encode to 25 fps framerate. But recently I tried to encode some Quicktime trailers (the 300 movie taken from Apple site) and I get in VDub an error:
Cannot start video compression:
An unknown error occured (may be corrupt data).
(error code -100)
After much trial and error I come to the conclusion that the framerate is the culprit. This Avisynth script works well:
Blackness(300,640,480,"yv12",25)
but this one gives the error:
Blackness(300,640,480,"yv12",23.976)
I noticed that encoding to mpeg4 gives the error too. I tried several ffdshow versions, including latest beta 2a and latest icl9 build, with the same result.

[clsid]

Read the first post of this topic for the answer.

[marcellus]

I see.
:thanks:

Any ideea, can be fixed in the future? Not that it can't be worked around the way it is now, I can change fps in avisynth to 25 fps an then change it back in restream but would be nice to work with any input framerate.

[cc979]

after further tests using xvid anamorphic clips with ffdshow rev1069

this is what happens with AR changes in avi's and mp4's

http://www.mytempdir.com/1269513

test1 = anamorphic-source.avi+xvid-par16x9-dar1x1.avi
test2 = anamorphic-source.avi+xvid-par1x1-dar16x9.avi

in vlc

xvid-par16x9-dar1x1.avi: AR works
xvid-par1x1-dar16x9.avi: AR works
container16x9-bitstream5x4.mp4: AR works
container-unspecified-bitstream16x9.mp4: AR works
test1: AR works
test2: AR works

in mpc/ffdshow

xvid-par16x9-dar1x1.avi: AR does not work
xvid-par1x1-dar16x9.avi: AR does not work
container16x9-bitstream5x4.mp4: AR works
container-unspecified-bitstream16x9.mp4: AR works
test1: AR does not work
test2: AR does not work

not sure if the naming scheme is correct, using xvid using the 2 types of 16x9 - 1 using PAR 16/9 and other DAR 16/9

and the mp4's i used yamp and told it to use DAR 16/9

cheers all

ps. using render overlay/haali/vmr9-renderless
but with vmr9-windowed all is fine

[Rectal Prolapse]

DeepMeepBeep said:

What seems to indicate that this a ffdshow bug is that like multiblitz mentioned it, if you don't use ffdshow as a VC1 decoder but only as a postprocessor filter you get the issue while the same VC1 decoder alone without ffdshow works fine.

I can confirm this - everything gets reported as 29.97 by FFDShow when it is attached to a VC1 decoder in the chain. Verified by Reclock and VMR9.

[Fizick]

I try to use ffdshow for realtime encoding to MPEG2 in AVI (constant bitrate, lavcodec).
I found that I must set "decoder bitsteam buffer size" to big value (about 10000000) and "maximum bitrate" to non-zero at "ratecontrol" tab. In other cases encoder freeze my system. may be it is possible somehow protect user from wrong settings. For example, use mean bitrate as a max bitrate, and auto-enlarge buffers to safe value (how to calculate it?).

[haruhiko_yamagata]

AFAIK, Haruhiko's builds were all ANSI. That could explain the manifest issue (see link above or here (http://www.codeguru.com/forum/showthread.php?t=386815&highlight=manifest))
It's all so much easier in a managed environment though :D
ANSI or UNICODE does not mater. Leak's builds are UNICODE, but it have the same behavior.

[ferchon03]

Hello, I have a problem and I can't find a solution.
Every time I close the ffdshow video or audio decoder configuration I get this error:
"RunDLL An exception occurred while trying to run 'ffdshow.ax,configure'."
I've installed XP SP2 and the latest ffdshow from free-codecs.com.
Anyone know how I can fix it?

Thanks in advance.

[Dr Pizza]

The only thing I've found that's obviously broken in x64 is the OSD; the display is completely corrupt. This isn't new, though; the old ffdshow x64 build has the same problem, so whatever's broken has been broken for many months.
Oh, now I see that the OSD uses MMX. Didn't think of that. I'm beginning to implement the MMX stubs (using SSE2 integer ops), as that seems to be the easiest way to move forward (as it means the code using MMX intrinsics can remain unchanged). The code is now better than it was, but there are still visual problems with the OSD. It's readable, kinda, but something's getting screwed up with the outlines. I might be doing some arithmetic backwards or something. Though I haven't compared to see how it looks in a 32-bit build, so maybe not.

I've only put in the functions needed for the OSD so far, because it's easy to tell when they're working and when they're not. I don't know what the other things (kernelDeint, TomsMoComp) should do, so I don't know how to tell if they're working or not.

Edit: aha, the problem was I was using a signed function for something that should have been unsigned. With that fixed I now have a properly working OSD.

[Leak]

Every time I close the ffdshow video or audio decoder configuration I get this error:
"RunDLL An exception occurred while trying to run 'ffdshow.ax,configure'."
I've installed XP SP2 and the latest ffdshow from free-codecs.com.
Anyone know how I can fix it?
That usually happens if you have more than one version of ffdshow installed somewhere and configure brings up the config dialog of an older installation (i.e. not the one you installed last).

Look where the links are pointing to and make sure they're pointing at the directory where you last installed ffdshow, and probably clean up older installations if that's what's causing this... :)

Removing all ffdshow versions and then installing just one should also fix it, I guess.

[SealTooGreat]

@haruhiko_yamagata
HERE (http://www.mytempdir.com/1270174) is the ffdshow video preset that i use, referring my previous post (http://forum.doom9.org/showthread.php?p=974892#post974892)...btw I've tried ffdshow_rev1074_20070325_clsid_sse_icl9 and it doesn't strech my anamorphic encodes (xvid/ac3/avi) while keeping green line in the right border.

[cc979]

@haruhiko_yamagata
HERE (http://www.mytempdir.com/1270174) is the ffdshow video preset that i use, referring my previous post (http://forum.doom9.org/showthread.php?p=974892#post974892)...btw I've tried ffdshow_rev1074_20070325_clsid_sse_icl9 and it doesn't strech my anamorphic encodes (xvid/ac3/avi) while keeping green line in the right border.

what video card have you got ?

[haruhiko_yamagata]

@haruhiko_yamagata
HERE (http://www.mytempdir.com/1270174) is the ffdshow video preset that i use, referring my previous post (http://forum.doom9.org/showthread.php?p=974892#post974892)...btw I've tried ffdshow_rev1074_20070325_clsid_sse_icl9 and it doesn't strech my anamorphic encodes (xvid/ac3/avi) while keeping green line in the right border.
Thank you for your preset, I'm begining to understand.
I found milan's 20051115 works, 20051129 does not work.
beta1 works? I can't reproduce this.

It's "Output->Allow output format change during playback".
3 mode for this was introduced between 20051115 and 20051129. Intermediate state does not work. It works with it checked and "Connect to compatible filters only" unchecked. If "Connect to compatible filters only" is checked MPC fails over VMR on Windows Xp.
Haali's renderer is not working for me (probably because of my hardware). I have black screen. The AR changes correctly with 20051115 though. So my testing is not reliable.

Should Haali's renderer be added to compatible filters?
Please test rev 1075 with "Output->Allow output format change during playback" checked and "Connect to compatible filters only" unchecked.

With the setting above, beta1 works, beta2a does not work. So I think I have understand your issue :D.

[wozio]

Hi,

I have a question: is it possible to use VMR9 deinterlacing using ffdshow for decoding for example mpeg2 tv broadcasts? It is possible with dscaler5 when set to auto deinterlacing and nv12 output color space and also in coreavc (for h264 of course) when set to directshow deinterlacing.

Has it something to overlay mixer hw deinterlacing option in outut tab?

Regards
Piotr

[SealTooGreat]

@haruhiko_yamagata
"Use overlay mixer" also has to be checked in order to get correct AR and I always keep "Connect to compatible filters only" unchecked 'cause Haali Renderer is not on the list. And you are right it doesn't work with "Output->Allow output format change during playback" intermediate state.
It seems that rev 1075 doesn't exist at ffdshow download page. Have you meant rev 1074?!
With the setting above, beta1 works, beta2a does not work.
I wonder why?!:confused:
...I've rolled back to rev1006_clsid_sse_icl9 - works fine for me considering this problem I came on.(beta1 works fine too)

[Leak]

It seems that rev 1075 doesn't exist at ffdshow download page. Have you meant rev 1074?!
He definitely meant rev. 1075 (http://svn.sourceforge.net/viewvc/ffdshow-tryout/?view=log) - it's just that nobody has compiled it yet...

[kurt]

...I've rolled back to rev1006_clsid_sse_icl9 - works fine for me considering this problem I came on.(beta1 works fine too)
here beta2 doesn't work only for avi files (par.testfile) (http://home.arcor.de/evil.bert/test/xvid.par.avi)
other containers like mkv/mp4 don't show this behavior...

info:
latest mpc build by drevil_xxl
latest haali splitter

Edit: nevermind, rev 1075 should fix it

[marcellus]

I try to use ffdshow for realtime encoding to MPEG2 in AVI (constant bitrate, lavcodec).
I found that I must set "decoder bitsteam buffer size" to big value (about 10000000) and "maximum bitrate" to non-zero at "ratecontrol" tab. In other cases encoder freeze my system. may be it is possible somehow protect user from wrong settings. For example, use mean bitrate as a max bitrate, and auto-enlarge buffers to safe value (how to calculate it?).
In ffdshow encoder dialog use Presets->Load All->Libavcodec DVD PAL (or anything you find apropriate). This should give good defaults. I myself use that preset but I change bitrate from 6000 to 1500-2000 kbs (plus I use 2 b frames). With that setting I obtain acceptable tv captures.

[Fizick]

marcellus,
thanks. (My CPU is slow, so I try use 352x576. with max bitrate 4000. )

[Leak]

Download (http://leak.no-ip.org/AviSynth/ffdshow/)
Patch (https://sourceforge.net/tracker/index.php?func=detail&aid=1676882&group_id=173941&atid=867362) (against rev. 1076)

Another small update - I've added "applying pulldown" to the OSD info when pulldown is actually being applied.

I've also added debug printouts, which can be activated by starting the Avisynth script with a "#debug" comment. Those printouts can be logged with utilities like Sysinternal's DebugView.

np: Denzel + Huhn - Karlsruhe (Paraport)

[SealTooGreat]

@haruhiko_yamagata
rev 1075 works fine with MPC(+Haali Renderer), considering PAR/anamorphic issue.
Thanx :)

[marcellus]

@Fizick
I don't know what cpu you have, mine is an AthlonXP 2600 (I wouldn't call it fast these days :( ) and I capture full PAL->Kerneldeinterlace->Mpeg2 encode. (the audio part is captured as wav, I convert it to mp2 after the capture). I always use icl9 ffdshow compiles, it really makes a difference.

But keep in mind that a larger bitrate means more cpu needed (i don't know why) so that's why I keep it low at 1500 kpbs. 4000 kbps at 352x576 I think is overkill with libavcodec quality.

I forgot to mention that I "cooked" a custom quantization matrix that works very well with libavcodec mpeg2 encoder (I tried it with Xvid too but although I got good results I sticked with H.236 for Xvid encodes):

8,14,18,21,23,24,26,27
14,21,24,27,29,30,32,33
18,24,28,30,33,34,36,37
21,27,30,33,35,37,38,39
23,29,33,35,37,39,40,41
24,30,34,37,39,40,42,43
26,32,36,38,40,42,43,44
27,33,37,39,41,43,44,46

16,20,23,24,26,27,28,29
20,24,27,29,30,31,32,33
23,27,29,31,32,33,34,35
24,29,31,33,34,35,36,37
26,30,32,34,35,37,37,38
27,31,33,35,37,38,39,39
28,32,34,36,37,39,40,40
29,33,35,37,38,39,40,41

[Kado]

Hello

I was wondering if someone could add the wma decoder to the list of selectable decoders for WMA 7/8/9 like it it was done before but for WMV formats, so instead of using libavcodec I could use wma decoder within ffdshow.

Also I found the following issue:
If the plug in "dsp_stacker.dll" for Winamp is in the plugins folder after opening a file that uses ffdshow audio decoder if you try to open the audio decoder panel, the player (MPC in my case) hangs and you need to use the task manager to close it. This only happens if the winamp dsp option is enabled in ffdshow (i'm using enhancer for audio processing).

Steps to replicate this event:
Enable dsp processing in ffdshow with a plugin like Enhancer (http://www.winamp.com/plugins/details.php?id=81361) , install Multiple-DSP Stacker (http://www.winamp.com/plugins/details.php?id=128742) , open a file that uses ffdshow audio decoder and try to open the decoder panel.

If you load the file then install the dsp stacker the panel will open normally and the dsp stacker will be listed in the dsp plugins list.

Kado

[foxyshadis]

The wmv decoders do list wmv, as long as you have them installed correctly. I think it requires at least WMP9, I can't say for sure though, since I have 11 now.

Will have to test the DSP issue, although some DSP plugins just plain aren't coded very nicely, and practically none support greater than 16-bit stereo.

[Leak]

The wmv decoders do list wmv, as long as you have them installed correctly. I think it requires at least WMP9, I can't say for sure though, since I have 11 now.
Right, but Kado wanted the same for WMA in the audio decoder - only libavcodec can be chosen for WMA 7-9...

[Kado]

Will have to test the DSP issue, although some DSP plugins just plain aren't coded very nicely, and practically none support greater than 16-bit stereo.

In my case, Enhancer only suports audio up to stereo and 16bits max, so if you put the mixer (i'm using quadro settings) in front of the winamp dsp (in ffdshow) there won't be any sound processing from Enhancer, the same happens if the original file is 24bits or greater since ffdshow only converts the sample format in the end of all processing. Anyway the problem comes from the dsp stacker plugin, and since I only use it when shoutcasting (very rare) I remove the dll extension from the file and everything goes great.

Regarding WMA, I want exactly what Leak said, and I have WMP11 installed aswell.

Also since I have an Audigy 2 ZS is it worth to force 24bit output?

Kado

[iron2000]

The wmv decoder doesn't show for me.
Now on WMP10 with WM11 runtime.
WMP11 seems even more bloated and occupy more disk space so I choose 10 instead.

Are the Real and QT decoders suppose to show too?
Mine is only libavcodec.

[clsid]

Are the Real and QT decoders suppose to show too?
Mine is only libavcodec.
Only libavcodec is available for those. Also note that RealMedia is only partially supported, so it is pretty much useless.

[Inventive Software]

What the hell happened to ffdshow while I was away from Doom9 eh? :p :D ;)

[oddball]

ffdshow has problems with high bitrate DTS (1536kbps). AC3Filter plays them back flawlessly but ffdshow output stutters. I use SPDIF out in both ffdshow and AC3Filter.

[David1105]

Which ineterface in source of ffdshow, which grants access to change settings for "Audio settings" (Specifically, check/uncheck "Volume", "Normalize" and "Max. amplification", etc.), in the Audio Decoder Properties page?

IaudioFilterVolume seems like not exposed to outside. Is IffProcAudio the right one?

Thanks!

[Kado]

Which ineterface in ffdshow which grants access to change settings for "Audio settings" (Specifically, check/uncheck "Volume", "Normalize" and "Max. amplification", etc.), in the Audio Decoder Properties page?


Do you mean this (http://img.villagephotos.com/p/2004-6/737150/ffdshow_audio_panel.jpg)?

Kado

[Peuj]

Hi,

Just to know, have the Didee's YlevelsG, YlevelsS, YlevelsC been fixed finally?

http://forum.doom9.org/showthread.php?p=902088#post902088

Thanks

[Revgen]

I have problems capturing video using Dscaler 4.1.15 ( www.dscaler.org ) and FFDShow tryouts version 1058 March 22nd.

This happens on all previous versions of FFDshow too. Whenvever I try to cap using the VFW interface with ffdshow, my caps look good for half-a-sec then they turn into garbage.

Here's a clip for you guys to look at.

http://www.mediafire.com/?cog4grzzqzz

It's a huffyuv cap. It plays in VLC player. I can't open it in Vdub though.

Is it possible that Dscaler and FFDshow are using shared resources and possibly causing a conflict?

[David1105]

Do you mean this (http://img.villagephotos.com/p/2004-6/737150/ffdshow_audio_panel.jpg)?

Kado

No, I mean to change those settings programmatically, rather than manually.

[_xxl]

ffdshow has problems with high bitrate DTS (1536kbps). AC3Filter plays them back flawlessly but ffdshow output stutters. I use SPDIF out in both ffdshow and AC3Filter.
Can you do a regression test to see if this is a ffdshow/tryout bug?

[multiblitz]

DeepMeepBeep said:



I can confirm this - everything gets reported as 29.97 by FFDShow when it is attached to a VC1 decoder in the chain. Verified by Reclock and VMR9.

Can someone look into this bug ? PLEEEEEEEEAAAAAASE

[Leak]

I can confirm this - everything gets reported as 29.97 by FFDShow when it is attached to a VC1 decoder in the chain. Verified by Reclock and VMR9.
Well, if it works anything like DVD does then that's correct.

23.976FPS material on DVDs actually states that it's 29.970FPS in the video info header, but then the decoder only returns 23.976 frames per second (with timestamps that alternate between 33 and 50ms duration) and sets flags on the video to mark it as interlaced - that's what's called soft pulldown.

It's then up to the Video Renderer to make sense of everything and recognize that mess as 23.976FPS (or just treat it as 29.970FPS interlaced, like your good old analog TV set would); of course it could be that ffdshow doesn't pass all flags along correctly, but the video stream probably is marked as 29.970FPS in the video info header.

[haruhiko_yamagata]

Hi,

Just to know, have the Didee's YlevelsG, YlevelsS, YlevelsC been fixed finally?

http://forum.doom9.org/showthread.php?p=902088#post902088

Thanks
Yes, it has been fixed.

[Randi]

i can confirm the high bitrate (1.5mbit) DTS bug, also that it used to work in older versions, I don't have time to test which version, though

[DeepBeepMeep]

it could be that ffdshow doesn't pass all flags along correctly, but the video stream probably is marked as 29.970FPS in the video info header.

Yep that's my theory (see my previous post). However the issue isn't only about fps reported incorrectly, the video is really accelerated as if it had really 29.970 fps and consequently audio and video gets out of sync

[fastplayer]

i can confirm the high bitrate (1.5mbit) DTS bug, also that it used to work in older versions, I don't have time to test which version, though
Does it happen too when downmixed to stereo?

[_xxl]

also that it used to work in older versions, I don't have time to test which version, though
I don't have time to fix your bug.

[Peuj]

Yes, it has been fixed.

Thanks :)

[Peuj]

I have some questions about the "Levels":

I want to do this YlevelsS(0,1.6,255,0,255).

So I select YlevelsS but after I'm lost :p

Should I set input "0,255" || ouput "0,255" and Gamma 1.6 ?

because I see in this thread http://forum.doom9.org/showthread.php?t=113991&highlight=ColorYUV that for Levels(16,1,240,0,255) I have to do: input "16,240" || ouput "1,255" and Gamma 1

I think I miss something with the Gamma value :confused:

Some other questions about the Levels:
Do I need to check "Show histogram" and "Full range"?
What is the Automatic check box?

Thanks for your help.

[foxyshadis]

Yes, that's exactly how to translate from avisynth syntax to a levels gui. (Some gui put the gamma slide in with the input, which is where avisynth got its version.)

Show histogram just overlays a grey histogram on the levels window, so you can watch your changes in realtime. (More difficult with a stupid modal window like MPC's, my #1 complaint about the player.) Full range depends on what you're doing with the levels, if input and output are both in clipped "tv" range you don't want to use it, but otherwise you do.

Automatic should be avoided unless you want jumpy levels. Try it and see what happens.

[Rectal Prolapse]

Well, if it works anything like DVD does then that's correct.

Unfortunately this results in massive A/V sync problems, and sometimes the video plays too fast as well!

Without FFDShow, the renderer reports 23.976 fps - which is the correct framerate. This is true for ALL VC1 decoders I've tried - they ALL feed 23.976 to the renderer.

So, I'm not sure what is going on here. With an AVC decoder attached to FFDShow, it works as expected - 23.976 fps reported by FFDShow and the renderer.

[skystrife]

Oh, now I see that the OSD uses MMX. Didn't think of that. I'm beginning to implement the MMX stubs (using SSE2 integer ops), as that seems to be the easiest way to move forward (as it means the code using MMX intrinsics can remain unchanged). The code is now better than it was, but there are still visual problems with the OSD. It's readable, kinda, but something's getting screwed up with the outlines. I might be doing some arithmetic backwards or something. Though I haven't compared to see how it looks in a 32-bit build, so maybe not.

I've only put in the functions needed for the OSD so far, because it's easy to tell when they're working and when they're not. I don't know what the other things (kernelDeint, TomsMoComp) should do, so I don't know how to tell if they're working or not.

Edit: aha, the problem was I was using a signed function for something that should have been unsigned. With that fixed I now have a properly working OSD.

Gonna pester you about progress again. =P

[Peuj]

Yes, that's exactly how to translate from avisynth syntax to a levels gui. (Some gui put the gamma slide in with the input, which is where avisynth got its version.)

Show histogram just overlays a grey histogram on the levels window, so you can watch your changes in realtime. (More difficult with a stupid modal window like MPC's, my #1 complaint about the player.) Full range depends on what you're doing with the levels, if input and output are both in clipped "tv" range you don't want to use it, but otherwise you do.

Automatic should be avoided unless you want jumpy levels. Try it and see what happens.

OK thanks !!

Just to be sure, when you say "in clipped "tv" range" you mean "16,240" right ?
So as I use haali renderer all my video will be in 0,255 and I should always use Full range.

[Dr Pizza]

Gonna pester you about progress again. =P

OK, I can put the build up somewhere if people can look at it (it's pure VC++ 8 SP1), but I need somewhere to put the source (GPL, you know, I can't redistribute my build without redistributing source, and I have no practical place to stick the source at the moment).

It's worked well for all the things I've thrown at it (various MPEG-4 variants, 1080p H.264), but I don't have an exhaustive set of test files or anything like that, and don't use much in the way of postprocessing/filtering. There are things I know should be problematic, as I still haven't implemented all the MMX routines and had to #ifdef out a couple of bits of build-breaking inline asm.

[Leak]

OK, I can put the build up somewhere if people can look at it (it's pure VC++ 8 SP1), but I need somewhere to put the source (GPL, you know, I can't redistribute my build without redistributing source, and I have no practical place to stick the source at the moment).
You might want to read the GPL again :p - you can redistribute it sans source as much as you like, as long as you provide the source (even if it's just via email) when someone asks for it. Of course, just putting the source next to the binary is easier.

Also, why don't you just add a patch to the patches tracker on SourceForge?

np: Bleubird - Writer (Rip USA (The Birdfleu))

[Dr Pizza]

You might want to read the GPL again :p - you can redistribute it sans source as much as you like, as long as you provide the source (even if it's just via email)
That leaves me on the hook for 3 years, and that's not something I'm willing to commit to. Of the three source provision options that the GPL has (source alongside, 3 years on-demand access, or a reference to another person's source distribution of the kind mentioned previously), only one is practical; the first. Distribution with source alongside (in the same package).

when someone asks for it. Of course, just putting the source next to the binary is easier.
Right.

Also, why don't you just add a patch to the patches tracker on SourceForge?
I have no idea how I would go about doing that, and I don't see how I can make a 3 year guarantee about SourceForge.

[clsid]

Don't worry too much. It's not like someone is going to sue you.

[cc979]

problem compiling faad when using the new mingw headers - w32api-3.9 and runtime.3-12

gcc -c -I. -I../.. -DSTDC_HEADERS -mno-cygwin -mdll -mthreads -pipe -mmmx -DNDEBUG -UDEBUG -DWIN32 -D_WIN32 -O2 -march=pentium-mmx -mtune=i686 -fomit-frame-pointer -finline-functions -finline -frename-registers -fweb -funit-at-a-time -Os -MMD -o bits.o bits.c
In file included from bits.c:33:
common.h:220: error: conflicting types for 'uint32_t'
d:/mingw/bin/../lib/gcc/i686-pc-mingw32/4.0.4/../../../../include/stdint.h:32: error: previous declaration of 'uint32_t' was here
common.h:223: error: conflicting types for 'int32_t'
d:/mingw/bin/../lib/gcc/i686-pc-mingw32/4.0.4/../../../../include/stdint.h:31: error: previous declaration of 'int32_t' was here
common.h:225: error: conflicting types for 'int8_t'
d:/mingw/bin/../lib/gcc/i686-pc-mingw32/4.0.4/../../../../include/stdint.h:27: error: previous declaration of 'int8_t' was here
make: *** [bits.o] Error 1

this there a quick fix or should go back to previous headers

cheers

[Randi]

Does it happen too when downmixed to stereo?

i just pass-through to the receiver, but will check tonite (about DTS bug)

[_xxl]

problem compiling faad when using the new mingw headers - w32api-3.9 and runtime.3-12

Did you delete *.d files after "make clean"?

[cc979]

Did you delete *.d files after "make clean"?

oops, i meant faad in ffdshow-rev1082

ps. it can be compiled if i modify ffdshow-tryout\src\codecs\faad\common.h

line 220:
//typedef unsigned long uint32_t;
//typedef unsigned short uint16_t;
//typedef unsigned char uint8_t;
//typedef long int32_t;
//typedef short int16_t;
//typedef char int8_t;

maybe the new runtime broke the defines

cheers

[foxyshadis]

I have no idea how I would go about doing that, and I don't see how I can make a 3 year guarantee about SourceForge.

You have tortoisesvn? Right click on ffdshow/trunk, in the tortoise menu select Create Patch. By default it'll check all the in-svn files that are changed (usually mostly projects, those should be unchecked), if you added any files you'll want to look for them in the unchecked masses.

Adding a patch to a repository like Sourceforge would probably constitute enough of a good-faith effort if it ever disappears and someone was really intent on suing anyone who didn't supply a source, and for added insurance you could attach it to this thread. :p If you supply the patch and have someone else do the building, you have no obligation whatsoever, of course.

[skystrife]

Oh, one thing you might want to check closely is H.264 playback on the 64-bit build. That's the main reason I'm seeking out a separate 64-bit ffdshow- the celtic_druid build has artifacts galore whenever I play a file with H.264 video content.

If the H.264 playback is artifact-free, I'll be one freaking happy camper. =)

...and I'm assuming this won't have an installer, so I'll have to research how to install it manually. It can't be as simple as running regsvr32 on a few .ax files, can it?

[haruhiko_yamagata]

@Dr Pizza
Great!
I've orderd my new PC, including Xp64, I'm willingly to test it (if I could successfully build it :p ).

@Kado
I'm debuging Winamp plugin feature. ffdshow's code around there is not written very nicely, as is unusual with Milan. I have fixed some (not commited yet), but Multiple-DSP Stacker still seems to be incompatible with ffdshow. I'm thinking of excluding it. Do you have any other such plugins?

[cc979]

regarding my post http://forum.doom9.org/showpost.php?p=977216&postcount=953

just had a reply back from the mingw developers

http://sourceforge.net/tracker/index.php?func=detail&aid=1691376&group_id=2435&atid=102435

[Dr Pizza]

Oh, one thing you might want to check closely is H.264 playback on the 64-bit build. That's the main reason I'm seeking out a separate 64-bit ffdshow- the celtic_druid build has artifacts galore whenever I play a file with H.264 video content.
Can you provide a sample problematic file? I don't have much H.264 content, but the things I've tried (some TV caps and some 1080p samples) have all been fine.

...and I'm assuming this won't have an installer, so I'll have to research how to install it manually. It can't be as simple as running regsvr32 on a few .ax files, can it?
I'll include an installer, though I won't be able to promise it works. I don't have a machine without VC++ installed so I won't be able to test it puts the libraries in the right place (on my system I can just rely on the SxS libraries and don't have to worry about the private assembly nonsense).

[skystrife]

Can you provide a sample problematic file? I don't have much H.264 content, but the things I've tried (some TV caps and some 1080p samples) have all been fine.


I'll include an installer, though I won't be able to promise it works. I don't have a machine without VC++ installed so I won't be able to test it puts the libraries in the right place (on my system I can just rely on the SxS libraries and don't have to worry about the private assembly nonsense).

I have larger files, but I don't have a tool to cut down the mkv into a sample. =P


http://skystrife.com/uploads/avp.zip

@Dr Pizza
Great!
I've orderd my new PC, including Xp64, I'm willingly to test it (if I could successfully build it :p ).

@Kado
I'm debuging Winamp plugin feature. ffdshow's code around there is not written very nicely, as is unusual with Milan. I have fixed some (not commited yet), but Multiple-DSP Stacker still seems to be incompatible with ffdshow. I'm thinking of excluding it. Do you have any other such plugins?

XP64 doesn't have a 64-bit media player, IIRC.

[TheShadowRunner]

Hey All, I got a small question regarding CLSID's builds.
Those come in 2 flavors, "standard" and "ICL9".
Why would i chose a ICL9 build, does that depend on the processor i own?
(I got 2 laptops, 1 with Pentium 3-M 1.2Ghz, and 1 centrino pentium-m @ 1.7ghz, first one supports SSE, second one supports both SSE & SSE2). So would the ICL9 have any advantage for those cpus?
Thanks for the info.
Regards,

TSR

[clsid]

It does not have any advantage unless you use some specific filters in ffdshow. See the FAQ link in the opening post of the topic.

[clsid]

XP64 doesn't have a 64-bit media player, IIRC.
Both XP64 and Vista64 use a 32-bit version of WMP by default.

On Vista, switching between 32/64 bit WMP can be done with these commands:

%windir%\system32\unregmp2.exe /SwapTo:32
%windir%\system32\unregmp2.exe /SwapTo:64

Maybe that also works on XP64.

[skystrife]

If XP64 has a 64-bit WMP, it's probably located in the Program Files folder as opposed to the Program Files (x86).

I don't have XP64, so I couldn't tell you. =P

[Px]

If XP64 has a 64-bit WMP
It doesn't have :)

[clsid]

WMP11 (64-bit) for Windows XP64:
http://www.microsoft.com/downloads/details.aspx?familyid=C6A4CECB-9CB2-42F7-8814-7CB5067EE958&displaylang=en

[Px]

WMP11 (64-bit) for Windows XP64:
http://www.microsoft.com/downloads/details.aspx?familyid=C6A4CECB-9CB2-42F7-8814-7CB5067EE958&displaylang=en
Yeah, I knew about it, and it already downloaded and installed, but after install it lay in Program Files (x86). Later, I take some time to look on it more closely....

[skystrife]

If you don't have a x64 build of ffdshow installed, throw an XviD or DivX file at it. It it works, you know that the player is 32-bit (using FFDShow x32), and if it doesn't you know that the player is 64-bit and doesn't have a codec for the file. =)

[zambelli]

WMP11 (64-bit) for Windows XP64:
http://www.microsoft.com/downloads/details.aspx?familyid=C6A4CECB-9CB2-42F7-8814-7CB5067EE958&displaylang=en
Unfortunately, that player is 32-bit. The only 64-bit WMP is available on Vista x64, and even there it's not the default player.

[Dr Pizza]

Don't worry too much. It's not like someone is going to sue you.
But I should worry about it, because those are the terms of the GPL. And if you're not going to follow the terms of the license, it seems to make the whole thing a bit pointless.

[Dr Pizza]

I have larger files, but I don't have a tool to cut down the mkv into a sample.
OK, well that looks fine to me (using my own media player app), there didn't seem to be any artefacting.

[skystrife]

But I should worry about it, because those are the terms of the GPL. And if you're not going to follow the terms of the license, it seems to make the whole thing a bit pointless.

I'd give you a ftp for my website, but if the download was hit a lot it might knock the whole thing offline, seeing as I don't have a lot of bandwidth.

Hmm...

[skystrife]

Hate to double-post, but this would probably not be seen otherwise.

http://code.google.com/hosting/

Maybe there? (Just throwing out ideas)

[Dr Pizza]

OK, so, anyway, here we go.

Patch:
https://sourceforge.net/tracker/?func=detail&atid=867362&aid=1692613&group_id=173941

Installer (which is substantially untested):
http://www.file.am/d/file/ffdshow_rev1081_20070328_x64.exe/

x64 VC8SP1 redist:
http://www.file.am/d/file/vcredist_x64.exe/

[skystrife]

OK, so, anyway, here we go.

Patch:
https://sourceforge.net/tracker/?func=detail&atid=867362&aid=1692613&group_id=173941

Installer (which is substantially untested):
http://www.file.am/d/file/ffdshow_rev1081_20070328_x64.exe/

x64 VC8SP1 redist:
http://www.file.am/d/file/vcredist_x64.exe/

Installer worked perfectly, no problems with it to report. =)

H.264 decoding is no longer crap, so that makes for a very happy Strifey! =P

Vorbis decoding in libavcodec was really wonky, but switching to tremor fixed that in a jiffy.

[siim04]

Unfortunately, that player is 32-bit. The only 64-bit WMP is available on Vista x64, and even there it's not the default player.

You can switch between 64-bit and 32-bit WMP in Windows XP (I think it worked similar to Vista version). I tried using 64-bit WMP, but had to change back to 32-bit version as most codecs were not available. Also Windows Media Encoder is only 64-bit on Windows XP x64, which means that I can't use it for converting files either (and I really would like to compress my mpeg1, mpeg2 and xvid videos as they are taking up too much space).

I tried to compile ffdshow tryouts with Visual Studio C++ 2005 Express Edition and yasm, but in vain (I think something actually compiled but did not work). I think the project was not set up correctly in my case (I do not exactly know, which files should be replaced to compile 64-bit binaries).

[Peuj]

Hi,

I don't know if it's ffdshow issue but I have a problem with this mov file: http://www.jeuxvideo.fr/telecharger-video-1-gameplay-43971.html

I use the build ffdshow_rev1082_20070328_clsid.exe and the latest MPC version compiled by drevil_xxl.

When I play the video everything is perfect excepted that I don't have sound at all and no error message from MPC.
It's like there is no audio stream.

Does somebody know what can be the problem?

Thanks

[_xxl]

64-bit compile:
.\src\TffDecoder.cpp(1511) : error C3861: 'SetThreadIdealProcessor': identifier not found
.\src\TffDecoder.cpp(1515) : error C3861: 'SetThreadIdealProcessor': identifier not found
.\src\TffDecoder.cpp(1520) : error C3861: 'SetThreadIdealProcessor': identifier not found
C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(42) : warning C4005: 'INT_MIN' : macro redefinition
../..\inttypes.h(70) : see previous definition of 'INT_MIN'
C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(43) : warning C4005: 'INT_MAX' : macro redefinition
../..\inttypes.h(74) : see previous definition of 'INT_MAX'
C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(67) : warning C4005: '_UI32_MAX' : macro redefinition
../..\inttypes.h(58) : see previous definition of '_UI32_MAX'
C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(72) : warning C4005: '_I64_MIN' : macro redefinition
../..\inttypes.h(46) : see previous definition of '_I64_MIN'
C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(74) : warning C4005: '_I64_MAX' : macro redefinition
../..\inttypes.h(50) : see previous definition of '_I64_MAX'
C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(76) : warning C4005: '_UI64_MAX' : macro redefinition
../..\inttypes.h(54) : see previous definition of '_UI64_MAX'
Creating library Releasex64/ff_x264.lib and object Releasex64/ff_x264.exp
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_p_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_ddl_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_dc_top_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_dc_left_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_v_mmx
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_v_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_4x4_vl_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_vr_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol x264_intra_sa8d_x3_8x8_core_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_dc_top_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_vl_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_ddl_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8c_dc_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8c_v_mmx
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_ddr_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_dc_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_4x4_ddl_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8c_p_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_dc_mmxext
libavcodec should be compiled by GCC.No asm code is used in MSVC.

[TheRyuu]

I getting a problem with the avisynth portion of ffdshow.

For some reason it's not releasing the RAM if I should change or deactivate the script.

So if I start a script with the setmemorymax(512) parameter then the 512mb of ram that avisynth is now using will not get released for as long as the video is showing.

Everything gets released after ffdshow closes (movie ends, I close out MPC, etc...) so it's not like the memory is permanently taken. Never had this problem before and I haven't changed anything in avisynth so I figure it has to be ffdshow.

I'm using this build:
ffdshow_rev1082_20070331_sse_xxl.exe

The build I had on before the one above was this one (which I had no problems with):
ffdshow_rev1074_20070325_clsid_sse_icl9.exe

And everything before that one I've never had this problem before.

So I guess it could've come up anywere between those two.
I haven't tried clsid's build of the same rev yet.

[MidnightWatcher]

OK, so, anyway, here we go.

Patch:
https://sourceforge.net/tracker/?func=detail&atid=867362&aid=1692613&group_id=173941

Installer (which is substantially untested):
http://www.file.am/d/file/ffdshow_rev1081_20070328_x64.exe/

x64 VC8SP1 redist:
http://www.file.am/d/file/vcredist_x64.exe/
Dr Pizza, you rock! Keep up this excellent progress!

:thanks:

[clsid]

Please test the VFW interface of ffdshow 64-bit. I am curious if my workaround is working.

[Dr Pizza]

64-bit compile:
.\src\TffDecoder.cpp(1511) : error C3861: 'SetThreadIdealProcessor': identifier not found

I'm using the Vista SDK's and that's working for me; that error suggests that _WIN32_WINNT isn't being set correctly for you (that function prototype has a guard around it requiring NT4 and up).

C:\Program Files\Microsoft Visual Studio 8\VC\include\limits.h(42) : warning C4005: 'INT_MIN' : macro redefinition
Harmless, albeit irritating. Doesn't seem to be 64-bit specific?

Creating library Releasex64/ff_x264.lib and object Releasex64/ff_x264.exp
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_p_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_ddl_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_dc_top_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_dc_left_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_v_mmx
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_v_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_4x4_vl_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_vr_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol x264_intra_sa8d_x3_8x8_core_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_dc_top_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_vl_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_ddl_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8c_dc_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8c_v_mmx
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_ddr_sse2
predict-c.obj : error LNK2001: unresolved external symbol predict_16x16_dc_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_4x4_ddl_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8c_p_core_mmxext
predict-c.obj : error LNK2001: unresolved external symbol predict_8x8_dc_mmxext
This means that the asm has failed to build.

[Dr Pizza]

Please test the VFW interface of ffdshow 64-bit. I am curious if my workaround is working.

How would one go about doing that?

What is the workaround?

[clsid]

The workaround of putting it in the appdir.

You should be able to test if it works with 64-bit VirtualDub:
http://virtualdub.sourceforge.net/

[zambelli]

You can switch between 64-bit and 32-bit WMP in Windows XP (I think it worked similar to Vista version). I tried using 64-bit WMP, but had to change back to 32-bit version as most codecs were not available.
My point was that WMP11 is only available in 32-bit version for XP. The version marked 64-bit is meant to be installed on XP x64, but is actually just the same 32-bit version for most part. I think the only 64-bit parts of that package are some portable device drivers and the WMP shell object. The only true 64-bit WMP11 is in Vista x64.

Also Windows Media Encoder is only 64-bit on Windows XP x64, which means that I can't use it for converting files either (and I really would like to compress my mpeg1, mpeg2 and xvid videos as they are taking up too much space).
Actually, WME9 32-bit installs on XP x64 just fine.

Anyway, this is all off-topic to this thread.

[Dr Pizza]

The ExpandConstant thing you suggested doesn't seem to work; it's not expanding it (so my registry is getting stuff that says literally ExpandConstant in it). However, even with that changed (so it puts the right strings into the registry) vdub is claiming to be unable to find a suitable VFW codec. It doesn't look like it's even trying to load ff_vfw.dll, in fact; using depends to profile vdub shows no attemp to load the .dll.

[Jeremy Duncan]

Can somebody add Leak's Patch to Drevil's and XXL's FFDshow builds.

[clsid]

Can somebody add Leak's Patch to Drevil's and XXL's FFDshow builds.
Split personality :confused:

[clsid]

The ExpandConstant thing you suggested doesn't seem to work; it's not expanding it (so my registry is getting stuff that says literally ExpandConstant in it). However, even with that changed (so it puts the right strings into the registry) vdub is claiming to be unable to find a suitable VFW codec. It doesn't look like it's even trying to load ff_vfw.dll, in fact; using depends to profile vdub shows no attemp to load the .dll.
Spaces in the path cause problems. Hence the usage of GetShortName. Try if this works:

[Registry]
Root: HKLM; Subkey: SOFTWARE\Microsoft\Windows NT\CurrentVersion\Drivers32; ValueType: string; ValueName: VIDC.FFDS; ValueData: {code:GetVFWLocation|}; Flags: uninsdeletevalue; Components: ffdshow\vfw

[C0de]
function GetVFWLocation(dummy: String): String;
begin
Result := GetShortName(ExpandConstant('{app}\ff_vfw.dll'));
end;

[haruhiko_yamagata]

Hi,

I don't know if it's ffdshow issue but I have a problem with this mov file: http://www.jeuxvideo.fr/telecharger-video-1-gameplay-43971.html

I use the build ffdshow_rev1082_20070328_clsid.exe and the latest MPC version compiled by drevil_xxl.

When I play the video everything is perfect excepted that I don't have sound at all and no error message from MPC.
It's like there is no audio stream.

Does somebody know what can be the problem?

Thanks
8KB/s for 360MB. I cannot download.
Please use faster file sharing service.

[siim04]

Actually, WME9 32-bit installs on XP x64 just fine.

Anyway, this is all off-topic to this thread.

On my PC WME9 32-bit says that the machine type is not supported and won't install.

On the bright side. I tested the x64 version of ffdshow tryouts and it worked well (at least with WME x64, so it should work with graphedit as well).
:thanks: Thank you for the x64 version!

[Peuj]

8KB/s for 360MB. I cannot download.
Please use faster file sharing service.

Hi, thanks for looking at my problem.

I don't have problem to download the video maybe there was too much download when you have tried?

Anyway I have upload the file on http://www.megaupload.com/fr/?d=0O31ORW2
If you know a faster file sharing service let know I will upload the file on it.

Here's the information about the file:

Format : QuickTime
Format/Info : Original Apple specifications
Format/Family : MPEG-4
File size : 360 MiB
PlayTime : 1mn 45s
Bit rate : 29 Mbps
StreamSize : 4.57 MiB
Encoded date : UTC 2007-03-27 01:17:32
Tagged date : UTC 2007-03-26 16:39:21

Video #0
Codec : Sorenson 3
Codec/Info : Sorenson Media Video 3 (Apple QuickTime 5)
PlayTime : 1mn 45s
Bit rate : 28 Mbps
Width : 720 pixels
Height : 576 pixels
Aspect ratio : 5/4
Frame rate : 30.000 fps
Standard : PAL
StreamSize : 351 MiB
Encoded date : UTC 2007-03-26 16:17:32
Tagged date : UTC 2007-03-26 16:39:21

Audio #0
Codec : PCM
PlayTime : 1mn 47s
Bit rate : 353 Kbps
Channel(s) : 2 channels
Sampling rate : 44 KHz
Resolution : 16 bits
StreamSize : 4.50 MiB
Encoded date : UTC 2007-03-26 16:17:32
Tagged date : UTC 2007-03-26 16:39:21

Thanks

[clsid]

It is probably a splitter issue.

[thuan]

Same problem with the file as OP. mplayer can play the file fine. Here's the log

Playing infernal_video_1_-_gameplay_anglais_34539.mov.
Quicktime/MOV file format detected.
VIDEO: [SVQ3] 720x576 24bpp 30.000 fps 0.0 kbps ( 0.0 kbyte/s)
[ass] auto-open
[ass] Init
Fontconfig error: Cannot load default config file
[ass] Updating font cache.
==========================================================================
Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
Selected video codec: [ffsvq3] vfm: ffmpeg (FFmpeg Sorenson Video v3 (SVQ3))
==========================================================================
==========================================================================
Opening audio decoder: [alaw] aLaw/uLaw audio decoder
AUDIO: 44100 Hz, 2 ch, s16le, 705.6 kbit/50.00% (ratio: 88200->176400)
Selected audio codec: [ulaw] afm: alaw (uLaw)
==========================================================================
AO: [dsound] 44100Hz 2ch s16le (2 bytes per sample)
Starting playback...
VDec: vo config request - 720 x 576 (preferred colorspace: Planar YV12)
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [directx] 720x576 => 480x384 Planar YV12
A: 6.2 V: 6.2 A-V: 0.001 ct: 0.006 187/187 10% 19% 0.1% 0 0

[clsid]

The file plays correctly with the Nero Digital MP4/MOV parser (NDParser.ax). So this is NOT a ffdshow problem.

[Peuj]

The file plays correctly with the Nero Digital MP4/MOV parser (NDParser.ax). So this is NOT a ffdshow problem.

So it's a MPC problem ?

[clsid]

Yes, more specifically with the internal MP4/MOV splitter. Haali's splitter doesn't work at all with this file.

[clsid]

A lot of people are having trouble finding the 'presets' functionality in ffdshow because it is not in a very obvious place.

Some suggestions to improve the GUI:

* Rename 'Image settings' to 'Profiles/Presets'
* Rename 'Show/hide filters' to 'Filters' and remove the indentation.
* Remove indentations for 'Decoder options', 'Queue & misc' and 'Output'.

And for ffdshow audio decoder:

* Rename 'Audio settings' to 'Profiles/Presets'.
* Rename 'Show/hide filters' to 'Filters' and remove indentation.
* Remove indentation for 'Output'.

[chros]

@Kado
I'm debuging Winamp plugin feature. ffdshow's code around there is not written very nicely, as is unusual with Milan. I have fixed some (not commited yet), but Multiple-DSP Stacker still seems to be incompatible with ffdshow. I'm thinking of excluding it. Do you have any other such plugins?
Thanks for this act! I have asked about a month ago about the new DFX plugin (ver. 8.x) which is able to connect to multichannel audio (it works in winamp), but not working in ffdshow !
Can you do something about it ? (dts, aac, ac3, etc.)

Thank you !

[vlada]

H, I'd like to report some problems with Aspect Ratio of MPEG-4 ASP in AVI. I prepared a 320x240 file where I set DAR to 2.35:1 (using MPEG4 Modifier). I did some tests with following results:
XviD decoder - WMP OK, MPC OK
ffdshow decoder - WMP 4/3, MPC 16/9

So I guess there is something wired in ffdshow regarding AR setting. In fact how does AR signalling work? Does ffdshow sends any special signals to a player? I think it would be very nice to have a feature which would resize the video to a correct AR (expand or shrink). This might then also solve the problem with DVD letterboxing I mentined before.

Another strange thing is, that ffdshow reports the AR to be 4/3 when playing the file.

[kurt]

@ vlada: which rev. do you use? recently there was a fix for this AR-bug (since rev 1075)...