ffdshow development #2

[bond]

ok, now that, after more than one year of development and testing, a new build of one of the best mpeg-4 (and more) directshow decoders existing, ffdshow, has been officially released, i take this opportunity to close the good old 65 sites strong ffdshow development (http://forum.doom9.org/showthread.php?s=&threadid=48511) thread, as its simply not possible anymore to find any infos in it

everyone is invited to continue the discussion about ffdshow development in this thread or start new threads for discssing specific issues in seperate threads :)

grap ffdshow here (http://sourceforge.net/projects/ffdshow/)

[Soulhunter]

@ Coroner

Thanks for the info !!!

Only asked coz there is this winamp plugin...
Version 1.2

SBLive! Resample BugFix output plug-in for Winamp2

Freeware component!

USE THIS SOFTWARE AT YOUR OWN RISK !!!

I will not be responsible for anything regarding the use of this software!

If you are not looking for highest audio quality or have no idea about what
distortion is, than this component is of no use for you ... it will only make
your CPU work more on music

but for those with SBLive! or Audigy series of soundcards and there are a lot more
with this issue! this will give you the perfect sound you wanted to hear when
you've bought that soundcard !!!

[The SBLive! issue]

SBLive! has a low quality resample to 48KHz algoritm (or multirate filter
kernel too small...) maybe because to keep low latency to audio stream ...

This is the SBLive! (and not only) problem:

absolutely any audio stream is converted to 48KHz.
the worst performance is 44.1KHz audio streams (cd audio).

Audigy has the same issue except it resamples to 96000Hz but still
kernel size too small so 1% IMD at 16KHz for Audigy2 (www.tomshardware.com)

[The solution:]

my high quality resample to 48KHz (or any >= 48000, that means 96000Hz too)
component for Winamp3.

Testing results:

using SBLive! to resample to 48000:
19KHz IMD=20%
18KHz IMD=10%
16KHz IMD=5%
14KHz IMD=1%

You don't believe me ...???? I've included in component pack generated tones
used in my test versions of 44100Hz and 48000Hz ... Should sound the same !!!

using my component for KernelLength=16000 samples:
19KHz IMD=0.01%
18KHz IMD=0.001% ; can't measure less than this
16KHz IMD=0%

CPU Load:
<14% on Duron 825MHz for KernelSize 16000
<5% on Duron 825MHz for KernelSize 6000
<1% on AthlonXP2000+ for KernelSize 20000

[known issues]
there are no known issues

<TargetSampleRate> specifies resample target frequency of filter (should be
48000 for SBLive! cards, 96000 for Audigy, and this must be the working
Samplerate of the DSP on your soundcard!!!

<KernelSize> is the multirate kernel filter length in samples of the lowpass
filter. Here are some hints for 48000&96000:
16000 samples for quality described above
6000 samples are enough for 0.1% IMD @ 19KHz and no audible IMD
(IMD is above 20khz and should be rejected by the hardware
lowpass filter of your soundcard)
altering this results:
higher KernelSize => higher quality more CPU load.
lower KernelSize => lower quality less CPU load.
don't alter this if you do not know what means this or if you do not check
the results on a spectrum analyzer!
I've used as spectrum analyzer Goldwave 4.23 and CoolEdit2000 to measure
IMD. To record result just loopback your soundcard output to Line-In and
view the result over SpectrumAnalyzer in realtime.

[some hints]
If you really want your SBLive/Audigy/Audigy2 to sound perfect I recommand you
to download and install KX Drivers wich you can find at: www.kxproject.com.
These drivers are opensource and also freeware.
Best sound : KX Drivers + SBLive!Bugfix plugin. Try it!

[Contact:]
email: adi111p@yahoo.com


P.S. Haven't you noticed that DVD movies (sampled to 48000Hz) sounded perfect
on your sound system and Audio CD's didn't????? This is why!

another advice: Use KX drivers for SBLive and Audigy and Audigy2 soundcards. You will get
professional EQ-s, filters, 5.1,6.1,7.1 Decoders, but you will need some time to understand
how to configure it

download it from www.kxproject.com

So, still need to know this about ffdshows upsampling...

- Is Kaiser "the best" upsampling algorithm of ffdshow ???

- Is it better than the Audigy's "bad" resampling ???

- When yes, how much better is it ???


Tia n' Bye

[unmei]

great it even comes with a built in microwave :eek:

I don't know it this is ultimately new, but it wasn't in the 2004-06-29 i used so far: keyboard shortcuts to toggle PP, noise etc ..good idea, very useful!

[Coroner]

@Soulhunter

Haven't seen that one before. Same trick as resample in foobar by the looks of it. Would avoid the resample at 48Khz, although it's still not going to sound as good as a non-resample card, depending on the resampler it would probably sound better than the creative one. Suggest you might get better info by posting a question here Head-Fi Computers as a source forum (http://www4.head-fi.org/forums/forumdisplay.php?f=59)

Hope that helps

[saratoga]

Any chance of an Athlon or P4 optimized build? Or is that not worth the effort?

[opsis81]

Please fix the broken MPEG-1 and MPEG-2 decoding...
Please make me get rid of any other MPEG-2 decoder I have registered in my PC...
Thank you in advance!:)

[SeeMoreDigital]

Originally posted by opsis81
Please fix the broken MPEG-1 and MPEG-2 decoding...
Please make me get rid of any other MPEG-2 decoder I have registered in my PC...
Thank you in advance!:) I second this request!

It would be great if it could support Mpeg2 HiDef pixel frame sizes right up to 1920x1080/88...


Cheers

[SeeMoreDigital]

Originally posted by bond
... Grap ffdshow here (http://sourceforge.net/projects/ffdshow/) It's probably worth pointing out that the FFdshow builds on the SourceForge web site are by Milan Cutka.


Cheers

[athos]

Originally posted by SeeMoreDigital
It's probably worth pointing out that the FFdshow builds on the SourceForge web site are by Milan Cutka.
Cheers
The ffdshow-alpha builds are actually by me.

Did you notice the new ffdshow homepage that Milan put up?
http://ffdshow.sourceforge.net/tikiwiki/

[bond]

Originally posted by athos
Did you notice the new ffdshow homepage that Milan put up?
http://ffdshow.sourceforge.net/tikiwiki/hm, cant reach it atm :(

[hellfred]

It sure took its time, but finally a side came up in my browser.
So do not give up.
Here is a deep link (http://ffdshow.sourceforge.net/tikiwiki/tiki-view_articles.php)
Hellfred

[SeeMoreDigital]

Originally posted by athos
The ffdshow-alpha builds are actually by me.

Did you notice the new ffdshow homepage that Milan put up?
http://ffdshow.sourceforge.net/tikiwiki/

I apologise Athos. But why is Milan's name slapped all over the change logs?

http://sourceforge.net/project/shownotes.php?group_id=53761&release_id=255667

If this has confused me, I can't imagine how the newbies are going to get on :confused:


Cheers

[athos]

Originally posted by SeeMoreDigital
I apologise Athos. But why is Milan's name slapped all over the change logs?

http://sourceforge.net/project/shownotes.php?group_id=53761&release_id=255667

If this has confused me, I can't imagine how the newbies are going to get on :confused:


All CVS commits are by Milan. I just compile the binaries.

[DAvenger]

Many thanks to Milan for 'fixing' the wildcard issue. Now RadLight users can use FFDShow once again :D

(Milan, to bolo teda rychle hehe :p )

[faxmactor]

Sorry for taking over this matter from the previous thread, but I haven't replied this reply ;), and besides the problem still exists.

Originally posted by faxmactor
I have video that is likely to be made with an buggy, early Xvid 1.0 (or earlier) build. It has lots of b-frames but no keyframes in the beginning.
Xvid is able to display it correctly as well as mplayer (which uses libavcodec, too), so I came to a conlusion that this issue could be fixed in ffdshow as well. Milan, please have a look at.
A screenshot: HERE (http://ozdtersegi.axelero.net/ffdshow_issue.jpg), a short clip (2,5M): HERE (http://faxmactor.port5.com/Excel%20Saga%20-%2001%20-%20The%20Plan%20To%20Murder%20Koshi%20Rikudo%20(AHQ).ogm)
Originally posted by Bogalvator
Did you try setting the IDCT setting to XviD? (under the "Miscellaneous" section)
Yes, I did. No use...

[Andy2222]

new SSE2 release up

NOTE: This is just a preview version, means its not as "rdy" as the older releases since im reworking kinda lots of stuff in the resizer and had no time to finish all.

I released this versions mainly cause milan added all this new audio stuff and this seemed usefull for the SSE2 version too.

PS: read the sse2 changelog first before u update since some small stuff is missing in this preview version!

u can download it from the sig location

[Soulhunter]

Yay, milan added the SRC resampling Ive requested... :cool:

Cant await the non SSE2 version !!!


Tia n' Bye

[Audionut]

Originally posted by Andy2222
new SSE2 release up


Thanks Andy2222 for your sse2 releases.
And to Athos for giving you somewhere to stash your realeases.:D

[Alvy]

Ffdshow really got awesome in its possiblilities :) !
Thank you for your work.

It would be great if the hotkeys settings could get more powerful. So the user could comfortably take control over ffdshow.

For my personal usage it would be fantastic it one could control:

misc:
video delay +/- (10 ms) per keypress

resize:
black borders
+/- per keypress or at least toggle on/off

much less important
zoom +/-
shapen strength +/-

I would be very glad if someone could improve the functionality of the hotkeys.

Thanks

Alvy

[Soulhunter]

Please could someone with a Audigy2 ZS help me to confirm something ???

Do the following steps with the 20040725 build...

- Set your Audigy2 to 5.1 setup

- Disable all extras (EAX, CMSS, EQ etc.)

- Connect a headphone to the L/R output of the card (or a 2ch amp)

- Activate ffdshows raw audio processing (no decoding/filters/mixer)

- Playback some stereo content (a simple mp3 file should do it...)


So, playback should be balanced (same loud @ L/R channel) !!!

When yes...

- Enable ffdshows audio resampling (96000Hz / Kaiser)

Do you hear a difference in the channel balance now (is one channel louder) ???


EDIT:

Addition to the last post...

Plz, try the above stuff with the default integer output (all checked -> 16/24/32) !!!

If the R channel gets louder (ca. 25%) after activating the resampling...

Disable 16bit integer and use 32bit floating point to see if it helps !!!


Tia n' Bye

[Soulhunter]

Ok, forget the stuff above !!!

Problem is solved with the last build... :)


Bye

[omion]

I've been having a problem with the latest (20040808) ffdshow build. It gives wierd 8x16 blocks in most b-frames of most movies. All of my movies are XviD or DivX (a whole bunch of versions... I don't know which ones are which) The blocks only appear when decoding b-frames with the latest build, and decoder set to libavcodec.

I have tested it in MPC and The Core Media Player, with the same results.

I have a P4 1.5GHz with 512MB RAM.

You can see an example here. (http://people.ucsc.edu/~rswilson/other/sakura-vs-ffdshow.jpg) Note that all the things that look like JPEG artifacts are actually JPEG artifacts. :p

Thanks!

[Ana]

I have the same problem with ffdshow (20040808) that omion has, when libavcodec is used to decode xvid. I think this has something to do with custom matricies.

Xvid (1.0.1) setting used to encode:

matrix: Andreas 78er
Qpel
GMC
max b-vobs: 2
VHQ 4
chroma on

[Gooop]

I don't think it has something to do with custom matricies, as I have the same problem on many videos which were encoded using h263.
2 bframes as well, QPEL but no gmc. I don't know if it can help but it seems that the problem is not exactly the same if the video is muxed in a mkv : in avi I can see the same artefacts as omion, but once muxed in mkv the video just crashes my player, mpc 6482 with its builtin matroska splitter. Another weird thing is that the player doesn't crash if I play the same mkv file with the same mpc but another matroskasplitter I registered (matroskasplitter_20040111-2). In this case the video plays with the same bugs as described before.
And I think (not sure, long time ago...) the problem occurs with videos that I encoded with older xvid builds as well (koepi-19062003-1).

[arno]

Is there already somebody out there that has some info on what "Mplayer Accurate Deblocking" exactly does (or has a link with an explanation)? If I interprete it right it sounds like a nice feature but a solid explanation would be nice :-D

[masken]

(repost from closed thread)

Is there any chance of improving the installer so that it supports silent installs? (/Silent switch for example).

...and perhaps also return an errorcode upon failure.

Perhaps also some additinal switches, like:
/DisplayProgress
/PostProcessing:[0-6]
/AutomaticQualityControl
...etc.

Or alternatively, the HKLM keys explained so one could just import values silently after an install to set the preferences.

This would really help me (and many other I guess) when installing new PC's....

[swalker]

Originally posted by masken
(repost from closed thread)

Is there any chance of improving the installer so that it supports silent installs? (/Silent switch for example).


It already does. The ffdshow installer is your standard NSIS based installer which supports /S for silent installations.

[winman]

@omion & ana

I also has the same weird blocks with some XviD and DviX movies. However, after changing the IDCT from the default "simple" to "auto", the weird blocks went away.

Previous version ffdshow-20040725 has IDCT default at "auto".

[omion]

Originally posted by Gooop
I don't think it has something to do with custom matricies, as I have the same problem on many videos which were encoded using h263.
2 bframes as well, QPEL but no gmc. I don't know if it can help but it seems that the problem is not exactly the same if the video is muxed in a mkv : in avi I can see the same artefacts as omion, but once muxed in mkv the video just crashes my player, mpc 6482 with its builtin matroska splitter. Another weird thing is that the player doesn't crash if I play the same mkv file with the same mpc but another matroskasplitter I registered (matroskasplitter_20040111-2). In this case the video plays with the same bugs as described before.
And I think (not sure, long time ago...) the problem occurs with videos that I encoded with older xvid builds as well (koepi-19062003-1).
I have strange things happen with Matroska, too. The screen shot I posted was from a Matroska, decoded with the external splitter (1.0.2.3 + sub patch) I demuxed the video, and tried to play the resulting AVI, but it crashed. Using the internal Matroska splitter, it also crashed.

Again, everything is magically fixed when using either an older version of ffdshow, or not using libavcodec.

I also just noticed that the blocks ONLY occur on the right half of a 16x16 macroblock, NEVER on the left.

@winman:
Doesn't help for me. :( I have mine set at XviD (libmpeg2) all the time.

[Blkbird]

OK, I asked this in the "old" thread, but it wasn't answered, so here I try it again.

Why are only TrueType fonts listed for subtitle display? The Type-1 and OpenType fonts are not showing up in the list.

[ViCroié]

I've been having a problem with the latest (20040808) ffdshow build. It gives wierd 8x16 blocks in most b-frames of most movies. All of my movies are XviD or DivX (a whole bunch of versions... I don't know which ones are which) The blocks only appear when decoding b-frames with the latest build, and decoder set to libavcodec.

I have the same problem, i tried both H.263 and Custom (HVS) matrice... but the only way for me to not have these blocks was to disable QPEL... when i use XviD's own decoder... it plays fine... for some encodes i used XviD with VHQ for B-Frames, but i tried XviD 1.0.1 and had same problem. :(

[aketon]

One good and one bad thing about ffdshow 20040808!

The good thing:
h.264 decoder can finally play encoded videos with all the analyzer flags enabled with no problems! :)

The bad thing:
Theora is broken!!!:(

[timeismoney]

I feel 0806 is better than 0808

I use 0808 to decode any AVI/MKV encoded by DivX Pro 5.11 will cause any player crash on my system, only videos encoded by DivX Pro 5.11, other version of divx is just fine

BTW, the audio codec is perfect under MPC and any version of WMP below 9.0, but under WMP9 it says like "Can't connect PINs", both AC3/DTS/AAC, and AAC is not friendly to some .aac and media contains aac audio track, m4a is all OK

[arno]

Originally posted by timeismoney
I feel 0806 is better than 0808

I use 0808 to decode any AVI/MKV encoded by DivX Pro 5.11 will cause any player crash on my system, only videos encoded by DivX Pro 5.11, other version of divx is just fine

BTW, the audio codec is perfect under MPC and any version of WMP below 9.0, but under WMP9 it says like "Can't connect PINs", both AC3/DTS/AAC, and AAC is not friendly to some .aac and media contains aac audio track, m4a is all OK

Yep, I agree. And 0808 also seems to hang my BSplayer a lot more often than 0806...

[LigH]

I just tried to run a test with makeAVIS, and read from another user that he does not get filtered output when he uses ffdshow's VfW interface to decode YV12 (he wanted to compare MPEG2Dec3's PP with ffdshow's PP). So I tested three configurations: ffdshow as VfW converter for YV12, ffdshow as frameserver for AVIS fake AVIs, and ffdshow as true decoder for FFV1 AVIs.

Result: No filter (deblocking/deringing, brightness/gamma, noise/grain) was applied to the decoded video in either case.

Is this intentional? Then I'm confused because the VfW interface looks like it provides image filtering, just like the DirectShow filter. Or may it be possible to miss one simple checkbox to enable image filtering in the VfW interface in general?

[marcellus]

Hi, I have two issues to bring to attention:

1. There is a problem with the Logoaway function. In mode "XY" and with a blur value equal and greater than 1 a green shade is cast over the logoaway rectangle that makes it easy to spot on an uniform background (otherwise it wouldnt be seen). The amplitude of this green shading increases in direct proportion with the blur setting (at 0 the problem doesn't show, at 10 is obvious). I use this function in realtime capture to increase compresibility.

Here is a frame that shows the issue. Look at the arrow to see greenish rectangle that should have the same colour as the surounding background.
Logoaway issue example (http://www.marcellusvcd.netfirms.com/Logoaway_01.html)

2. It seems there is an internal error in converting interlaced YUY2 to YV12. When I capture from YUY2 TV and I deinterlace in real time with almost any deinterlacer provided (including dscaler dll's) chroma of the resulted movie is somehow delayed half a frame from luma (This is easy seen at a scene change). When I check the "swap fields" box chroma is behind luma with half a frame. It's like the chroma of 2 consecutive frames is blended together. The only deintelacing method that does not show this "chroma delay" issue is "5 tap low pass" but that is normal since with this method the result is blended frames anyway.

Here is a succesion of frames that show the bug:
Frame1 (http://www.marcellusvcd.netfirms.com/Chr_blend_01.html)
Frame2 (http://www.marcellusvcd.netfirms.com/Chr_blend_02.html)
(Look at the arrow to see the chroma coresponding to the face of the character in frame number 1. The face of the curent frame character is pale because it doesn't have proper chroma like in Frame3.)
Frame3 (http://www.marcellusvcd.netfirms.com/Chr_blend_03.html)

best regards
marcellus

[stephanV]

Some issues with ffdshow's vfw MPEG4 encoder:

1. Enabling b-frames causes files to be undersized by a ratio of 1:5 or more

2. B-frames in combination with qpel crash VirtualDub (several versions) and also graphedit. Here's a crashinfo.txt (http://www.fesite.nl/grotesteph/crashinfo.txt)

Using qpel and b-frames with avs2avi gives me a ICSeqCompressFrame failed: The system cannot find the file specified

I cant remember this behaviour from earlier builds, although im not sure about the b-frames undersizing problems.

hope this is useful. :)

[Bogalvator]

I remember reading somewhere (here?) that ffdshow MPEG-4 with B-frames results in all B-frames being encoded at quant 31 when used in 2 pass - can someone confirm?

The avs2avi could be caused by the fact that ffdshow no longer installs the "ffvfw.dll" for the vfw section (guess work)


On a separate note: I notice that KernelDeint has succesfully been introduced into ffdshow. Would it also be possible to have DGBob integrated in to the deinterlace section without too much difficulty? This would be useful to those who store interlaced encodes on their PC (for things like sports clips, which aren't smooth enough when deinterlaced)

[stephanV]

just to be clear: if i didnt enable b-frames and qpel at the same time, avs2avi did work.

[Andy2222]

Originally posted by Bogalvator
On a separate note: I notice that KernelDeint has succesfully been introduced into ffdshow. Would it also be possible to have DGBob integrated in to the deinterlace section without too much difficulty? This would be useful to those who store interlaced encodes on their PC (for things like sports clips, which aren't smooth enough when deinterlaced)

DGBob was added by milan some days ago and can be used in a new release.

[LigH]

@ milan, athos, Andy2222:

Does ffdshow's VfW interface apply image filters at all? (^)

[Soulhunter]

Source...

http://img69.exs.cx/img69/2770/Source.jpg


Source -> VFW -> Blur 255 -> FFV1

http://img69.exs.cx/img69/1444/Blur255.jpg


Bye

[LigH]

When used as encoder: A little bit, it seems to me (shouldn't 255 soften a lot more?)...

But I'm rather interested in "when used as VfW decoder for AVIS".

[bond]

Originally posted by aketon
The good thing:
h.264 decoder can finally play encoded videos with all the analyzer flags enabled with no problems! :)what analyser flags?

The bad thing: Theora is broken!!!:/

[aketon]

Originally posted by bond
what analyser flags?

:/


Previous releases of ffdshow couldn't play properly the videos encoded with "Inter Analyzer Flag PSUB8x8" and "CABAC" ON at the same time! Now it can play them with no problems! Mplayer can play them also! VLC media player still can't play them properly!


Every time I am trying to encode with THEORA using ffdshow 20040808, the video I get is full of garbage in it! When I swithed back to previous releases of ffdshow, I could encode with no problems!


AS USUAL: "Sorry for my english"


BYE!!!

[bond]

ah yes indeed, afaik this has nothing to do with "analysing" but is basically the block size.
it should have worked for some time already, cause basically the ffmpeg guys fixed this because i made a bugreport at sourceforge, which was two months ago or so
i think i updated my h.264 sticky already with this info, i will have a look, ffmpegs h.264 decoder should be fully useable already i think

[athos]

New release up on SF. I decided not to include skl_drv_mpg.dll. Get it from earlier release, or here:
http://athos.leffe.dnsalias.com/skl_drv_mpg.zip

libavcodec.dll is smaller because i compiled without -funroll-loops.

mpeg-1/mpeg-2 decoding through libmpeg2 seems to work fine in this release, including seeking.

[Gooop]

thanks athos for the new build
xvid with bframes / qpel (don't know what caused the bug actually, didn't have time enough to test) works again, at least for me.

[athos]

Originally posted by Gooop
thanks athos for the new build
xvid with bframes / qpel (don't know what caused the bug actually, didn't have time enough to test) works again, at least for me.
It might have had something to do with the compiler flags i used for GCC. I tried to use only safe flags for this one.

[Worf]

Hi Athos,

i can't see the ffdshow-20040828 files in http://athos.leffe.dnsalias.com/?M=D
Im do something wrong?

Thanks

[Worf]

Oupsss!!! I find your file in SouceForge but not in your link... Sorry for my reply!

Thanks for your builds...

BR

[SeeMoreDigital]

Originally posted by Worf
Oupsss!!! I find your file in SouceForge but not in your link... Sorry for my reply!

Thanks for your builds...

BR Everybody please look here: -

http://sourceforge.net/projects/ffdshow/

[Shinobu]

too bad that vp3 decoding is alway bugs.
great works on audio filter.

++

[aketon]

At last! The latest version of ffdshow 20040828 has a new option to use bitrate in x264! It was about time!:D


[edit]
It is awesome! It has scene detection too (not so good though)! I've just encoded LOTR 3 trailer (resolution: 640x256 bitrate:900) and the result was good in the first 30 seconds, because after that 30 seconds, something strange happened to the codec! The encoder start to lower the bitrate from time to time and to use quantizers between 40-51!:confused: :confused:

[edit2]
I've just found where the problem is with the bad encode I mentioned! When I use "Inter Analyzer Flag PSUB8x8" and "CABAC" at the same time something strange is happening in the encoder and the result is to have (after a certain period of time) a video with quantizers 40-51! If you disable "Inter Analyzer Flag PSUB8x8" or "CABAC", then everything is going well!!!

[SeeMoreDigital]

Hi Athos and Milan,

Just a small thing. But can the 'menu' headings be altered slightly. My proposal is as follows: -

http://img9.exs.cx/img9/2975/SMD_FFdshow_proposal.gif

If it's okay with everybody that is?


Cheers

[BoNz1]

Originally posted by aketon
At last! The latest version of ffdshow 20040828 has a new option to use bitrate in x264! It was about time!:D


There is also a 2pass rate controller that is waiting to be added to the subversion.
EDIT: It would be nice if justin or min chen could update the vfw and the project file for the 1pass and 2pass. Please remove mc-c.c from the build also. Thanks.

[Phanton_13]

It's not posible to play VP3 using the theora library using a preprocesor filetr based in the reference lossles vp3 to theora trancoder .

[virus]

Originally posted by aketon
At last! The latest version of ffdshow 20040828 has a new option to use bitrate in x264! It was about time!:D

too bad the "1 pass fixed quantizer" option doesn't work anymore... it always encodes at roughly the same rate no matter what quantizer you use.
One good thing is that it apparently uses adaptive quantization within a frame (unless ffdshow's visualization tool is b0rked ;)). Some blocks are encoded at QP=24, others up to QP=30. Nice. But still too much bugs! :(

[aketon]

Originally posted by virus
too bad the "1 pass fixed quantizer" option doesn't work anymore... it always encodes at roughly the same rate no matter what quantizer you use.

Yes! The only working good (not so good though) option is "1 pass constant bitrate"! "1 pass fixed quantizer" is useless right now and it hasn't got scene change detection also!

Originally posted by virus
But still too much bugs!

Yes, inded!!!:(

[pankov]

@Athos, Andy & Milan
Gyus can you cooperate for integrating the "Green Fix" that Andy did
for the SSE2 builds for the non SSE2 builds?

2004-07-01 Andy2222 (ffdshow-20040701_SSE2.exe)
...
* fixed "green" shift resize bug
...


I really can't believe that this is not the the top priority fix as this is visible on all NVidia cards and especially in VMR9 mode which is supposed to be the mode used in the future.
I'm on the brink of stop using FFDShow for resizing and I will really hate to do this

PLEAAAASE, guys, kill this annoying bug

[Andy2222]

Originally posted by pankov
@Athos, Andy & Milan
Gyus can you cooperate for integrating the "Green Fix" that Andy did
for the SSE2 builds for the non SSE2 builds?

2004-07-01 Andy2222 (ffdshow-20040701_SSE2.exe)
...
* fixed "green" shift resize bug
...


I really can't believe that this is not the the top priority fix as this is visible on all NVidia cards and especially in VMR9 mode which is supposed to be the mode used in the future.
I'm on the brink of stop using FFDShow for resizing and I will really hate to do this

PLEAAAASE, guys, kill this annoying bug

i will write a normal mmx2 version for the normal build if my relocation chaos (new job, new location ... need to visit ikea, get a fridge..) is over

[marcellus]

Hi, in latest build (20040828) avisynth processing inside ffdshow is broken :(.

[dZeus]

Hi, are there any plans to improve the RGB conversion routines in ffdshow? I wanted to do a comparison of mpeg2 decoder filters by capturing frames with ffdshow, but I keep getting horrible artifacts caused by the conversion algorythm in ffdshow to rgb colourspace rather than the clean output I see on the overlay/vmr on my screen.

The routines in dvd2avi (http://arbor.ee.ntu.edu.tw/~jackei/dvd2avi/) produce much, much better results than the one in ffdshow (and also are GPL, so easy to copy/implement? ;) ).

[Affar]

i Don´t know if developers know this error in last versiones of FFDSHOW (ALPHA) when is decoding old XviD compressed videos.

http://www.divxhouse.com/temp/error_ffdshow.28.8.2004.jpg

And here little sample of this error >> here (http://www.deco-flower.com/error_ffdshow.28.8.2004.zip)

Thanks and seeya :)

[stephanV]

file works fine with ffdshow 28082004 in WMP and MPC...

[Affar]

Ops, the problem is in old configuration. Unistall and install and error is disappears :)

Sorry for the error :/

[Rober2D2]

About MPEG1, MPEG2 support.

1) Thank you for fixing decoding. It's been a long time :)

2) Aspect ratio does not work correctly for the moment. I am using bsplayer 1.0 and ffdshow 28-08-2004. Using an alternative MPEG decoder (intervideo for example), display correct aspect ratio.

[Krismen]

Hi

I have problem with mov video clips encoded with Sorenson video 3. I've got ffdshow-20040808 and 3ivx 4.5.1 installed. In ffdshow configuration svq3 is set to libavcodec. Allow Unsupported Decoders is checked in 3ivx media splitter configuration.


Here is a sample frame (no. 324) form Ocean 12 trailer (http://www.apple.com/trailers/wb/oceans_12/medium.html):

http://priv.twoje-sudety.pl/~divxpl/Dj_AnT/o12_bad_324.png


It was saved in VDMod using following script as input:
LoadPlugin("C:\Program files\AviSynth 2.5\plugins\directshowsource.dll")
DirectShowSource("o12.mov")


I've checked all IDCT algorithms if ffdshow properties but clip still dosen't plays properly.


I can watch it without any problem when I install QT Alternative, but it would be nice to wrap some SV3 clips into AVI container using GraphEdit (without any recompress, except qdesign audio :() and decode it through ffdshow.


In this (http://forum.doom9.org/showthread.php?s=&threadid=48511&perpage=20&highlight=svq3&pagenumber=16) thread:
Originally posted by ferrous66
Now that the latest unofficial build includes the Sorenson 3 codec, can you playback .mov files with ffdshow + 3ivx .mov splitter?Originally posted by Stux
It works with our latest builds, not sure what happens with 4.0.4

I thought it shouldn't be any problem with playing SV3, but I was wrong :-(.


Regards

[SeeMoreDigital]

Have you at anytime installed QuickTime Alternative?

If you have and are also running an official version of the QuickTime player, you might run into a few problems!

If you have not installed QuickTime Alternative you could try installing and registering this QuickTime DSdecoder filter (http://82.2.167.24/Uploaded_Files/Doom9_Forum_files/QuickTime_DSdec_Filter.zip).

Let us know how you get on please?


Cheers

[Krismen]

Have you at anytime installed QuickTime Alternative?
Yes, but I completly unistalled it (I had to uninstall QT Player as well) and installed QT Player and Cyberlink qt ds source filter. When I done this I couldn't manually construct filter graph in GraphEdit. GE crashed all the time when connecting output pin of File source async. to 3ivx media splitter. So I unistalled ds filter and everything back to normal except decoding :(.

If you have and are also running an official version of the QuickTime player, you might run into a few problems!
Yes I was aware of that. That's why I installed only ds filter later.

In a few days time I'll check how this trailer is decoded on another PC and then I'll post the result.

[lalo]

Hi Krismen!

You can try Nero's directshow filters in GE with installed quicktime alternative. I saved the graph and load it with directshowsource in avisynth. I've reencoded some trailers with this method in virtualdub.

You need: NeQTVdec.ax, NeQTAdec.ax, ndparser.ax and Avisynth 2.54 or newer.


It works fine me.

lalo

[SeeMoreDigital]

Originally posted by lalo
You can try Nero's directshow filters in GE with installed quicktime alternative. I saved the graph and load it with directshowsource in avisynth. I've reencoded some trailers with this method in virtualdub.

You need: NeQTVdec.ax, NeQTAdec.ax, ndparser.ax and Avisynth 2.54 or newer.Hi lalo,

Can you confirm if you've had any success using Nero's filters to decode "Sorenson/QD2 .mov" files (as currently used in all Apple QuickTime Trailers)?

I've only been able to use Nero's QT filters to decode "Mpeg4/AAC .mov" files - It would be good to have clarification?


Cheers

[lalo]

Hi SeeMoreDigital,

Yes, I managed to decode "Sorenson3/QD2 .mov" files using Nero's filters.


Cheers

[SeeMoreDigital]

Originally posted by lalo
Yes, I managed to decode "Sorenson3/QD2 .mov" files using Nero's filters. Jeez this is good news!

Have you able to play "Sorenson3/QD2 .mov" files in Nero's ShowTime player? I can't!


Cheers

[easyfab]

I see that the cvs was update with 2 pass support for x264 :)
With the next binaries release the true x264 test will begin.

[Blight]

The nero filters are riddled with bugs, they don't detect all audio tracks, they won't connect to ffdshow for Sorenson 3 decoding, they won't accept a disconnect/reconnect of the filter (required to fix the VMR9 scaling bug that is STILL in DirectShow, even after SP2/DirectX9c, etc... a shame, they are quite close to playing most quicktime files.

And I did get them to decode the QDesign Music 2 audio format.

[aketon]

Originally posted by easyfab
I see that the cvs was update with 2 pass support for x264 :)
With the next binaries release the true x264 test will begin.

Great news!:)

[aketon]

Is there anybody that can encode properly to theora with ffdshow?? In the last two versions of ffdshow, theora seems to be broken! Am I the only one who see this problem? Or it is just something that I did wrong during encoding?:(

[Krismen]

I've tested trailer mentioned ealier on another PC and it was played the same way as on my machine (using 3ivx + ffdshow). Then I converted this trailer into AVI with mov2avi using Sorenson Video 3 compression and ffdshow played this AVI correctly. So I think it's probably (???) 3ivx's fault not ffdshow.

[virus]

*bump* :D

@Athos, Andy
please pretty please, could you prepare a new build? The last one is 1 month old right now and there are a lot of improvements in x264 that we like to test.

Also: could you pass on to Milan the request to change the Quantizer slider when encoding to H.264 to the proper range defined in AVC (1-51) instead of the 1-31 used by default? That would be very appreciated ;)

cheers & thank you :)
virus

[Kurtnoise]

I've made a compile of ff_x.264.dll (http://atlas2.tgv.net/~media-video/forum2/download.php?id=1452) from the CVS...;) Just put it in the ffdshow directory and test it.

[Tommy Carrot]

Originally posted by Kurtnoise13
I've made a compile of ff_x.264.dll (http://atlas2.tgv.net/~media-video/forum2/download.php?id=1452) from the CVS...;) Just put it in the ffdshow directory and test it.
Thank you for your effort, but it crashes when i want to encode with it. And another thing is that we cannot use the 2-pass RC if only the dll is updated, since the older ffdshow doesn't have it in the GUI.

[Kurtnoise]

arff yes. Sorry :( I haven't test it.

Stay tune...

[virus]

Originally posted by Tommy Carrot
And another thing is that we cannot use the 2-pass RC if only the dll is updated, since the older ffdshow doesn't have it in the GUI.
yep. also the new options for subpel refinement (level range from 0 to 5) would require a new GUI element.

[colin.findlay]

Sorry if this is a little off-topic, but does anyone know if it's possible to force Windows Media Player 9+ to use the ffdShow audio MP3 decoder by default? WMP seems to only use l3codeca.acm, and not even try to use any of the directshow filters.
It's bizzare....

[LigH]

Since I know why, I would never install any newer MS-WMP version, but rather use really powerful media players instead (like Media Player Classic, VideoLan Client, ZoomPlayer, BSPlayer, The Core Media Player, ...) - I can control them as I decide, and they don't phone home.

You may try to disable or even uninstall this outdated MP3 codec (Control Panel, Sound & MM, Audio Codecs...). If this makes problems, you may re-install the Radium codec afterwards. And if you decide to substitute it, you may try the LAME ACM (RareWares, MP3, latest stable pack, right-click on .INF, Install).

[DKDIB]

LigH wrote:
> [...] if you decide to substitute it, you may try the LAME ACM
> (RareWares, MP3, latest stable pack, right-click on .INF, Install).

lameACM is not an audio decoder (unlike other implementations of LAME).

[LigH]

You mean, it is an "encoder only"? It does not contain mpg123 for decoding MP3? - Well... Is there a MAD "decoder only" ACM codec? :p

Anyway: For playback, a DS filter shall be enough, and for conversion, better use import plugins of your favourite editor, BeSweet/HeadAC3he for direct conversions, or MAD bundle for HQ decoding to WAV.

[aketon]

Originally posted by virus
yep. also the new options for subpel refinement (level range from 0 to 5) would require a new GUI element.

I believe that x264 must start have its one GUI! Using ffdshow to encode isn't something I really like! Unfortunately I don't know anything about programming to make one and the only person that I know who made a GUI ("Justin Clay" I thing) has stopped! I hope that someone can spend some time to make a new one with the new options on it!!!

[Sirber]

FFDshow ain't that bad. Just get used to it. Less troubles :)

[Tommy Carrot]

Originally posted by aketon
I believe that x264 must start have its one GUI! Using ffdshow to encode isn't something I really like! Unfortunately I don't know anything about programming to make one and the only person that I know who made a GUI ("Justin Clay" I thing) has stopped! I hope that someone can spend some time to make a new one with the new options on it!!!

Oh, FFDShow does have the subpel refinement option (and IMO there is nothing wrong with the ffdshow GUI), just a build from it has not been released so far. But don't worry, you don't miss too much, the new option doesn't make wonders, i couldn't see any significant quality gain with subpel ref. 5. Not to mention that the h.264 encoder of the current cvs version is severely buggy (i don't know if this is x264's or ffdshow's fault).

[aketon]

Originally posted by Sirber
FFDshow ain't that bad. Just get used to it. Less troubles
Originally posted by Tommy Carrot
Oh, FFDShow does have the subpel refinement option (and IMO there is nothing wrong with the ffdshow GUI), just a build from it has not been released so far. But don't worry, you don't miss too much, the new option doesn't make wonders, i couldn't see any significant quality gain with subpel ref. 5. Not to mention that the h.264 encoder of the current cvs version is severely buggy (i don't know if this is x264's or ffdshow's fault).

It is not just that! I know some poeple that don't really like to use ffdshow (because when they see all these stuff in it, they find it difficult, they see it difficult)! They prefer to have each one codec installed separately in their system from the others! It has to do with how a simple user see it!!! I can't explain exactly what I fully want to say (because my english are not very good), but I think that you can understand what I am trying to say!:rolleyes:

Simple things are better to use! All in one things are difficult!(from the newbie aspect)!!!;)

[SeeMoreDigital]

Can anybody confirm whether FFdshow's Mpeg1/2 decoder filter is able to auto DAR content correctly?


Cheers

[Sharktooth]

No new builds since august... uhm...

[athos]

New build up @ SF.

[BoNz1]

Originally posted by athos
New build up @ SF.

Cool :cool: 2-pass x264 encoding :).

[Sirber]

I just checked this morning :(

http://prdownloads.sourceforge.net/ffdshow/ffdshow-20041003.exe?download

[BoNz1]

Originally posted by Sirber
I just checked this morning :(

http://prdownloads.sourceforge.net/ffdshow/ffdshow-20041003.exe?download

Yeah, it isn't there but it is in the changelog :(.

[SeeMoreDigital]

Just installed ffdshow-20041003

Is it just me, or is Mpeg1 and Mpeg2 decoding b0rked :(

DTS works nicely though!

[Sirber]

Is x2674 updated? Haven't a chance yet to install it. I'm in the middle of an encoding :(

[SeeMoreDigital]

Originally posted by Sirber
Is x2674 updated? Dunno about x2674 my friend ;). But here's what Doom9 says about the changes: - ffdshow 20041003 (http://osdn.dl.sourceforge.net/sourceforge/ffdshow/ffdshow-20041003.exe) contains more than a long page of release notes (https://sourceforge.net/project/shownotes.php?release_id=272514) (as compared to the August 8th release I previously had for download).

We're basically talking about improved MKV and OGM support, various improvements in the subtitle area, support for Nero MPEG4 video, AAC audio support using Nero file source, x264 improvements (a H.264 codec) and the latest libavcodec is being used

Cheers

[Tommy Carrot]

Originally posted by Sirber
Is x264 updated? Haven't a chance yet to install it. I'm in the middle of an encoding :(

Yes, there is the new subpel refinement option, but still no 2-pass. However, "more iterations" mode (the slowest subpel ref. mode) seems to be borked, ugly artifacts appear with it sometimes. The default mode (the second) seems to be ok though. I didn't try the other modes yet, so i can't comment on them.

[netchris]

Yes x264 is updated compared to the ffdshow built of 28/8/04.
Two pass is not available yet.

Some days ago Tommy Carrot (if I remember correctly) posted an ffdshow built. The built was removed some hours later, but I played with it and the x264 codec had some encoding bugs - white blocks appearing randomly, when using anything else than halfpel.
The next day I finally managed to compile on my own the ffdshow code (the 2/10/2004 one), and my built does not have these encoding bugs whatever subpel refinements I choose, but it is quite slower. Testing now the new 03.10 built and the encoding bugs are back (with the speed being as fast as tommy carrots).
The speed difference can be because i did not use any cpu optimazations (I dont know how), or because i got too many warnings when building the code. I used the visual studio net compiler.
I dont understand why the bugs are back,I guess they used older x264 code than the 2/10 one but cant understand why they would use older code.

Anyway the encodes made with my built are very nice, I can only hope the two pass option will make x264 more competitive.

Thanks for your time :)

*********************************************
update : Yes Tommy I am talking abouth these artifacts that are not present in my built

[colin.findlay]

Now, the reason I want WMP to use the ffdshow audio decoder is 2-fold.
a) I'm using a HTPC solution (mediaportal) which integrates with WMP only at this point.
b) I'd like ffdshow to auto-load a different preset for mp3 files. In this case 2.0 stereo out for mp3/wma/ogg & 2.0 stereo mixed to 5.1 out for any other stereo source. (ie. divx with mp3)

I may just end up assigning the presets to hot keys (if poss) and setting up my remote so I can manually switch (not nearly as cool)

BTW. You can replace the ACM with others, but what I would really need is an ACM wrapper for ffdshow with decode-only support, and maybe encoder pass-thru.

Regards
C.

Originally posted by LigH
Since I know why, I would never install any newer MS-WMP version, but rather use really powerful media players instead (like Media Player Classic, VideoLan Client, ZoomPlayer, BSPlayer, The Core Media Player, ...) - I can control them as I decide, and they don't phone home.

You may try to disable or even uninstall this outdated MP3 codec (Control Panel, Sound & MM, Audio Codecs...). If this makes problems, you may re-install the Radium codec afterwards. And if you decide to substitute it, you may try the LAME ACM (RareWares, MP3, latest stable pack, right-click on .INF, Install).

[Tommy Carrot]

Originally posted by netchris
update : Yes Tommy I am talking abouth these artifacts that are not present in my built
If they didn't occured in your build, i suspect that using the cpu optimizations at compilation causes this bug, since the x264 code wasn't touched in the last few days, probably that's the only difference between the 2 builds.

Edit: ok, i was wrong, there were changes in the x264 code 3 days ago, but i still stay with my opinion that using too strong optimization settings at compiling can cause those bugs.

[obieobieobie]

This latest ffdshow seems to crash the Windows Explorer shell when viewing a folder with video files.

[aketon]

Originally posted by Tommy Carrot
Yes, there is the new subpel refinement option, but still no 2-pass. However, "more iterations" mode (the slowest subpel ref. mode) seems to be borked, ugly artifacts appear with it sometimes. The default mode (the second) seems to be ok though. I didn't try the other modes yet, so i can't comment on them.

But in the changelog:

2004-09-09 20:20 milan_cutka

* x264 two pass encoding

:confused: :confused: :confused:

[marcellus]

Although the change log says this:
2004-09-06 08:37 milan_cutka

* use avisynth filter again

2004-09-06 08:30 milan_cutka

* use avisynth filter again
avisynth filtering inside ffdshow is still not working. Last working version was 20040808.
And I really need this.
:( :( :(
Edit:
To give more details:
-When I use it via vfw interface for realtime captures avisynth filtering simply does nothing
-When I open files via MPC (directshow interface) avisynth filtering does nothing and it shows this error on screen:
http://img59.exs.cx/img59/2360/snapshot_01.jpg
Edit 2:
Forgot to mention that is not particular script related, the simplest script like invert() or Info() triggers the error

[aketon]

At least, with this new build, I can encode to theora again!:)

[Leak]

Originally posted by marcellus
Although the change log says this:

avisynth filtering inside ffdshow is still not working. Last working version was 20040808.
And I really need this.

I'd assume he was writing about the AviSynth filter that's now included with ffdshow so you can use ffdshow's functions in AviSynth instead of the other way round. Of course, a fixed AviSynth support in ffdshow would be very nice too...

np: Yoko Kanno - Cream (Ghost In The Shell Stand Alone Complex OST Be Human)

[marcellus]

I'd assume he was writing about the AviSynth filter that's now included with ffdshow so you can use ffdshow's functions in AviSynth instead of the other way round.
Yes, that passed thru my mind too.
Of course, a fixed AviSynth support in ffdshow would be very nice too...
For me is a must. And I don't understand why got broken, for long time, untill version 20040828, it worked fine. If I copy ffdshow.ax ver. 20040808 over the current one it works for my purpose but it's not 100% safe, strange things could happen anytime.

I use ffdshow with my TV capture program to encode real time mpeg2 (SVCD as target resolution). This task is very demanding for my CPU so to be sure I don't have dropped frames I must use this chain:

Capture(704*576)->
(ffdShow)Crop(656*512)->
(ffdShow)Resize(448*512)->
(ffdShow)denoise3D->
(avisynth)[Decomb(eliminate phase shift)->KernelDeintMMX->AddBorders(480*576)]->
(ffdShow)Encode

This way intensive tasks as denoising and deinterlacing are made on a 448x512 picture.
The last part, especially Decomb and AddBorders I can't do it with FFDshow's filters, so I have to do it in AviSynth.
Of course I could do it at a later encode but my purpose is to obtain a SVCD compliant mpeg2 file without the need of a reencode (at least the video part), I'm pretty happy with the quality and I don't want to spend more CPU time for my every day quick and dirty captures.
To dream a little, I wish we could put in ffdshow 2 or more instances of one filter, for example I could use denoise3d after deinterlace and before addborders (so I would have to use avisynth 2 times).

[virus]

Looks like x264 2-pass is available, but not in the usual way. Here's what Milan wrote on SourceForge after I've submitted my requests:

by Milan Cutka:

x264 uses its own two pass algorithm which is similar to libavcodec's and so is the way to use it.
To do the first pass, select constant quantizer (preferred) or constant bitrate and on the Ouput page select Write from "Libavcodec stats" group (sorry, forgot to add x264). In the edit box below you can enter first pass stats file name. For second pass, select constant bitrate and select "Use"
from "Libavcodec stats group. Sorry, I know this isn't very user friendly, but it's because ffdshow's own two pass routines can't be used.

[SeeMoreDigital]

Originally posted by obieobieobie
This latest ffdshow seems to crash the Windows Explorer shell when viewing a folder with video files. Yes, I've noticed this too!

As soon as I have H.264 "libavcodec" enabled and go a hunting for my h.264 encodes, I receive a "Windows Explorer" warning...


Cheers

[obieobieobie]

It is added to the bugtracker over at the ffdshow sourceforge project page. Milan posted a new build (requires an sse capable cpu) which has fixed the problem at least for me.


link to bugthread (http://sourceforge.net/tracker/index.php?func=detail&aid=1039593&group_id=53761&atid=471489)

[Stereodude]

Any change FFDShow can be fixed to read the aspect ratio from Matroska files correctly?

I have 3 computers and only on 1 of them does FFDShow correctly use the aspect ratio / display size value in the .MKV even though the same exact software and filters are used on all 3.

[SeeMoreDigital]

Originally posted by Stereodude
Any change FFDShow can be fixed to read the aspect ratio from Matroska files correctly?

I have 3 computers and only on 1 of them does FFDShow correctly use the aspect ratio / display size value in the .MKV even though the same exact software and filters are used on all 3. What streams have you muxed into MKV?

The current build can't correctly AR Mpeg1 or 2. I've also tried FFdshow with anamorphic 3ivx, XviD, Nero and DivX Mpeg4 streams, in both AVI and MKV. And sadly it does not work!

So for me, I'm sticking with one of XviD's test DSdec filters with auto AR detection!


Cheers

[Stereodude]

Originally posted by SeeMoreDigital
What streams have you muxed into MKV?

The current build can't correctly AR Mpeg1 or 2. I've also tried FFdshow with anamorphic 3ivx, XviD, Nero and DivX Mpeg4 streams, in both AVI and MKV. And sadly it does not work!

So for me, I'm sticking with one of XviD's test DSdec filters with auto AR detection!


Cheers
I was doing Anamorphic 1080x720 xvid decoded to 1280x720 with AC3 audio. It works on one of my computers with FFDShow, the other 2 do not play it back correctly with FFDshow. They all work fine with using the official 1.02 xvid decoder.

[SeeMoreDigital]

Originally posted by Stereodude
I was doing Anamorphic 1080x720 xvid decoded to 1280x720 with AC3 audio. It works on one of my computers with FFDShow, the other 2 do not play it back correctly with FFDshow. They all work fine with using the official 1.02 xvid decoder. If you're re-encoding you're source to 1280x720 (with square) pixels then the encode is no longer anamorphic. It's a "true 16:9 frame" encode!


Cheers

[Stereodude]

Originally posted by SeeMoreDigital
If you're re-encoding you're source to 1280x720 (with square) pixels then the encode is no longer anamorphic. It's a "true 16:9 frame" encode!


Cheers
As I said I am encoding to 1080x720, and playing back at 1280x720. That is anamorphic.

[SeeMoreDigital]

Sorry my mistake!

1080x720 seems a rather odd source size. How did you manage to arrive at it?


Cheers

[virus]

OK, about x264: I've tested both ffdshow-20041003 and ffdshow-20041005-sse and both have the "blocking bug" when subpel refinement is set to something different than "hpel only".

It has been suggested that the problem may be due to compiler optimizations. So: what compiler has been used for those builds? What compiler flags? Is there a way to have a different build with different compiler/params to test it? At least, if we understand where the error actually is, we will be able to submit a bugreport...

some help please? :)

[celtic_druid]

http://celticdruid.no-ip.com/test/ff_x2641.7z
http://celticdruid.no-ip.com/test/ff_x2642.7z

First one I tried and it looked fine. Second one I don't know but it should be faster.

http://celticdruid.no-ip.com/test/ff_x2643.7z
http://celticdruid.no-ip.com/test/ff_x2644.7z
http://celticdruid.no-ip.com/test/ff_x2645.7z
http://celticdruid.no-ip.com/test/ff_x2646.7z

[virus]

thank you celtic_druid, I'm currently testing your 2nd build and the results seem strange to me. 95% of the blocks are gone but not all! There are still very few small artifacts... I'm trying to narrow down the problem and test more.

Anyway the PSNR results for x264 start to be pretty good (better than XviD). I'll report back on the x264 development thread about them later.

[netchris]

Thanks a lot celtic_druid!
The first and second builds are fine,they have exactly the same behaviour as my builds, so they are slow but block-bug free.
The files from 3 to 6 are twice as fast but they all have the block-bug.
What did you change in these builds?

95% of the blocks are gone but not all! There are still very few small artifacts...

The artifacts you see might be unrelated to the bug. I believe they are encoder immaturities.

Also the last mplayer/mencoder build that I did has x264 enabled. Build is optimised for athlon-xp's though.

could you please provide a link for the mencoder build?

[celtic_druid]

2nd build should be reasonably fast. Think it was ICL 7.1 with /O3 /G6 /Qipo /Qunroll /QaxiMK

3rd one was I think -O4 march=pentiumpro mtune=pentiumpro -pipe -ffast-math -fomit-frame-pointer
4th was basically default without -funroll-loops
5th was basically default without -finline
6th was basically default without -finline-functions

Here ya go:
http://celticdruid.no-ip.com/test/ff_x2647.7z
gcc without march, ffast-math or anything.

Hmmm, hang on I don't think that the MSVC project files have nasm enabled. Perhaps that is where the problem lies? Would also explain the speed difference.

[netchris]

The 7th build is slow again (a little faster than second) but without the block bug.


Hmmm, hang on I don't think that the MSVC project files have nasm enabled. Perhaps that is where the problem lies? Would also explain the speed difference.

I hope you are right so that we find the source of all evil.

[virus]

things start getting damn messy... 2nd build as before, but this time I've rised max ref. frames from 1 to 3... and guess what? A blockfeast! Even big blue blocks on white water... :(

I'm gonna check the 7th build with the same settings to see what happens. Right now the "bad" options are subpel ref. 4 ("qpel on all") + max ref. frames 3 + CABAC + in-loop filter + all analyze flags on.

@netchris: could you please try to re-encode with the settings above and see what happens?

[netchris]

I tried 2nd and 7th build that seem identical in quallity, with the options you asked but i dont see any problems here. I always used 3 reference frames. Subpel ref. 4 as well as 5 are clean in my tests.
Maybe there is something wrong with reference frames?(i doubt it)
I'll be gone for some hours so no more tests till then.

[celtic_druid]

Definatly no asm in the msvc project files. I couldn't get it to compile with asm when I added it either. Build 7 has asm, just no gcc opts so if 7 is ok then it isn't a problem with the asm anyway.

[virus]

Tested build 7. No way...
I believe it's a decoding problem. I've verified it under VDub and MPC. Seems to happen when you watch the "buggy" scene, go backwards and then rewatch the same scene. First display is OK, the 2nd is a blockfeast. According to a failed assert thrown by Vdub during decoding, the culprit may be libavcodec (or maybe the libavcodec sources embedded in ffdshow). More later.

[Tommy Carrot]

Originally posted by virus
I believe it's a decoding problem.
Definitely not, because the corrupted blocks are still there with the ateme decoder filter.

I checked build #1, and while it's slow as hell, at least it's working correctly (except the upper-left corner underquantizing problem).

[netchris]

It is not a decoding problem, when you have disabled 264 decoding support in ffdshow and decode it with nero's decoder the problem remains.

First display is OK, the 2nd is a blockfeast. According to a failed assert thrown by Vdub during decoding, the culprit may be libavcodec

what do you meen? In the first display the picture is ok? (without white block bug?)If that is the case, then there is a decoding bug that is unrelated to the white blocks. Appart from the decoding bug, the white blocks are definatelly an encoder's bug.

The problem must lie somewhere in icl's optimizations. I compiled with msvc(version 2003)choosing different options (with and without optimizations) and there never is a white block bug.The only problem is the resulting x264 dll is always quite slow (about as slow as builts 2 and 7). I also tried compiling with mingw-msys an the resulting dll is as fast as icl's fast builds but it also has the white blocks (though the bug might be less frequent with mingw,i am not yet sure about this).
I Also agree that nasm is not the problem as when compiling with msvc it is always enabled (though I dont know if it is being used by the code).

******************************************************************

I apologise in advance if I say something stupid,because I have no previous experience with the compilers, I am currently learning many basic stuff, thanks to x264 :) .
So if i dont make sense in some cases please be patient and you are welcome to correct me.


edit :
Tommy I didn't see your post while I was writing mine :) nice timing

[virus]

oh well, let's start from scratch then :rolleyes:

Let's say the decoding troubles are just a consequence of a broken bitstream then. I've tried to cut the video at the keyframe just before the Bad Things kick in... when I open the resulting clip VDub says that the first frame is bad, the rest good but undecodable (of course!).
But if I cut it a bit early everything works... and the bad blocks only appear if you go forward and backwards when decoding... well, in fact the decoding is rather unpredictable. Sometimes it displays almost no blocks, sometimes a lot of big bad blocks :rolleyes:

Anyway: build 2 and 7 both have the problem for me. Build 2 is ICL, build 7 is gcc... so it cannot be a bug in the compiler I'd say. I'll give a shot at build 1, then I'll try to invent something different.

:confused:

[virus]

All attempts failed miserably. Same problem.

Here's a sample (http://www.webalice.it/riccardo.stievano/video_stuff/x264-b0rked.avi) (930 KB) of my troubled clip, encoded with celtic_druid's build #1.

If you decode it starting from the 1st frame, you should not see any blocks (though I have the impression that some artifacts still lurk somewhere, I should make a frame-by-frame comparison with the source for that). BUT if you decode starting from the second keyframe, bad blocks should appear. If they don't appear, try going forward and backwards and then restart decoding again from the 2nd keyframe. I've checked that with both VDub and MPC.

I hope some of you can confirm what I'm seeing here.
No clue about what's going on. :(

[snacky]

That's not an encoder bug.

In H.264, a p-frame can be predicted from multiple previous reference frames, including (possibly) frames preceding the last iframe. This is the case for the first i-frame in your sample video. This particular i-frame shouldn't be used for seeking directly to, and shouldn't be used as a keyframe.

[virus]

Originally posted by snacky
This particular i-frame shouldn't be used for seeking directly to, and shouldn't be used as a keyframe.
Are you saying that seeking/cutting reliably in H.264 is impossible if you use multiple ref. frames?
If you don't seek to a keyframe, where are you supposed to seek then?

This won't explain the blocks some of us saw using different builds, though. I've seen blocks appear in the middle of a 1000 frames long sequence, decoding from the start (see the scrshots in the x264 thread). Others reported similar problems. Maybe 2 different issues?

[netchris]

I've seen blocks appear in the middle of a 1000 frames long sequence, decoding from the start (see the scrshots in the x264 thread). Others reported similar problems. Maybe 2 different issues?

Please dont mistake it with the white blocks bug. This is an encoding bug, and what you mention is a decoding one. Two seperate bugs. I understand why you mistake them, they produce a similar effect so it might be confusing.

And one question gcc = mingw? isn't gcc linux only?

[akupenguin]

Originally posted by virus
Are you saying that seeking/cutting reliably in H.264 is impossible if you use multiple ref. frames? No, but the seeker/cutter has to distinguish between an I-frame (can be decoded independently, but says nothing about the following frames) and and IDR-frame (which is an I-frame that is guaranteed to be seekable)

Originally posted by netchris
And one question gcc = mingw? isn't gcc linux only? gcc has been ported to everything under the sun. mingw ("Minimalist GNU for Windows") is one such port.

[Stereodude]

Originally posted by SeeMoreDigital
Sorry my mistake!

1080x720 seems a rather odd source size. How did you manage to arrive at it?
I used the native content ratio as DVDs do 1.5:1 which gets anamorphically stretched to 1.78:1.

[virus]

Originally posted by akupenguin
No, but the seeker/cutter has to distinguish between an I-frame (can be decoded independently, but says nothing about the following frames) and and IDR-frame (which is an I-frame that is guaranteed to be seekable)
uhm, I can foresee several problems because of this. No tool that I know of make such distinction (but maybe bond can say something more on that :))

Is there an option in x264 to avoid that P-VOPs reference frames before the I-VOP they depend on? Something similar to "closed GOV" for XviD? Maybe it would be useful.

[Alvy]

Versions before I controlled ffdshow via girder with keyboard commands.
This does not work any more.
I tried all options like "remote control api" and "accept keyboard messages".
Is there a way to get this to work or is that a bug :confused:

[akupenguin]

Originally posted by virus
uhm, I can foresee several problems because of this. No tool that I know of make such distinction (but maybe bond can say something more on that)
If the tool is H.264-aware, it's trivial to make the distinction by looking in the slice-header. Of course, no existing codec-agnostic tool does so, because the distinction didn't exist before H.264.
And as long as the muxer is H.264-aware, the demuxer/seeker doesn't need to be: You just mark non-IDR I-frames as not keyframes (because they really aren't).

Is there an option in x264 to avoid that P-VOPs reference frames before the I-VOP they depend on? Something similar to "closed GOV" for XviD? Maybe it would be useful.
GOVs are already closed. It's just that 1 GOV can contain multiple I-frames.
To answer your question, set IDR interval = 1. (may also be called idrint or idrframe, I don't know what the ffdshow interface looks like.) That will force every I-frame to be IDR.

... Come to think of it, I can't see any function in non-IDR I-frames, except in the rare case of (scene 1)(scene 2)(scene 1) where scene 2 wants an I-frame, but the return of scene 1 could use a long-term reference instead of another I-frame. Still, the codec would have to be smart enough to look that far ahead. And for that matter, you only lose a small overhead for using a P-frame full of I-blocks instead of an I-frame.

[virus]

akupenguin, every time you post we learn something new about H.264 :)

I've opened a feature request item on Sourceforge to have the "IDR interval" option added to the ffdshow GUI (it's not available right now) along with the "deblock{alpha|beta}" one which is missing as well.

[snacky]

Originally posted by akupenguin
... Come to think of it, I can't see any function in non-IDR I-frames, except in the rare case of (scene 1)(scene 2)(scene 1) where scene 2 wants an I-frame, but the return of scene 1 could use a long-term reference instead of another I-frame. Still, the codec would have to be smart enough to look that far ahead. And for that matter, you only lose a small overhead for using a P-frame full of I-blocks instead of an I-frame. [/B]
I have an interest in encoding video game replays, and in this specialized field, non-IDR I-frames are probably useful quite often. Very large flashing objects or sections sometimes occur, and the flashing alone can trigger scene changes. Much more commonly (but also less importantly), meters, counters, and icons usually persist unchanged across genuine scene changes.

[akupenguin]

That gives me an idea... whereby the first pass uses no IDR frames, but records how many times each frame uses each reference. Then the second pass can decide which I-frames should really be IDR (most of them, so you don't sacrifice seeking) but lose no efficiency on snacky's flashes (or even force them to be P, if necessary).

[yaz]

Originally posted by akupenguin
That gives me an idea... whereby the first pass uses no IDR frames, but records how many times each frame uses each reference. Then the second pass can decide which I-frames should really be IDR (most of them, so you don't sacrifice seeking) but lose no efficiency on snacky's flashes (or even force them to be P, if necessary). yep, a kinda auto-idr-ing would be quite beneficial. just like vhq in xvid :-) for simpletons like me it's pretty hard to set a reasonable idr. anyway i doubt that there'd be any good way of it other than analysing the 1st pass.
the bests
y

ps can anyone outline me how's this x264 development going on now? officially x264 is a part of the vlc project but vlc itself provides the less encoding options (at least in the wizzard) ffvfw offers much more but it has serious implementation problems (say, the gui does not accept a/o sets options properly) mencoder gives the more options (including some very exotic stuffs like 3pass encoding and such) but some of them have been never mentioned anywhere else. so, how's it all going ?

[skynetman]

I have got BIG problems with ffdshow-20041003.
Some movies have a strange "colorizing" effect that changes image color balance from standard, to red, to green, to blue without any reason an repeatedly within few seconds .
I just launched gspot for some testing.
Current problematic AVI is XVID with packed bitstream.
Same problem with simple MMX or XVID IDCT , with both postptrocessing and picture properties on and off (obvioulsy colorizing option is always set to zero).
I'll check for other avi with same problems to see if it is xvid related.

[virus]

@celtic_druid

here are the PSNR scores for the 3 builds I've tested. Same material, same encoding params, decoded once from the start so to avoid any problems with non-IDR frames.

build 1: 43.3490 dB
build 2: 43.3490 dB
build 7: 42.0704 dB

So, first 2 are OK, while build 7 has problems... no, not big blocks, but a generalized poor quality on some sections (with PSNR falling below 38 dB!). So maybe it's gcc with ASM the culprit then?

[netchris]

:D :D :D :D :D :D :D :D
OK virus I found whats wrong.
The problem is with the -O3 option.
You have to use -O instead and you get a dll than encodes fast and without the white blocks bug.
If you use -O2 the white blocks reappear.
If you dont use the -Ox at all the encodes are white blocks-free but they are as slow as builds 1,2,7.
All the other options (-funroll-loops, -finline-functions -mtune -march etc) have no relation to the white blocks

So just use -O :cool:

(The compiles were made with mingw-msys)

[akupenguin]

A nice constrast to libavcodec, which fails to compile with anything less than -O2. :D

You might also try running the 'checkasm' tool in x264 svn. It was presumably designed for detecting this sort of situation?

[virus]

Don't get me wrong, but I still have some doubts. I'm really just presenting my opinion, and I have no real proof that things are as I think they can be.

If there's something that I've learned in the last 7 years using gcc, is that gcc almost never fails when doing basic optimization of some C code. To be more precise: I've never seen -O2 produce b0rked code on my programs with -O1 solving the problem. And when -O2 failed, it was because of my errors... you know, sometimes bugs are not exposed until the code gets really and deeply optimized, instructions reordered, redundant expressions eliminated, and so on. Again, I might be wrong, but I'm still not 100% sure that the error actually comes from gcc itself.

just my 2 cents

[netchris]

Since virus you have the experience please try to make your own builds with -O and the other options, and prove me right or wrong.
I would like to give you my build, so you can test it but it seems that attachements here are not working, and I dont have any free web space to upload it.
I didn't change anything in the code, I took what was given by the ffdshow and x264 team and played with it. It was trial and error.
The point is that in the current state even if there is something wrong with the code, there is a workaround to achieve a fast build without the white blocks bug.

[celtic_druid]

8th build up, with -O

[virus]

@celtic_druid:
build 8 works OK, thank you. I've crosschecked build 7 and 8 to see what was wrong. It turned out that the "broken" build 7 encode uses higher QPs on some scenes (example: 21-22 instead of 20). The picture looks OK, just with a slight detail drop. Don't ask me why this happens.

@netchris:
uh! oh! don't be so upset :)
I do appreciate that you found a workaround. And I'm not trying to "prove you wrong"... I'm just outlining a possible source of the problem, maybe unlikely, but still possible. And no, whatever the error is, you're not the culprit ;)


EDIT: btw, Milan has already added the GUI options I requested yesterday... what a speedster :)

[netchris]

@celtic_druid
Has the 8th build been removed? I dont see it.

@virus
Sorry I sounded like this, I was a little drunk yesterday so I was maybe a little harsh,but i didnt feal offended or upset.



I have a lot of respect for you virus as well as for celtic_druid and akupenguin (and all the other members of the community).
:thanks: for your time, efforts and explenations!

[celtic_druid]

There never was a link for me to remove.

I didn't bother posting a direct link to 8... kinda figured that people could just swap an 8 with one of the other links.

[Kurtnoise]

ffdshow_20041012 is out on SF (http://sourceforge.net/projects/ffdshow) !!! Available in 3 versions (Normal, SSE and SSE2).

Changes:
MS ADPCM support
two preferred vobsubs languages
configurable swscaler gaussian blur strength for vobsubs
support bmp and gif as bitmap overlay image
option to use WAVEFORMATEX instead of WAVEFORMATEXTENSIBLE on output if no custom channelmapping
bug fix: don't always recompute letterboxing size and don't always check for subtitles file
more blending modes
IDR interval and deblock parameters for x264
working on image overlay
vobsubs scaling
proper alignment in ICL+GCC libavcodec build
key shortcut for subtitles language cycling
possibilty to override vobsubs position
dwstring substr and c_str related fix
recognize cz for Czech in idx, don't show duplicate languages in vobsub config
don't set merit to "not used" if it previously had different value than one of defined in ffdshow
yet another fix in corrections
another subtitles correction fix
fix in capital letters correction
more keys working
keyboard shortcuts and remote messages for cycling presets
orthography correction from subrip
separate page for text subtitles options, more complete subtitles correction from Subrip
adding text subtitles correction from subrip
support for 8-bit samples on input
fixed another crash in vobsub
fixed crash in getNext with pictFull==true and pictHalf==true
enabled trellis quantization for H.263(+)



Notes:
This build should fix Windows XP explorer crashes.

Three editions are available: special for SSE and SSE2 capable CPUs and generic for all others.
SSE a SSE2 build were compiled using Intel C++ Compiler 8.1, generic build was compiled with Microsoft VC6.

Known bug: in SSE and SSE2 builds Avisynth image processing filter in ffdshow isn't working.

Subtitles correction may not work correctly in all cases (but should not crash).
These feature was ported from SubRip and if SubRip will give different resulf for a given subtitles, please send them to me.

[Inc]

First I do apologize in andvance if some questions now of mine have been already treated in here BUT the browsing/searching through this forum actually is a pain via t-online provider in germany as it seems.

These following questions do refer to the last ffdshow version BEFORE this actual release of 2004/10/12! Im at work on a MAC and just found this new release of 20041012 in here, so I cant test it right now.

1. A Makeavis generated "AVIS" AVI wont be decoded in an fvw decoding environment like Vdub or CCE, even if it has benn set in the vfw configuration of ffdshow (decode Avisynth AVIS). In an Dshow decoding environment its no problem.

2. Also according to vfw decoding environments: If in ffdshows vfw decode setup a decoding of for instance RAW YV12 decoding has been setup, Vdub still uses my XVID decoder in the informations Box of Vdub when decoding an avsiynth .avs with an outgoing YV12 colorspace (same thing in case of other outgoing colorspaces as FFdshow/vfw wont be used in Vdubor CCE).
If for instance capturing an AVI by using "FFDS" FourCC in ffdshow via VirtualVCR .... VirtualDub does decode that one via FFdshow/vfw! BUT no filtering within ffdshow's vfw setup will be performed. I have to do a Graphedit .grf workout via Directshowsource() to get filtering via ffdshow enbaled when opening an avs in vdub.

3. When trying to use Avisynth filtering within! ffdshows filter section, it doesnt work anymore. An access violation error is reported by avisynth WITHIN ffdshows decoding when doing the preview of that via ffdshow decoded AVI file.




I do love ffdshow! But to me it seems beside some very nice additions in new releases, there are some risks that things of the past wont work anymore as for instance Forcc "AVIS" decoding via Vdub or CCE never was a Problem when using the old FFvfw from autum of 2003 for instance.

Thanks a lot in advance!!
Inc.

[Leak]

Originally posted by incredible
3. When trying to use Avisynth filtering within! ffdshows filter section, it doesnt work anymore. An access violation error is reported by avisynth WITHIN ffdshows decoding when doing the preview of that via ffdshow decoded AVI file.

This seems to be a problem caused by compiling ffdshow with Intel's C compiler; it should work in the latest ffdshow version without SSE/SSE2 optimizations - see here (http://ffdshow.sourceforge.net/tikiwiki/tiki-read_article.php?articleId=12) and here (http://sourceforge.net/tracker/index.php?func=detail&aid=1040852&group_id=53761&atid=471489)...

np: Vladislav Delay - Anima E (Nohuume)

[pogo stick]

Is it just me or new ffdshow is not appearing in VDub's encoders list? :confused:

[netchris]

I have the same problem with the latest build, so I after i istalled it, I copied the ffdshow's files to a temp folder, then uninstalled it.
I installed an older build that doesnt have this problem, and copied the files from the temp folder to the ffdshow folder, to replace the old ones with the new.

[stephanV]

doesnt show up here either

[celtic_druid]

Go into your registry and change the vidc.ffds entry to progra~1 instead of "program files". Should sort it.

[stephanV]

yep, thats it... its weird though

[Yong]

Can i post the ffdshow audio decoder bug report in here?

[yaz]

i got the same problem(s!) as incredible.

no way of using 'makeavi' anymore :-((( loading the fake avi into mencoder makes a crash (windows say msvcrt.dll is crashing???) is there anyone succeeding in this business? pls, drop a workaround, if u're aware of any.

avisynthing in ffdshow still seems to crash (or i do in a wrong way?)

thx
y

[Leak]

Originally posted by yaz
avisynthing in ffdshow still seems to crash (or i do in a wrong way?)

Works for me - what exactly are you doing?

Also, did you make sure to install the unoptimized version? The SSE/SSE2 versions will still crash.

np: Vladislav Delay - Huone (Nohuume)

[yaz]

Originally posted by Leak
... The SSE/SSE2 versions will still crash ... well. i was playing with the sse ver. i give a try to the non-opt ver. thx for the answ.
thx
y

[Inc]

Originally posted by yaz
i got the same problem(s!) as incredible.

no way of using 'makeavi' anymore :-((( loading the fake avi into mencoder makes a crash (windows say msvcrt.dll is crashing???) is there anyone succeeding in this business? pls, drop a workaround, if u're aware of any.

avisynthing in ffdshow still seems to crash (or i do in a wrong way?)

thx
y

a) For using avisynth scripting "within" ffdshow it seems that you have to install the generic build - I hadnt time to install the newest for testing till now.

b) Mencoder worked without problems in the past when using ffdshows! Makeavis (I use it via commandline to just go in using an avs into the batch file). I just copied the "old" ffvfw.dll & mplayerlib.dll into the same folder as mencoder.exe and added the description in the codecs.conf file thats it. But as I do remember now ... wasnt it you who gave me that advice? So did something change also related to "this" in the very recent release of ffdshow??? :eek:

[yaz]

Originally posted by incredible
... wasnt it you who gave me that advice? So did something change also related to "this" in the very recent release of ffdshow??? :eek: yep, it was me ... but yesterday evening it all drove me mad. whatever i tried all i got was a crash in msvcrt.dll. and no reading, of course :-( i'll give it another go this evening. maybe, a 'clean-the-house-before' will help.
i'm just annoyed, becuse the log says 'makeavis support is back'. but for me 'back' means 'away' :-(

the bests
y

[Audionut]

The latest SEE2 version works fine for me.

Just had to run "regsvr32 c:\**\ffdshow.ax" from the command line.

It appears that it does not register itself.:confused:

http://sourceforge.net/tracker/index.php?func=detail&aid=1046034&group_id=53761&atid=471489

[Yong]

Here's my ffdshow audio decoder settings,
Format Decoder
MP2 libmad
MP3 mp3lib
AAC libfaad2
Vorbis themor
MSADPCM Libavcodec
IMA ADPCM Libavcodec
FLAC can't work.

Allowed sample format for sound processing:
16 bit integer

Support output sample format:
32 bit float
Floating to integer conversion:
Dithering(Selected), Noise shaping: heavy

When playing AAC ,MP2 ,WMA, and Vorbis with only the 24 bit output sample format selected, the audio will become full of distrosion.
AAC only have problem with 24 bit out.
While MP3 only can work with 16/32 bit float output, when select 24 bit the sound will lowered and full of distrosion, select 32 bit integer out will crash.
Vorbis only can play with the 32bit integer /float ouput but have some distrosion, select 16 bit out will crash.
playing MP2 with 16 bit out will crash.
IMA / MS ADPCM crash with 32 bit integer out.
WMA is work fine but can't properly decode some WMV audio.
Only the 32 bit float is safe to use.

ffdshow video decoder bug:
Playing mpg with ffdshow now can work, but still can't play some video-stream-only mpg file.

Sorry for my english...:stupid:

[Mug Funky]

w00t! the vobsub functionality is really good.

i'm pushing my machine to it's limit, and it seemed vobsubbing through ffdshow and disabling Dvobsub was just enough to decode in real time.

how's the development of the loading of embedded subs? i'm experimenting with muxing the .idx into a .mkv file, but nothing except media player classic can find these subs, and it renders them too slow.

i'm playing back xvid @ 50fps, 512x384... just to see if i can. i was so sick of field-blends and what to do with them that i just up and smartbobbed the whole thing and encoded it without touching the blends.. . this presents the problem of decoding speed, of course, which is why i'd like ffdshow to do as much of the processing as it can.

i suppose for now i can just leave the .idx and .sub in the same directory, but i find whole files are tidier.

oh, also, yellow subs seem to come out grey...?

[APF_Gandalf]

Originally posted by Audionut
The latest SEE2 version works fine for me.

Just had to run "regsvr32 c:\**\ffdshow.ax" from the command line.

It appears that it does not register itself.:confused:

http://sourceforge.net/tracker/index.php?func=detail&aid=1046034&group_id=53761&atid=471489

it seems to me that it's the "old" folder name bug that is back. if you install it in a folder without spaces, it works: "c:\ffdshow" works, "c:\program files\ffdshow" fails even if I try to register it myself using "c:\program files\ffdshow\ffdshow.ax" or "c:\progra~1\ffdshow\ffdshow.ax"
I submitted a way to avoid this bug long ago when it first appeared. it worked, but it seems that it was forgotten in the latest official builds.

[arman68]

Originally posted by Mug Funky
w00t! the vobsub functionality is really good.

i'm pushing my machine to it's limit, and it seemed vobsubbing through ffdshow and disabling Dvobsub was just enough to decode in real time.

Yes, it is coming along very nicely. I did some tests on a P4 2.2Ghz, and VSFilter adds an extra 30% cpu usage, as opposed to VobSub in ffdshow. And that's using swgaussian blur, which is very nice. For the first time I can now use the post-processing options I wanted :D

Seems that the new keyboard shortcut for subtitle cycling are broken, and broke the toggle as well though.

[kittychan]

yooooo

I am quite happy ffshow can replace the vsfilter.dll and decrease the cpu load ^_- , but I can't play subtitle embedded in .ogm ot .mkv...
I did tick the "accecpt embedded sutitles" box, but no subs appears, so I still have to use vsfilter for .ogm (and .mkv)

Did I do something wrong?

++

kittychan

EDIT : OK my wrong....
I have to set the VRM7 for the video direct show in MPC option...
sorry for having ask ^^

++

2nd EDIT....
re-sorry, when set the VRM7 for the video direct show in MPC option, sub are played by MPC but still not by ffdshow...
I still can not make ffdshow play ogm sub.

++

[Mug Funky]

there's a little tooltip that says "very incomplete and experimental. Send me samples which don't work".

i'm guessing it'll only do textsub? not sure, but i'm certain it'll be completed soon.

in the meantime you'll have to live with cluttered directories full of .idx and .sub files :)

[The Shemeta]

hi,

i upgraded to the latest ffdshow-20041012 & now i don't have the vfw when i open a movie in VirtualDubMod.

anyone with the same problem?

the good thing is that if fixed the explorer.exe crash.

[esby]

@The Shemeta:

you could READ this thread a bit...

Now take this answer, go back 4-5 replies ago, and READ!

esby

[therealjoeblow]

The .idx file of a VobSub set has numerous user-definable settings that can be changed to suit, including custom colors:

custom colors: OFF, tridx: 1000, colors: D2D2D2, 000000, 000000, 000000

Would it be possible to add support for overriding this setting when ffdshow reads the file (colors would do fine for now, and maybe in the future, all of the special settings when you have time)?

If it's a big deal to add this functionality, then don't worry about it, but I just thought the ffdshow configuration interface could have a couple of more tabs where a user could enter their preferred color (and other) settings, and then ffdshow would use those if they were entered, instead of the ones contained in the .idx file.

Many thanks,

[therealjoeblow]

One more note on the vobsub colors:

There appears to be a small glitch in basic interpretation and display of colors. The following line shows correctly with directvobsub as white text outlined in black for one particular set of subs:

custom colors: ON, tridx: 1000, colors: 000000, 000000, 000000, D2D2D2

However, with ffdshow, it shows the text in all black with black outline, impossible to read. If I change the line as follows, then the subs appear correctly with white text outlined in black:

custom colors: ON, tridx: 1000, colors: D2D2D2, 000000, 000000, 000000

I'm not sure exactly which one is correct (directvobsub or ffdshow), but since vobsub was around much longer, I'm assuming it's interpretation would govern, having been burned to thousands of cd's around the world already.

[LigH]

If that was about the order of the masking (transparency) bits, it would have been at least understandable:

Either "tridx: 3210" (bit-position order) or "tridx: 0123" (ascending order) would look possible to me; I'd prefer "ascending" order.

But the colors... shall always be connected in ascending order, shouldn't they?

colors: (col. for pal.-index 0) (col. for pal.-index 1) (col. for pal.-index 2) (col. for pal.-index 3)

I would be very confused if it wasn't so! :scared:

So the second line shall be nonsense, because color 0 is transparent, so it shouldn't matter which color it got - the fact that it is white, obviously shows a shift or mirror in the palette index interpretation.

[Yong]

ffdshow version Oct 12 2004 08:06:04 SSE2
Bug report:

My computer spec,
Pentium 4 2.4GHz
512 MD DDR 266
Windows XP SP1
DirectX 9.0c
ATI RADEON 9600XT

Software:
Media Player Classic 6.4.8.2
Matroska container,
Using Haali Matroska Splitter Sep 8 2004 14.06.02

ffdshow configuration:
All video decoder using libavcodec,
except MPEG1 and MPEG in AVI using libmpeg2.
Only the postprocessing filter is used.
Other setting use default setting.

ffdshow postprocessng configuration:
[* = Selected option]
*process whole image
*only right half
*Postprocessing

*Custom
--*Dering, Luminance
--*Deblock H, Luminance
--*Deblock V, Luminanve
Processing Strength: 512

Processing method:
--*mplayer
--*Accurate deblocking
Level fix
--*luminance
--*Full luma range

--*Nic's Nic's First
--* X threshold: 20
--* Y threshold: 40

I can start playing video normally sometimes when using setting above, But i drag new video to the player, it crashed.
If the video can start normally, it will crashed if i close the video.
Sometimes can't start playing at all, crashed.
If i unselect the "Dering, Deblock H, Deblock V(All Luminance)" option, crashed won't happen again.
This three settings crashed with "only right half" option selected.

select "Dering , Deblock H, Deblock V(all Chroma)" don't have this problem.

[BoNz1]

Encoding video with x264 with ffdshow video codec in virtualdubmod crashes here. Anyone else with the same problem?

[LigH]

No crash with x264; but some strange over-softening occured (I wonder if the loop filter was the reason, or enabling all subdivisions); I miss a "reset" to default options (except for re-installing).

[yaz]

Originally posted by LigH
No crash with x264; but some strange over-softening occured (I wonder if the loop filter was the reason, or enabling all subdivisions); I miss a "reset" to default options (except for re-installing). the same here. with celtic's compile it 'cuts like hell' for me.
the quality is amazing (for me) but, yes, the overall look is somehow softer than i like. (yes, i'm one of that crisp-freaks:-)
the bests
y

[akupenguin]

Originally posted by LigH
some strange over-softening occured (I wonder if the loop filter was the reason, or enabling all subdivisions)
Do you have normal ffdshow postprocessing on, in addition to the H.264 loopfilter?

[LigH]

Of course! ;) - I tried both; It looks like after a scene change, the first frames are quite crisp, then suddenly the following frames become flat. I'm not sure if this still occurs with the current version shipped with the latest ffdshow, it occured with the previous version and celtic_druid's test version 8. I'll try to check it more soon (sorry, don't have much free time)...

[celtic_druid]

As far as I can tell there haven't been any changes to src/codecs/x264 since I compiled those versions, although there have been a couple of changes to x264.

[hellfred]

VDubMod and ffdshows x264 works for me,too.
Win98
PIII
VDubMod 1.5.10.1 build2439
ffdshow 20041012 with cletic_druids ff_x264.dll #8
2pass with default settings. quant 12

Hellfred

[celtic_druid]

In case anyone is interested:
http://celticdruid.no-ip.com/test/ff_x2648-AthlonXP.7z
http://celticdruid.no-ip.com/test/ff_x2648-Pentium4.7z

Might give a bit of a speed boost over the previous pentiumpro build.

[yaz]

Originally posted by celtic_druid
In case anyone is interested ... many many thx ! just for sure; did u recompile the previous source or is it a new cvs-snapshot ?
this or that, it always occurs to me that would we ever be able to make any return for u for your contribution ? if u got a clue, just call me :-)))
thx again
y

[celtic_druid]

Well I have done several cvs updates since the last compile, however the src/codecs/x264 files have not changed so yeah it is basically a recompile.

[yaz]

Originally posted by yaz
... if u got a clue, just call me :-))) ... thx again
y Microsoft Internet Explorer Detected - Downloaded Blocked would it be that ? :-(((
y

[Yong]

I can't download...
When i click the link above,
after 2 swcond i was see "Microsoft Internet Explorer Detected - Downloaded Blocked"...
can someone plz tell me why...:confused:

EDIT: plz ignore my post, it worked:D :D :D

[LigH]

I'd guess, the MSIE detected a META-REFRESH or script initiated download, and the MSIE blocked a not-user-initiated download (for security reasons).

Your message looks like the webserver detected an access using MSIE, and the webserver blocked the download - what I would not expect at all... :D

[celtic_druid]

Originally posted by LigH
Your message looks like the webserver detected an access using MSIE, and the webserver blocked the download - what I would not expect at all... :D

That is exactly the case, however I turned off the IE block days ago.

[Yong]

Big Thanks celtic_druid for the ff_x264 dll and the mencoder!!:D
But still encounter an error when encoding video using the h264 codec... :eek:
My comp, spec. is P4 2.4GHz.
I've read the mencoder+x264 thread, i'll try the new build [mplayer20041022p4].
It's worked. :D

@LigH:
I'm using shared computer(internet cafe), so i don't know the computer admin's using what kind of secutity setting...
I'm using firefox(R) at my own computer!;)

New bug report here:
ffdshow config:
format: MPEG1/2
codec: libavcodec

ffdshow will crash when playing mpeg 1/2 video with the visualization "Motion vector" is on or selected while playing.

[Rumbah]

The ffdshow SSE2 build crashes Premiere Pro 1.5 with ffv1 (I have a P4). With the non SSE2 bild, everything works fine.

[celtic_druid]

updated x264: filesize (bits) in a 32 bit int will overflow after 250MB, screwin.
core/common.h
encoder/encoder.c

Someone want to test this (http://celticdruid.no-ip.com/test/ff_x2649.7z)
, not sure if it effects the problem with -o3 or not, so I compiled this new dll with -o3.

Doubt it fixes the -o3 so I also put up two new dll's with -o.

Links for the updated dll's with -o in case anyone was wondering:
http://celticdruid.no-ip.com/test/ff_x264-AthlonXP.7z
http://celticdruid.no-ip.com/test/ff_x264-Pentium4.7z

[netchris]

The problem with -O3 remains, the patch is irrelevant to the problem.

[celtic_druid]

Yeah, I thought as much.

[netchris]

Made some tests with this dll, and the 1 pass constant bitrate behaviour has changed. It gives a quite improved picture, but often gives oversized files. My guess is you have included all the latest patches so we have an up to date x264?
It seems the codec gives an even better picture now.
Havent made any 2 pass tests so take these with a grain of salt.

[bond]

i dunno if milan actually follows this thread (i hope so), but i wanted to mention that it would be great if ffdshow could support the small letter "s264" fourcc, which moonlight uses with their h.264 stuff, to make ffdshow able to connect to it

[gotaserena]

Perhaps it is worth mentioning that this seems to be the only way to play h.264 from .mp4s using ffdshow: since 3ivx mp4 splitter doesn't support h.264, we are forced to use moonlight's or pay for nero's.

[celtic_druid]

@bond, I added s264 myself, seems to work fine with the moonlight splitter.

ffdshow.s264.7z (http://celticdruid.no-ip.com/xvid/misc/ffdshow.s264.7z)
ffdshow.s264.7z (http://s14.yousendit.com/d.aspx?id=89A176BA463D60E2236FB90C1610B8C7)
Works fine with the version of ffdshow that you can find in the same directory, not tested with any other ones, although I guess it would work, just things like SNOW encoding might be stuffed due to an old libavcodec.dll.

[movmasty]

where do i could find a guide,or something similar, to ffdshow settings??????????

[Mug Funky]

just to check... are b-frames working in x264? i seem to remember using them in mencoder, so maybe b-frames should be un-greyed in ffdshow?

or are they currently way too unstable?

[edit]

aaah, that's right - they don't work with CABAC yet.

i suppose they can be allowed provided CABAC is unchecked, at least for now.

i'm just thinking the bitrate/quality is so good now, that b-frames would make it ven more amazing. but possibly there's inherently less to be gained from b-frames in H-264 than there is in ASP MPEG-4. i don't know enough about either to be able to tell :(

[celtic_druid]

bframes: CONF RANGE 0, 16.

Guess you have to stick with mencoder or wait for Milan to add it in ffdshow. Default for mencoder is 0 by the way, probably a good reason for it to.

[akupenguin]

libavcodec should be able to decode B-frames with CABAC now.
H.264 B-frames should help just as much as MPEG-4 ASP B-frames, or more.
But they're much more complicated than in ASP, and very few of the B modes are implemented in x264. So currently they make compression worse.

[celtic_druid]

Did a build with bframes enabled, seems to work. Playback is a different story. Just jitters.

[wata]

how about multi-language support for text subs, like vobsub

eg
moive.english.srt
movie.german.srt
movie.chinese.srt

etc

i know you can select files from submenu "Subtitle files", but if i put another location that contain alots of subs, it show every of those subs as well regardless if the filename is different or not

thanks

[Amour]

How do I know if my computer supports SSE or SSE2 ?
I think I got a Pentium 4 from Intel.

[LigH]

How about CPUInfo (Freeware)?

http://www.pcanalyser.de/download/cpuinfofree212.zip
__

BTW: You don't know your PC? So why did you buy it? :sly: ;)

[Leak]

Originally posted by Amour
How do I know if my computer supports SSE or SSE2 ?
I think I got a Pentium 4 from Intel.

I think you got both SSE and SSE2, then - since every P4 has them. :)

[Dark_Angel_PT]

When will we see a stable version of fddshow?

Its nice to add new stuff, but there's always bugs and problems being left behind.

I stopped using ffdshow because of this.
Its an incredibly powerfull software without a goal. There isn't like a "1.0 version" to achieve.
It's like it's trapped in an ever-ending beta stage, with no final version in the future.

Why isn't ffdshow frozen at current state, developers get a few goals planned, and set as a first objective to get all the bugs solved before any new stuff is added.

Probably the problem I had in the past is solved, but I bet that new ones will show up.

In its current state, ffdshow is capable of replacing every "regular" codec out there for the "average Joe".
With a full subtitle suppport, and all the filters and features it as, getting this to a 1.0 Final, would make a lot of people happy outthere.

All those crappy codec packs would come to a short end in no time.

But this is just my opinion...

[Sharktooth]

FFDShow is still in ALPHA!
if you want a stable version download the 2002 version...

[Defiler]

Originally posted by Sharktooth
FFDShow is still in ALPHA!
if you want a stable version download the 2002 version... The fact that the last stable branch came out 2.5 years ago is, of course, exactly the issue that Dark Angel is getting at. If someone could work up a time estimate for a new stable fork, I would be willing to contribute funds toward developer time.

[Yong]

ffdshow version NOV 1 2004 10,28,49
ffdshow audio decoder crash with following settings,
codec: mp2, vorbis
decoder: libmad, thermor
audio filter: libavcodec resampler filter.

I found a new audio codec ffmpeg sonic lossy/lossless in ffmpeg encoder,
but the new audio codec only can decode by recent build mplayer (bulid by celtic_druid), i hope the new release ffdshow can decode the ffmpeg sonic audio...

[Dark_Angel_PT]

Originally posted by Defiler
The fact that the last stable branch came out 2.5 years ago is, of course, exactly the issue that Dark Angel is getting at.

Does this answear your question, Sharktooth?

[vinouz]

In ffdshow encoder, in the "override framerate" setting in the "output" section, the value is an unchangeable 25fps.
It wrecks the constant bitrate 1 pass estimation with different framerates.
More than this, trying to encode a video for a nokia 3g phone using H.263, the raw h.263 stream contains information that makes it override the 15fps setting both in the original file I compressed and in the .mp4 information (using -rate=15 un mp4creator) and play at 25fps.

In an AVI container, the framerate settings of the AVI get the priorities so it doesn't appear (the file plays at 15fps)...

Hope it will be fixed soon. I'm confused such a big and simple bug hasn't been found and corrected earlier....

And, thanks to the developper for such a great piece of code !

[yaz]

@celtic_druid
in your compile (041108 hasn't been tested) the deinterlacer bug seems to be reinvented. any change in the pane (even wout pushing accept/ok!) breaks 'ffdshow.ax,configure'. hacking the registry (the good old workaround:-) seems to work. thx for the compiles, anyway.
y

[Sharktooth]

Originally posted by Dark_Angel_PT
Does this answear your question, Sharktooth?
Yes sure, but maybe milan has some goals to reach before thinking to a stable build...

[Dark_Angel_PT]

Originally posted by Sharktooth
Yes sure, but maybe milan has some goals to reach before thinking to a stable build...

Yes, but thats's what I mean.
To organize things a little bit, set some goals...something!

[celtic_druid]

@yaz, I think that could be another compiler issue. Unfortunatly I can't get it to compile with gcc/mingw.

[yaz]

@celtic_druid
sorry to say, 041108 goes the same way :-(
btw, should i change ff_x264.dll to yours or is that included ?
thx
y

[celtic_druid]

The ff_x264.dll that is included is a generic version. If you happen to have a P4, AthlonXP or I guess AMD 64 CPU then there might be some kind of speed gain from the XP/P4 versions.

Still used ICL for the 041108 version which is I think where the problem comes in, although using MSVC doesn't make any difference, it does however for the problem with libavcodec's resampler. If I could get it to compile with gcc/mingw then I suspect that the problem would disappear.

[mav_top]

i don't know if this is the right post to ask features in ffdshow, so if i'm wrong i'm sorry about it

anyway i would like to have the OSD with a timer, just to see it for
"n" seconds and then it would be disabled (the OSD) automatically, and also if possible know which filter i'm using actually

i ask for this, because sometimes i don't know if ffdshow is running and which settings it is using, so i need to go in the filter proprierties page and look,

if i could in the osd show all these stuffs for specified seconds it would be better

thanks in advance
mav_top

[pogo stick]

Is ffavisynth working already?
Here is what ffdshow's help says:

ffavisynth

ffavisynth - Avisynth filter to directly use ffdshow image processing filters from Avisynth scripts

Syntax

ffdshow(string "preset", string "options")

preset - existing ffdshow preset to be used
options - array of "name=value" pairs separated by commas

Both parameters are optional. If preset is not specified, a new preset called ffavisynth is created temporarly. Options override preset settings. List of allowed options names and values is to be written, for now look at registry key HKEY_CURRENT_USER\Software\GNU\ffdshow\default to get the idea.

Limitation

Input and output colorspaces are equal. Even if ffdshow image processing filter chain would produce image with different colorspace, it will be converted to match that on input.
I tried to follow instructions, but it doesn't seem to work or I am missing something.
Any help, please. Thanks in advance.

[Socio]

Originally posted by pogo stick
Is ffavisynth working already?
Here is what ffdshow's help says:


I could only get Avisynth to work in the latest non-SSE SSE2 version if you try it under the SSE or SSE2 version you get a script error.

[pogo stick]

I don't get any errors. It just don't make a difference in video. Can you post examples of options that work for you?

[Leak]

Originally posted by Socio
I could only get Avisynth to work in the latest non-SSE SSE2 version if you try it under the SSE or SSE2 version you get a script error.

Well, the SSE/SSE2 versions only have a problem if you use AviSynth from *WITHIN* ffdshow while playing. The functionality this thread is about is using ffdshow as a filter in AviSynth, so I don't think the same bug will byte, errr, bite here...

[Socio]

Originally posted by Leak
Well, the SSE/SSE2 versions only have a problem if you use AviSynth from *WITHIN* ffdshow while playing. The functionality this thread is about is using ffdshow as a filter in AviSynth, so I don't think the same bug will byte, errr, bite here...

Yes I was thinking from with inside as that is how I use it and Pogo was refering to from with outside my error.

[Flash_Git]

I was wondering if it was possible to have multiple instances of ffdshow. For some files I'd find it useful to have two instances of logoaway and some of the time two sets of subtitles would be useful although I could manage that with VSFilter.

[Leak]

Originally posted by Flash_Git
I was wondering if it was possible to have multiple instances of ffdshow. For some files I'd find it useful to have two instances of logoaway and some of the time two sets of subtitles would be useful although I could manage that with VSFilter.

I'd say being able to use the same ffdshow internal filter several times in one instance would be even better - you can reorder the filters already, so it might be possible to invoke them more than once with the current architecture. I'm not Milan, though... :)

np: Triola - AG Penthouse (2. Epoche) (Im Fünftonraum)

[swalker]

Originally posted by celtic_druid
If I could get it to compile with gcc/mingw then I suspect that the problem would disappear.

What gcc version are trying to compile with? gcc3.4.2 builds the mplayer, ffmpeg, libmpeg2, mpeg2enc, and x264 code although I'm currently building the main ffdshow project with Visual C++.

[pogo stick]

Originally posted by pogo stick
I tried to follow instructions, but it doesn't seem to work or I am missing something.
It was my mistake with options. :o
Seems to work fine! And since this filter is for preprocessing and not for postprocessing I thought it deserves it's own thread in Avisynth section. Here (http://forum.doom9.org/showthread.php?threadid=85447) it is. :)

[Yong]

@mav_top:

May be you should PM or mail to milan about the feature request, because he say he lazy to read this long thread!:eek:

[therealjoeblow]

Originally posted by wata
how about multi-language support for text subs, like vobsub

eg
moive.english.srt
movie.german.srt
movie.chinese.srt

etc

i know you can select files from submenu "Subtitle files", but if i put another location that contain alots of subs, it show every of those subs as well regardless if the filename is different or not

thanks

Or carrying that further -

1) support for multiple subs based on a standard naming convention (as noted above, could be language, or close-captioning as in "filename.cc.sub.srt", or the unicode version of the close-captining "filemame.unicode.cc.sub.srt"

-basically, the filter should look for the 'filename' portion up to the first '.' to match the movie name; then differentiate the various versions by the text up to the next '.'; then discard the rest; and finally, recognize the subs by the file extention after the last '.' - this appears to be the way that vsfilter works.

and

2) UNICODE support would be great - normal text subs can't display the special characters like musical notes that wrap around lycics in text subs - Unicode can. Vsfilter supports these, current version of ffdshow doesn't appear to.

also,

several replies ago I suggested the color interpretation of VobSubs wasn't handled correctly, it would be great if ffdshow could have settings for the 4 vobsub color settings to override to a user's preference anyways.

[Mug Funky]

um... i'm thinking this is just me here, but are vobsubs broken in the latest few compiles? or did i do something to break them?

ffdshow is the only thing that will be able to sub my movies and play them back at sufficient speed, so i'm a little eager to find out how to get them working again.

tried complete uninstall + reinstall, but they don't seem to come back.

i'm using fast bilinear if that makes a difference. haven't played enough to see if any other types make them come back :(

[edit]

it appears to load them but not display them. i can choose "language" in the vobsub config and everything.

maybe it's doing something odd like rendering them offscreen or something? thing is i don't think i installed a new version before the subs broke.

[hellfred]

Unvisible subtitles? That reminds me that in MediPlayerClassic one has to set a special DirectShow video output to get them displayer, I think it was DirectShow VMR9 renderless. Get yourself a copy of MPC and play with the settings:
View->Options->Playback->output.
I have looked it up, it is DirectShow VMR9 renderless. See footnote (**).

Hellfred

[Vitos]

Originally posted by hellfred
Unvisible subtitles? That reminds me that in MediPlayerClassic one has to set a special DirectShow video output to get them displayer, I think it was DirectShow VMR9 renderless. Get yourself a copy of MPC and play with the settings:
View->Options->Playback->output.
I have looked it up, it is DirectShow VMR9 renderless. See footnote (**).


Rendering subtitles in ffdshow is completely independent from output you choose in your video player. VMR9 is needed in MPC if you use its built-in subtitling...

[Mug Funky]

aha... and MPC's subtitling (though pretty) isn't anywhere near as fast as i need to avoid framedrops and frozen-till-next-keyframe issues. my machine is old.

i'm on VMR9 renderless, i've tries windowed as well, plus overlay, etc. (no VMR7 because i refuse to install winXP - my machine's in a happy place, and that kind of disruption would prove my undoing).

[edit]

now this has worked before, so i don't know what's happened.

i might go and use the last binary from sourceforge. i'm encoding now, so i'll do this once it's finished.

[Didée]

Originally posted by Mug Funky

... my machine is old.
i'm on VMR9 renderless ...
If you are short with CPU horsepower, why do you let VMR9 burden the CPU with all that drawing stuff, while the g-card could provide a HW overlay, but instead is cleaning its fingernails only?

[Mug Funky]

hmm... it didn't make a difference, actually. i think VMR is using 3d stuff from the card.

the difference to me was that (a) nvidia's overlay messes up chroma sampling, and (b) MPC gave me a decoding framerate under VMR9, whereas in overlay it always reported 0 fps.

as far as performance goes, i saw no difference - if a scene froze in VMR9, it would freeze in overlay too.

btw, after regressing back to ffdshow-20041012-sse.exe, subs came back.

has anyone else noticed this? i'm the only person reporting this, so i guess it's just an issue with my box.

it's a p3 733, 384 RAM (pc133 i think, probably not brand-name), with asus nvidia geforce DDR 32 (that's a geforce 1...) running latest drivers.

[edit]

in win2k...

[edit 2]

my monitor is fine, thankyou very much :) hitachi cm715. 19 inches of CRT goodness. degauss makes me feel like i'm at a rave :o

[Didée]

Originally posted by Mug Funky
it's a p3 733, 384 RAM (pc133 i think, probably not brand-name), with asus nvidia geforce DDR 32 (that's a geforce 1...) running latest drivers.
[edit] in win2k... Damnit! That's exactly my office PC at work! Except for a nvidia TNT2 with outdated drivers, and those 384 MB are built by three different no-name brands ;)

And the monitor is a bad joke.

[Mug Funky]

so my question to you:

do vobsubs work on that machine? :)

i love this computer. it's very loyal :) it's extremely stable, doesn't overheat (well, it went ga-ga once on a ~44 degree day when i was doing an xvid encode. no air-con here, and a tiny room).

i think if i were to put winXP on this machine the whole fragile house of cards i've constructed of it over the years would come crashing down. it runs slackware 'nix distros well, and i'm just working up the guts to mess the partitions up and put a real OS on this thing (linux :)). lots of backing-up to do first.

[Didée]

>> do vobsubs work on that machine?

Mind you, I don't know. My needs in respect to subtitles are rather low in general. In office, they are _zero_ ;)

[Mug Funky]

what? you mean you don't watch anime at work? hehe :)

[hellfred]

Originally posted by Mug Funky

it's a p3 733, 384 RAM (pc133 i think, probably not brand-name), with asus nvidia geforce DDR 32 (that's a geforce 1...) running latest drivers.
There is something seriously messed up in your OS, or the clip you want to play back is much harder to decode than anything i threw on my system up to now.
I have a P3-550MHz (PC100 chipset) running Win98 and WinXP, just installed. I only have one file with a VOBsub subtitle (MPEG4 with resolution of about 640x480, I am too lazy to search the file) and I clould play it on win32 using mplayer with only very few frames dropped. Booting to linux helped me to get a perfect smooth playback. I have never tryed the file with any dshow filters, though. Get yourself a movix boot cd and try playing the file with it to see what you can get from your hardware.
Hellfred

EDIT: I have digged up the file: it is MPEG4 603x306, I have a nVidia TNT 1 graphic adapter, 312MB ram, and i use DirectVOBSub and ffdshow v20041012 in BSPlayer v1.0 in this very moment to display the file. System reacts somewhat slow, but ffdshow OSD tells me that 100% CPU is hardly ever needed to decode the file. I will try using ffdshow to display the *.sub and *.idx file next.

EDIT2: On Win98 i did not get ffdshow to display the subtitles while DirectVobSub is installed, and as I do not want to touch a running system, i have booted into WinXP where there is no DirectVobSub installed. But there even without displaying any subtitle, decoding the very same clip rises the CPU load to 100% :(
I have a mplayer win32 compile on the system, and when using it to decode the clip and displaying the subtitle, i get a CPU load of 60 to 80% for most of the frames. You can get the a precompiled binary for win32 from the download section of mplayers homepage (http://mplayerhq.hu). Just unpack the archive and drag and drop the avi onto the mplayer.exe. When the subtitles are named after the main file and are lying next to the main file, they get loaded automatically. Rotate through the different subtitles by pressing 'j'.


Hellfred

[Mug Funky]

hhe.. my system is messed up, but it's not particularly slow considering it's age. it outperforms a lot of machines built this year, simply because i removed all the bloat from win2k that i could find. of course, on pure CPU stuff it is nowhere near as fast as todays stuff. but running programs and what have you, it runs awesomely.

the clip i'm decoding originated from a PAL field-blend from hybrid NTSC, and i couldn't be bothered with restore24 because it made 30p pans jerky. so i smart-bobbed and encoded at 50p, 512x384.

as you can probably guess from that, it struggles to play back in realtime (though it does :)), but when i was using vobsub, it was just too much for the poor thing. ffdshow didn't have this problem.

mplayer is a fine player, and i use it quite a bit - for one thing, "mplayer -fs *.*" is more useful than you'd think :)

unfortunately it's subtitle rendering seems to be b0rk on the compile i have (i'll try get a new one asap) - chroma isn't rendered, it is simply copied from the frame underneath. this is watchable a lot of the time, but when there's strong colour present, the subs look fugly like you wouldn't believe.

mplayer does indeed decode these files with speed to spare.

next time i'll encode interlaced, but i don't like doing this because then i need to bob on playback, and not everything does this all that well - my viddy card can do a dumb bob, but only when asked.

[edit]

holy crap! mplayer plays these files flawlessly at 50% CPU usage! that's amazing. i'm totally amazement :)

[edit 2]

decoding to -vo null gets me 30% CPU usage... but it's no fun unless you can watch it :)

[akupenguin]

Originally posted by Mug Funky
unfortunately its subtitle rendering seems to be b0rk on the compile i have (i'll try get a new one asap) - chroma isn't rendered, it is simply copied from the frame underneath. this is watchable a lot of the time, but when there's strong colour present, the subs look fugly like you wouldn't believe.
Not a bug, just a design limitation. It's being worked on, along with the SSA renderer. But I have no estimate for when the new version will be usable.
(BTW, I happen to like the semi-transparent look, though I agree it's ugly when applied to a really strong chroma.)

[BlindWanderer]

I have a file with muxed subs that i'm not exactly sure how they were prepared but...

ffdshow when rendering the subs displays a box character at the end of the each line (mpc doesn't do this). I suspect it's a line feed but i can't be sure. ( 0x0A or 0xOD); consequently could we have added a subtitle spy (text field that displays the last text to be rendered) and a character filter as to remove characters entered.

[eXistenZ_69]

Does anyone know how I can set ffdshow to output raw AC3 via s/pdif? It appears to be outputting padded AC3...

See the following thread for my problem description:
http://forum.doom9.org/showthread.php?s=&threadid=85535


Regards,

eXistenZ_69

[Mug Funky]

akupenguin: thanks for the update :) mplayer is cool, and i follow it's development avidly.

[pogo stick]

Good news, everyone! :)
Milan added deinterlacing part of Minolta's patch for XviD to FFDShow! So now when "Overlay mixer" checked FFDShow not only displays video using PAR from bitstream, but also auto-deinterlace it if it's interlaced. I tried Celtic Druid's 20041119 build it with anamorphic, interlaced XviD encodes in mp4 and it seems to work very nice and fast!
Some stuff added to interface (as commented by Milan):
"Bypass" method (in deinterlacing part) is useful for changing fields order.
"HW deinterlacing" (near to "Overlay mixer") is supposed to disable added changes (for backward compatibility), but doesn't work yet.
Thanks Milan and Minolta for new feature! :)

[SeeMoreDigital]

Hi pogo stick,

Can you provide a link to this build please?


Cheers

[pogo stick]

It's here: http://celticdruid.no-ip.com/xvid/ffdshow/

[minolta]

Even better, the auto-resize "expands" the window size in MPC (as opposed to shrinking image to window size). How...reinit image buffer? And it auto-deint divx as well. My old-favorite becomes my new-favorite decoder once again. Many thanks to Milan...:)

[Mug Funky]

hazzah! excellent.

i can start encoding interlaced again :)

my TV out is going to get SUCH a thrashing now.

[edit]

hmm... seems a little jumpy unless it's on "hardware deinterlacing". and in that case, i must output yuy2 for chroma to be done properly. the latter is probably a hardware issue with my archaic viddy card.

[edit 2]

output is really bloody good on HW deinterlace though. somehow nicer than powerDVD, even though i would have thought they did the same thing.

btw, for anyone out there with field-order problems, simply enable "offset" in ffdshow and move the pic up or down 1 pixel. turn it back off if field-order is correct, though. works a charm here.

[plonk420]

Originally posted by Mug Funky
output is really bloody good on HW deinterlace though. somehow nicer than powerDVD, even though i would have thought they did the same thing.

powerDVD 5 and earlier weren't very good IMO. watchable, but nothing you couldn't do with a simple remove even or odd field and reinterpolate. i don't know much about 6, tho.

OTOH, WinDVD 6's capabilities blew my socks off. seems to run at 50/60fps with non standard (24p/25p/30p) material, or something like a variable framerate. a friend said that FRAPS reported 50fps on a PAL source clip with sped-up sections (lazy editors) :p

[Mug Funky]

powerDVD 5 did dumb-bob, i think... i'm not sure. maybe i'm thinking of winDVD, as that's the first DVD player i had (ever!), along with the crappy and now dead Samsung drive it came with.

but a dumb-bob is all i need - just something to allow me to TV-out without randomly reversing field-order. i don't mind a little blur when the alternative is stuttering.

[IgorC]

would ffdshow support decoding of coming soon h.264 Ateme? It´s not comfortable to install (nero rom + nerovisonexpress3) approx. 50 Mbytes

[akupenguin]

yes, Real Soon Now (tm). But don't hold your breath, because I'm mostly trying to implement features in parallel with x264, not with Ateme.

[RadicalEd]

Aku, I feel the need to express my deep, explicitly heterosexual love for you at this instant.

[therealjoeblow]

Originally posted by BlindWanderer
I have a file with muxed subs that i'm not exactly sure how they were prepared but...

ffdshow when rendering the subs displays a box character at the end of the each line (mpc doesn't do this). I suspect it's a line feed but i can't be sure. ( 0x0A or 0xOD); consequently could we have added a subtitle spy (text field that displays the last text to be rendered) and a character filter as to remove characters entered.

Probably a unicode text sub muxed in. Try using disabling ffdshow's sub support and using vsfilter as a test. You then must make sure that you configure vsfilter to use a *unicode capable font* (one of the several variations of Tahoma is - you have to pick the correct one from the dropdown box when you choose Tahoma). If the box disappears or becomes a proper character, that's what the problem is.

FFDshow doesn't support unicode subs at the moment. I've asked for it here in this forum a couple of times. Hopefully it will come soon.

(of course, if that doesn't get rid of the box, then this wasn't the issue and you can feel free to yell at me...)

[therealjoeblow]

Originally posted by Mug Funky
hhe.. my system is messed up, but it's not particularly slow considering it's age. it outperforms a lot of machines built this year, simply because i removed all the bloat from win2k that i could find.

Not wanting to get too far off topic, could you provide a brief list of exactly what bloat you got rid of (and maybe how if it was complex)?

[Mug Funky]

i think there's possibly a memory leak while using AVIS for frameserving. i was serving a 2 gig DV file through avisynth into mencoder using AVIS, and towards the end my hard disk nearly exploded from swapping - i was using nearly 2 gigs of VM. not sure which part of the chain was doing it, but loading the DV file straight in gives me a constant memory usage, rather than an ever-increasing one.

tried both directshowsource and avisource in my script, but in both cases ffdshow was being used to decode, and also to serve. this may have something to do with it.

no big rush on this one - i should really start using avs2yuv anyway.

@ therealjoeblow:

hmm. while searching for a way to turn off windows messenger, i found (and subsequently lost) a page that described what all the windows services do, and which you could do without.

i'm sure a random google will turn up that page, or one like it. it's just a matter of a lot of registry hacking and disabling services through control panel's administrative tools.

i imagine it's harder to find all this in winXP... man, i am so glad i'm not running that - it's painful enough just to use it.

[iradic]

@therealjoeblow: try this ... http://www.techspot.com/guides-os.shtml

[leeperry]

i used to run ffdshow-20020617.exe which has always worked fine with all my WMV movies but the newest I could put my hands on : ffdshow-20041012-sse.exe, usually gives a BLACKSCREEN..

I've tried upgrading DirectX to 9.0c, and WMP to v10, to no success, and of course the old version worked like a champ when I downgraded..so something's broken like the default renderer or sumthing?

I run XP SP1, with the 61.76 Detonators on a Geforce4 Ti4200.

Could anyone see wth wrong ?
It's annoying to disable FFDSHOW each time :(

But apart from that, great piece of software!! 2 thumbs up :cool:

[Mug Funky]

hmm, as much as i don't like to deal with WMV, i've found that it works fine here. have you double-checked the codec settings so WMV is actually set to be decoded? the only thing i've tested with is the WMV of anna-nicole making a fool of herself at the AMAs the other week.

i'm playing through media player classic, on a p3 733, geforce DDR 32 (old old old).

[BlindWanderer]

Got a couple of bugs related to the Bitmap overlay.

1. can't use Vertical pos to push the overlay out of the frame.
2. extra line added to Bitmap overlay if overlay is of an uneven number of vertical lines (probably related to my input video being yv12). This extra line it adds is junk.

I'm using the December 1, build that milan posted for localizing ffdshow.

[Cyberace]

Hope I'm posting in the correct thread and are not off-topic with this request:

Microsoft DirectX Video Acceleration (DirectX VA) support

This feature/function request is both for hardware acceleration (motion compensation) and post-processing via Microsofts DirectX Video Acceleration (DXVA) API (http://msdn.microsoft.com/archive/default.asp?url=/archive/en-us/dx81_c/directx_cpp/htm/directxvideoacceleration.asp), (be nice if the options should be added in to FFdshow config settings when auto-detect if the video cards supported it and to what extent );

Hope FFdshow developer(s) can look into if this could be fully or partially used for most of its supported codecs?

Summery: Microsoft DirectX Video Acceleration (DXVA) allows DirectShow based software decoders to accelerate video playback directly on the graphics processors (GPU). If your graphics processor supports DXVA and has built-in technology to accelerate DVD and MPEG-2 file playback, then DXVA can provide (GPU) hardware acceleration. DVXA is an application programming interface (API) and a corresponding motion compensation (http://msdn.microsoft.com/library/en-us/graphics/hh/graphics/ddraw_2ab799f5-77bc-497c-9c64-68903bb940fb.xml.asp) device driver interface (DDI) for acceleration of digital video decoding. DDIs are also provided as part of DXVA; a deinterlacing DDI (http://msdn.microsoft.com/library/en-us/graphics/hh/graphics/dxvaref_d72180f0-84bc-4acf-b263-6f6c7e9f7b22.xml.asp) for deinterlacing and frame-rate conversion of video content, and a to support ProcAmp DDI (http://msdn.microsoft.com/library/en-us/graphics/hh/graphics/dxvaref_557ca488-5ca9-4595-994f-18eafa171f77.xml.asp) control and postprocessing of video content. DXVA provides an interface definition focused on support of MPEG-2 "main profile" video (formally ITU-T H.262 | ISO/IEC 13818-2), but is also intended to support other key video codecs (for example, ITU-T Recommendations H.263 and H.261, and MPEG-1 and MPEG-4).

A good exampel of the benifits of using DVXA for video acceleration is that you could decode HDTV (ATSC/DTV) MPEG-2 with a PIII 733Mhz CPU, a task that would normaly require a PIII 1400Mhz CPU if the card does not support DVXA HWMC. Or if you have a faster CPU then you can record something at the same time as you are playing back something else on the same Windows PC, (as with DXVA most of the playback effort gets loaded onto the GPU so the CPU can be doing something more/else).

Other post-processing featutes achivable via DVXA (if the video adapter hardware support it):
- Hardware iDCT motion compensation and subpicture decoding
- Hardware based IDCT (Inverse Discrete Cosine Transform)
- IQ (Inverse quantization)
- Fullframe video playback of HDTV and DVD content on slow CPU's (733Mhz +)
- Independent hardware color controls for video overlay
- Hardware colorspace conversion (YUV 4:2:2 and 4:2:0)
- 5tap horizontal by 3tap vertical filtering
- 8:1 up/down scaling
- Perpixel color keying
- Multiple video windows supported for CSC and filtering
- DVD subpicture alphablended compositing
- Alpha-Blending Surface

Developer links:
- Microsoft Developer information on DirectX Video Acceleration (http://msdn.microsoft.com/library/default.asp?url=/library/en-us/graphics/hh/graphics/dxvaguide_30c21a63-691d-4ada-b469-08790b3b6c29.xml.asp)
- How Decoders Use IAMVideoAccelerator (http://msdn.microsoft.com/archive/default.asp?url=/archive/en-us/directx9_c/directx/htm/howdecodersuseiamvideoaccelerator.asp)
- Mapping DirectX Video Acceleration to IAMVideoAccelerator (http://msdn.microsoft.com/archive/default.asp?url=/archive/en-us/directx9_c/directx/htm/52mappingdirectxvatoiamvideoaccelerator.asp)
- DVXA API/DDI Specification (Rev 1.0) (DirectX 8.1 C++ Archive) (http://msdn.microsoft.com/archive/en-us/dx81_c/directx_cpp/htm/directxvideoacceleration.asp)
- DirectShow DirectX Video Acceleration Video Subtypes (http://msdn.microsoft.com/library/en-us/directshow/htm/directxvideoaccelerationdxvavideosubtypes.asp)
- Enabling DirectX Video Acceleration in a custom player (http://msdn.microsoft.com/library/default.asp?url=/library/en-us/wmform95/htm/enablingdirectxvideoacceleration.asp)
- Windows Media Format 9.5 SDK Enabling DirectX Video Acceleration (http://msdn.microsoft.com/library/en-us/wmform95/htm/enablingdirectxvideoacceleration.asp)
- DirectX Video Acceleration Video Subtypes (DirectX 9.0 C++ Archive) (http://msdn.microsoft.com/archive/en-us/directx9_c/directx/htm/directxvideoaccelerationdxvavideosubtypes.asp)
- IAMVideoAccelerator Operational Specification (DirectX 9.0 C++ Archive) (http://msdn.microsoft.com/archive/en-us/directx9_c/directx/htm/522directxvaiamvideoacceleratoroperationalspecification.asp)
- DirectX Video Acceleration Motion Compensation Callbacks (http://msdn.microsoft.com/library/default.asp?url=/library/en-us/graphics/hh/graphics/ddraw_28178ef1-ce67-476d-bdf0-f854fc342b4c.xml.asp)
- Calling the Deinterlace DDI from a User-Mode Component (http://msdn.microsoft.com/library/default.asp?url=/library/en-us/graphics/hh/graphics/dxvaguide_6691221e-f352-47f9-8345-b38d0955fd2a.xml.asp)
- Per-Pixel Alpha Blending (DirectX 9.0 C++ Archive) (http://msdn.microsoft.com/archive/en-us/directx9_c/directx/htm/perpixelalphablending.asp)
....AND MUCH MORE available on a Microsoft Development Network search (http://search.microsoft.com/search/results.aspx?qu=dxva&View=msdn&st=b&c=0&s=1&swc=0)

Exampel of above DXVA technology in use is the newly announced "nVIDIA DVD Decoder", (as almost GForce/nForce GPU's support it).
nVIDIA DirectX Video Acceleration (DXVA) support MPEG-2 acceleration for:
Inverse quantization (IQ),
Inverse discrete cosine transform (IDCT)
Motion compensation (mo comp)
Enables advanced de-interlacing
Decodes high-definition MPEG-2
http://www.nvidia.com/object/dvd_decoder.html
http://www.nvidia.com/object/decoder_faq.html
(nVIDIA's part in the GPU that support this is called HDVP (http://www.nvidia.com/attach/149) (High Defininition Video Processor) / VPE (Video Processing Engine))

PS! My personal reason for wanting DXVA support by FFdshow is to have HDTV playback of MPEG-2/TS on slow PC

[Boco]

Think there might be a bug in divx 3 decoding. I got a file that crashes to the desktop when CelticDruid's latest build tries to decode it, but runs with the latest alpha on sourceforge (albet with a large green line on the bottem). I am using the cvs build (6.4.8.3) of MPC.

[celtic_druid]

No problems here with DivX 3 decoding either via dshow or VFW. Only tested two files though.

[Boco]

It was in a download of 26 similar video files. Only 1 crashed on open, and 1 other one crashed after playing for awhile. The rest played fine...

[dragongodz]

It was in a download of a 26 episode series.
:readrule:

[spyder]

@Cyberace

I believe it would be difficult to use DXVA in ffdshow's MPEG2 decoding as it's based on libmpeg2 which does all of those accelerated things internally. It could probably be used for scaling/deinterlacing etc though. On a side note, why do you need ffdshow for decoding HDTV MPEG2? Most software DVD player MPEG2 decoders will play it fine and even use DXVA to do it. I can play 1080i content with about 20-30% CPU usage on my 2500+ Athlon XP

[marius1]

Which version of AC3Filter's code is in the 1012 version of ffdshow? For AC3 encoding am I better off sticking to ffdshow or should I use the RC5 of AC3Filter? Thanks in advance.

[Cyberace]

Originally posted by spyder
@Cyberace

I believe it would be difficult to use DXVA in ffdshow's MPEG2 decoding as it's based on libmpeg2 which does all of those accelerated things internally. It could probably be used for scaling/deinterlacing etc though. On a side note, why do you need ffdshow for decoding HDTV MPEG2? Most software DVD player MPEG2 decoders will play it fine and even use DXVA to do it. I can play 1080i content with about 20-30% CPU usage on my 2500+ Athlon XP "difficult" or impossible? :rolleyes: ...every little helps and never say never ;)
(I got a 800Mhz PIII (underclocked) fanless PC as HTPC, yes I'm poor)
PS! Others could have faster PCs' but encoders running simultaneously

[kurt]

Originally posted by marius1
Which version of AC3Filter's code is in the 1012 version of ffdshow? For AC3 encoding am I better off sticking to ffdshow or should I use the RC5 of AC3Filter? Thanks in advance.

that depends on using spdif output or not .. on my machine ffdshow's AC3/dts audfilter doesn't work via spdif .. so I'm using latest AC3Filter (of course in wave out audio renderer) ..

[marius1]

Originally posted by kurt
that depends on using spdif output or not .. on my machine ffdshow's AC3/dts audfilter doesn't work via spdif .. so I'm using latest AC3Filter (of course in wave out audio renderer) ..

Actually ffdshow with AC3 audio works for me on spdif output. The only main diff between them I see in terms of functionality is that AC3Filter has the normalization feature.

[spyder]

@Cyberace

Of course it's not impossible. With enough rewriting you can make libmpeg2 decode MPEG4. The point is that it would probably require quite a few internal changes and I don't know that anyone is wanting to do that. As I said before, there are plenty of mpeg2 decoders that do this already. Using ffdshow to do it probably won't gain you much advantage over using the others as you can't do any post decode processing on the image(aside from deinterlacing/scalng maybe) using DXVA anyway.

[Cyberace]

Originally posted by spyder
@Cyberace, there are plenty of mpeg2 decoders that do this already.Are those mpeg2 decoders free and/or open source!?, in any case you can't argue with the fact that DvXA support in ffdshow would be a plus for all ffdshow users, if only some had the time and skill to code it for FFdshow, (and not just for MPEG2 as again DxVA support MPEG4 too if the GPU do that too (like VIA's latest GPU)


Originally posted by spyder
@Cyberace, you can't do any post decode processing on the image(aside from deinterlacing/scalng maybe) using DXVA.Personaly I think if deinterlacing (for MPEG-2) and scaling (for codec) could be done in GPU hardware that would be great!


PS! As all the specs/info for the DxVA interface are available online from Microsoft I would think it's more a matter of time and motivation than a matter of expertise on the developers side, (I could not do it myself as I'm not a programmer but probebely most existing FFdshow developers could?), don't you agree?

[Mug Funky]

cyberace: if your rig is too slow, try using mplayer to decode. it's not unusual to get double the performance under win32. i have no idea how fast it is in linux (i'd bet money that it's faster still).

the drawback is no hardware deinterlace and having to use commandline.

[nicco]

Is here any project to make ffdshow full compatible with nero digital?:)

[SeeMoreDigital]

Originally posted by nicco
Is here any project to make ffdshow full compatible with nero digital?:) Now that NeroDigital's Mpeg4/AVC codec is officially available, I guess it's much easier for the FFdshow developers to make a decoder that works with all of NeroDigital's implementations.... so lets hope so.


Cheers

[BlindWanderer]

cyber: microsoft has a bad habbit of publishing documentation that isn't always accurate.

[Cyberace]

I have to say that I am very disappointed that you all are so quick to dismiss the idea of DxVA :(
I just hope the mistake is mine, that I asked the wrong crowd or at the wrong time and place :p

(PS! I won't disturb you anymore in this thread unless get some insight for a FFdshow developer)

[filewalker]

ffdshow-20041212.exe is out.


http://celticdruid.no-ip.com/xvid/ffdshow/

But I can't find a changelog...

Cu

[Amour]

Originally posted by filewalker
ffdshow-20041212.exe is out.

http://celticdruid.no-ip.com/xvid/ffdshow/
I tried it, but now when I close a player, sometimes I got this error message (sorry, it's in French):
L'instruction à "0x038c1177" emploie l'adresse mémoire "0x03bd83a0". La mémoire ne peut pas être "read".

Cliquez sur OK pour terminer le programme.

And here is a translation made by myself:
The instruction at "0x038c1177" uses memory address "0x03bd83a0". Memory cannot be "read".

Click OK to close the programm.

[hhanh]

I debugged through the last CVS version of ffdshow. This AV comes from the destructor of the encoder but the real problem comes from this line

-- src/ffmpeg/h264.c
if((avctx->extradata_size == 0) || (avctx->extradata == NULL)) {
av_log(avctx, AV_LOG_ERROR, "AVC codec requires avcC data\n");
return -1;
}

The strange thing is that I think the avcC is present because I could see one with mpeg4ip and also, this code is not in the ffmpeg latest alpha. VideoLan works so I'm looking into how it parses the sequence headers.

--h

[Leo 69]

Does it decode Recode's .mp4 files (avc)?

[hhanh]

Yes, with some limitations. B-Frames & weighted prediction are not supported AFAIK.

[Leo 69]

But...bidirectional prediction and weightlifting prediction
has not been supported by VLC-player for a long time already.
And something might have changed,since FFDShow and VLC player use the same libraries,don't they?Thats why I'm asking.

[hhanh]

No, they don't use the same libraries. ffdshow uses libavcodec and videolan has its own implementation.

--h

[chilledoutuk]

erm afaik both ffdshow and vlc use the ffmpeg project codecs that includes libav codecs.

ffmpeg supports playback of avc video streams as long as neither bframes or weighted prediction is used at the moment. But i have been informed that bframes will be supported soon which is great.

ffdshow is a directshow playback filter for an array of formats and it can play avc but in order for you to be able to play avc in .mp4's you need a .mp4 stream splitter that can deal with it containing avc video. currently afaik only the nero splitter can do this (3ivx etc cant well not for me) which means you need vision express installed to playback .mp4 avc in directshow media players.

Vlc on the other hand because of its independance of system codec and directshow settings has its own .mp4 splitter that can cope with avc in .mp4 files but is still limited to ffmpegs implimentation of avc (meaning no bframes or wighted prediction yet!!)

FYI I am using a nightly build of VLC i am not sure if the current release can play avc in .mp4's or not though.

[hhanh]

Yes, you're right - the decoding part is the same. I was looking at the h264 splitter code.

[saratoga]

Originally posted by BlindWanderer
cyber: microsoft has a bad habbit of publishing documentation that isn't always accurate.

Err not for their APIs they don't. MS lives and dies on developers using their APIs. I've always found their stuff to be extremely good in this respect.

Are you thinking of documentation for their proprietary formats maybe?

[plonk420]

It was in a download of a 26 episode series.
Originally posted by dragongodz
:readrule:

that's not necessarily illegal, not like a downloaded movie... that's the creed fansubbers go by at least ;p

[celtic_druid]

26 eps though probably means Star Trek and last time I checked Paramount didn't like people downloading their stuff.

[yaz]

Originally posted by chilledoutuk
... in order for you to be able to play avc in .mp4's you need a .mp4 stream splitter that can deal with it containing avc video. currently afaik only the nero splitter can do this (3ivx etc cant well not for me) which means you need vision express installed to playback .mp4 avc in directshow media players ... remember e.g. the moonlight demuxer! it can chain nero ds and ffdshow as well (however, still has some seeking issues and conflicts w/nero parser). so, u can create and use avc.mp4 wout nve, pretty well.
the bests
y

[dragongodz]

that's not necessarily illegal, not like a downloaded movie... that's the creed fansubbers go by at least ;p
no but he doesnt say it is a fansub or whatever either. anyway i will kindly point people to this thread
http://forum.doom9.org/showthread.php?threadid=86626
which also contains this link
http://forum.doom9.org/announcement.php?s=&forumid=6

[Amour]

Originally posted by filewalker
ffdshow-20041212.exe is out.

http://celticdruid.no-ip.com/xvid/ffdshow/
Another problem : I cannot browse (forward, backward) my VCD (mpeg 1/2 ?) anymore using this version of ffdshow. :(
When I do it, the video and sound freeze both.

[SeeMoreDigital]

Originally posted by Amour
Another problem : I cannot browse (forward, backward) my VCD (mpeg 1/2 ?) anymore using this version of ffdshow. :(
When I do it, the video and sound freeze both. Also.... Can anybody confirm whether the DAR signalling detection works now?


Cheers

[therealjoeblow]

Can anyone give some *clear* bulletproof instructions on how to make this work? I encode from TV with my ATI All-in-wonder using latest ffdshow as the encoder, mpeg4 set to xvid, using the basic default settings but also enabling the 'interlacing' tickbox on the general tab under 'other options', nothing else.

I've enabled the "use overlay mixer" on the ffdshow 'output' tab, and it never deinterlaces automatically. I have to manually turn on the 'deinterlacing' tickbox in the sidebar for deinterlacing to actually occur.

It's a pain turning it on and off depending on what I want to watch, so was *really* hoping this automatic solution would work...

thanks for any help

[Teegedeck]

The 'interlacing' option ensures that interlaced content is encoded correcty - as interlaced. It doesn't deinterlace your video.

[therealjoeblow]

Originally posted by Teegedeck
The 'interlacing' option ensures that interlaced content is encoded correcty - as interlaced. It doesn't deinterlace your video.

I never meant to imply that I thought this would do the de-interlacing. I just wanted to make it clear that the file *was* properly encoded as interlaced, and therefore there should be no reason with the source that auto-deinterlacing doesn't work on playback.

So, knowing the source is ok, why doesn't the auto feature work??? That's my question and point of frustration.

[eb]

Amour wrote:
quote:
--------------------------------------------------------------------------------
Originally posted by filewalker
ffdshow-20041212.exe is out.

http://celticdruid.no-ip.com/xvid/ffdshow/
--------------------------------------------------------------------------------


Another problem : I cannot browse (forward, backward) my VCD (mpeg 1/2 ?) anymore using this version of ffdshow.
When I do it, the video and sound freeze both.

Take care, there are many versions of ffgshow from 12Dec.2004.
One designated for SSE, another for SS2 procesors...

http://img147.exs.cx/img147/1305/ffdshow4dc.jpg

For Athlon 2200+ I am using that with size 2,030,080/no troubles/.

eb

[celtic_druid]

Those are 12th Oct, not Dec.

[eb]

Those are 12th Oct, not Dec.

There are many ways to write the date, I prefer 12th Dec 2004,
eb

[celtic_druid]

Yes, there are many ways to write the date, but using 12th Dec 2004 to describe the 12th of October is not a valid one.

[eb]

@ celtic_druid,

I am very sorry, that is my foult.

... but care to all, use a proper version for your computer...

eb

[b0b0b0b]

Hi celtic druid, thanks for posting these builds. I appreciate them

[eb]

@ celtic_druid,
Now I know (but you knew before)why date 20041212 was so hardly coded in my mind. 1 minute ago I recoded it to 20041223.
No problems, no crush with recording on fly to .avi from digital sattv.(SS2 DVBcard)
How to name container with mp2/MPEG1 audio and Xvid compression?

to others, please visit http://celticdruid.no-ip.com/xvid/ffdshow/

eb

[Sharktooth]

latest ffdshow versions from celtic druid are mirrored here: http://ebola.gamersrevolt.it/celticdruid/ffdshow/

[eb]

@ Sharktooth,

Could you please suggest matrices for Xvid for recording on fly from digital sattv using Xvid compression from mpeg2 video stream, progressive, interlaced, PAL 25 frames/s ?

eb

[DeathTheSheep]

Will "slow"/"high quality" or "2-pass" mode ever be available for ffdshow? 'Cuz I need a nice vfw to do ma compression stuff, so yeah, I can't use those dang (hoooboy) MP4 files 'cuz they can't be ported, played (by most stuff), or, the clincher, edited in Vdub. So I was wonderin' whether or not major dev. of x264 was all done or if it is still in progress, lol

Cheers, baa

[bond]

i just wanted to open two different avc-in-avi files in good old avscompare tool, but it seems that its not possible to open two files at the same time via directshowsource() in .avs when using ffdshow as avc decoder (btw the same goes for the vss avc decoder)

anyone mentioning a similar problem?

[Neo Neko]

You might have to change the option to block more than one instance of the filter in the codec config. Just a guess. But it seems likely. ;)

[celtic_druid]

@DeathTheSheep, 2 pass x264 with ffdshow is available. It just doesn't work quite like the other codecs.

[bond]

Originally posted by Neo Neko
You might have to change the option to block more than one instance of the filter in the codec config. Just a guess. But it seems likely. ;) thats what i thought too, but it doesnt work (avscompare also crashes if you try to open one file with ffdshow and one with videosoft... :( )

[DeathTheSheep]

2-Pass lives?! In ffdshow? But... how? ;-*

[bond]

another issue:
i tried connecting latest ffdshow compile from celticdruid to the latest nero .mp4 parser, they can connect, but when i try to play the graph i get the error message: "the graph couldnt change state" :(

[DeathTheSheep]

I get a similar error, but it is usually remedied by modifying the ffdshow filter properties prior to its placement in a chain. This error is due to the fact that upon altering ffdshow's properties, the filter may not be able to change its own state. That's what I've gathered, at least. Hope I could be of assistance.

Cheers, baa

[bond]

whats the status on h.264 b-frames? somehow i remembered someone saying that b-frames without using cabac should work? actually my last tests showed that this isnt true :(

Originally posted by DeathTheSheep
I get a similar error, but it is usually remedied by modifying the ffdshow filter properties prior to its placement in a chain. This error is due to the fact that upon altering ffdshow's properties, the filter may not be able to change its own state. That's what I've gathered, at least. Hope I could be of assistance.hm, what for example do you change exactly? thx :)

[Sirber]

When I try to "use" the ffdshow stats for H264, I get Error #100 :(

[timeismoney]

celtic_druid, the ICL build seems broken in avisynth, would you like to make a simple msvc build?:p

[Shinobu]

use mod*16 and yv12 as an input you won't have de error code 100 anymore

++

[celtic_druid]

You mean using ffavisynth.dll? Works fine here. What exactly is the problem?

[timeismoney]

Originally posted by celtic_druid
You mean using ffavisynth.dll? Works fine here. What exactly is the problem? No, error in ffdshow options 'Avisynth' to load avs

milan's 20041012 is ok, but 20041012-sse or sse2 is fail, too

[celtic_druid]

http://celticdruid.no-ip.com/test/ffdshow.msvc.7z

Seems to work ok.

[timeismoney]

Yes, it works. Thanks.

Here is a question to make a ffdshow build: I install VS2003, DX9c SDK (Dec. 2004) and NASM, when I try to compiling the CVS, there're too much error (over 100+) in head file "simd.h", thus I can make all the dlls but not ffdshow.ax.

Is there something I have to install or something to config?

[slavickas]

Originally posted by timeismoney
Yes, it works. Thanks.

Here is a question to make a ffdshow build: I install VS2003, DX9c SDK (Dec. 2004) and NASM, when I try to compiling the CVS, there're too much error (over 100+) in head file "simd.h", thus I can make all the dlls but not ffdshow.ax.

Is there something I have to install or something to config?

just guess:rolleyes: :
have You build baseclasses of dxsdk, it's very often is forgotten iirc it's needed

btw can somebody build libavcodec with latest h264 patch ,pls :) , i'm outta luck smthng wrng :mad:

[eb]

http://img129.exs.cx/img129/7744/onliveto0aj.jpg



Using ffdshow + SS2 card + SkyAVC.ax from SkyView program by marfi we can record on live to .avi
In this example it is live record to cropped to 720x416 .avi with original mp2 audio + at the same time recording full "spare" video as mpeg2 to make quality comparison. Procesor Athlon 2200+ was working with 1890MHz with 80-98%.
Samples named as NEO NEKO /why, see the end of the sample/can be found at
ftp://www.eb.enterpol.pl
user name www.eb.enterpol.pl
pass eb


eb

edited
PS Ofcourse we can record to.avi from the SkyView program, but to make some experiments it is easer this way.
Similiar tests can be done with MyTheatre2.76 filter.on another sattv DVB cards
edit2
Another sample with
G day .Live record to .avi

[bond]

seems as if ffmpeg now has the ability to decode avc b-frames correctly! hopefully ffdshow can be updated too soon :)

[Leo 69]

I'm so glad to hear that Bond!Where do you know it from?
Seems like VLC player will be able to handle b-frames too!!!No?
I'm dreaming about it :) At last there'll be a possibility
to decode .mp4 normally without that huge NVE software :D

[bond]

Originally posted by Leo 69
I'm so glad to hear that Bond!Where do you know it from?on the ffmpeg maillist (michael niedermayer said the batch has been added)

Seems like VLC player will be able to handle b-frames too!!!sure, videolan like mplayer and ffdshow uses ffmpeg for decoding avc

I'm dreaming about it :) At last there'll be a possibility
to decode .mp4 normally without that huge NVE softwarewell thats already possible for a long time! .mp4 is nothing only nero uses...
have a look at the mp4 faq i link to in my signature

[Leo 69]

@ Bond

Did Michael tell anything about weighted prediction support?
That would be just perfect :)

[bond]

hm, there seems to be still one issue with b-frames left: the frames get outputted in coding order, not in playing order :(

Originally posted by Leo 69
Did Michael tell anything about weighted prediction support?nope

[CruNcher]

yep b-frames are better now very nice :D

[timeismoney]

Yep, cvs of 20041231 has already support avc b-frames, but with slight shrink/dithering

[CruNcher]

@timeismoney
you are Selling this Storm Codec pack ?


About Storm Codec

As an addon & enhancement of Windows Media Player,Storm Codec provide a solution to
most popular media files, streams & VCD/DVD. It can play RealMedia, QuickTime, MPEG-4
(DivX, XviD, 3ivx, MP4, FFGS, H264...), MPEG-2, HDTV, AC3/DTS/LPCM, VP3/VP6, Theora,
TTL2, Subtitles, OGG/OGM, Matroska, APE, FLAC, AAC, MPC, Voxware, FLiC, 3GP/AMR, etc.
Any other Media Player is not requisite (But latest Windows Media Player it recommended).

Latest Version is 5.00.-11 RC3

RealPlayer 10.5 Core Codec & ActiveX 6.0.12.1056
RealMedia Splitter 1.0.0.9
QuickTime Player 6.52 Core Codec & ActiveX (Including 3GP/AMR support)
Moonlight MP4 Demultiplexer 1.0.3.41110
Kopei's XviD Codec 1.1 alpha (Build 20041106)
ffdshow MPEG-4 Codec 20041201 (with libfaad2, liba52, libdts & libtheora)
On2 VP6 Codec 6.2.6.0
On2 VP3 Codec 3.2.6.1
Ligos Indeo Codec 3.2-5.2
TTL2 Decoder 30623
Cyberlink DTV Video/SP Filter 3.5.0.4009
Moonlight MPEG-2 Demultiplexer 3.1.190.41018 MOD
CoreVorbis 1.0
OGGSplitter 1.0.0.0
AACParser 1.1
CoreFLAC Audio Decoder & Source DirectShow Filter 0.4.0.46
DS Monkey's Audio Filter 1.0 + APE Audio Lib 3.99u4
TTA DirectShow Splitter & Decoder 1.0.0.203
RadLight MPC DirectShow Filter 1.0.0.3
Voxware Metasound & Metavoice ACM Codec
Matroska Splitter 1.0.2.4
VSFilter (DirectVobSub) 2.33
Media Player Classic 6.4.8.3 Inofficial Version
G.Spot 2.52 Beta1
LiveUpdate 20040922
Storm Codec Setting Tools 20041218
Nic's FourCC Changer
CNNIC Cilent (SimpChineseVersionOnly)

File: StormCodec5.00.-11.exe
Size: 14,266,338 Bytes
MD5: 51c00f03be85de1f1cfa562fa316dfdf

Version History please click here!

Important

Before you install this EVALUATION software you must agree that Ringz Studio decline all
responsibility for any mistake, error and the causing loss or harm of / by this EVALUATION
software. If you don't agree with it, please DON'T download or install it.

[timeismoney]

Selling? I think it must be some mistake in the expression of the intro. "EVALUATION" there not means you have to buy a "FULL" version, it just means it is a unstable version, maybe bugs or lacks in it, so the users should use it as their own risk...I'll tell them to change it.

Well, till now I know the "Storm Codec" is all free and will be free till it's end, just a enthusiast's pack for juniors. Maybe superiors will install and set everything solely, but some Codec Packs are useful for juniors who can't understand how to install and config everything. That's the reason of K-lite or other ones' existing.

So I think we can just take it easy...

And MPC 6483 comes from the CVS, they use celtic_druid's build or compiled it themselves.

PS: I said "They", because I'm the tow pretty girls' teacher.

[Amour]

Originally posted by eb
please visit http://celticdruid.no-ip.com/xvid/ffdshow/No longer available.
Originally posted by Sharktooth
latest ffdshow versions from celtic druid are mirrored here: http://ebola.gamersrevolt.it/celticdruid/ffdshow/
You don't have the 20041231. :(

[akupenguin]

Originally posted by bond
hm, there seems to be still one issue with b-frames left: the frames get outputted in coding order, not in playing order Fixed, at least for the common case.
(H.264 allows arbitrary frame order, not just the simple B-frame delay from ASP. But I don't know any encoders (not even JM) that do that yet.)

[marcellus]

No longer available.
Try again, is still there.
Originally posted by Amour
No longer available.

You don't have the 20041231. :(
The latest celticdruid's compile is still 20041223. Timeismoney talked above about cvs, not a compile.

[timeismoney]

Yes, CVS has some update in h264

http://tinypic.com/162iky

[celtic_druid]

Actually I did one the other day, I just never put it up. Been so many updates lately. I might put one up now. Takes ages to compile though compared to most other stuff.

[bond]

Originally posted by akupenguin
Fixed, at least for the common case.
(H.264 allows arbitrary frame order, not just the simple B-frame delay from ASP. But I don't know any encoders (not even JM) that do that yet.) great! so this means that b-frames are now supported in ffmpeg? :)

[Irwin]

Sorry OT here but

Where i find in source http://cvs.sourceforge.net/viewcvs.py/ffdshow/ffdshow/src/imgFilters/ sharpen-swscaler filter?

[Leo 69]

Hey guys!Does anybody know if videolan developers included
the latest ffmpeg (with b-frames support)into their last nightly build of VLC-player?I got a slow modem,so I can't just check it out,you know:D Besides they seem to be sleepy on their forum - no replies,etc...
Thanks in advance
Leo

-----------addition---
They've just submitted a reply where they say the ffmpeg update for VLC is only made once in a while...Gouch..I can't wait any more...

[Leo 69]

I am very disappointed that the latest ffdshow doesn't play Recode's
mp4 files smoothly (those which are with b-frames). Well, it works,
but with sporadic freezes, as if I got a slow PC.But processor load is only about 50% on my Athlon 2800+. Does anybody have similar problems?

Cheers
Leo

[slavickas]

Originally posted by Leo 69
Hey guys!Does anybody know if videolan developers included
the latest ffmpeg (with b-frames support)into their last nightly build of VLC-player?I got a slow modem,so I can't just check it out,you know:D Besides they seem to be sleepy on their forum - no replies,etc...
Thanks in advance
Leo

-----------addition---
They've just submitted a reply where they say the ffmpeg update for VLC is only made once in a while...Gouch..I can't wait any more...

nope seems nigtly vlc build uses older ffmpeg, although i have no problem playing with ffdshow latest, well if weighted pred not used

[giandrea]

I compiled the last Mplayer snapshot, that I suppose has the last FFmpeg CVS, because this night I encoded a short clip with NeroDigital AVC (B-frames, no weighted prediction, no psichovisual and no deblock) and it plays smoothly.
The only thing that I noticed is a possible lack of quality during playback, because the clip was encoded at 1000 kbps and its quality was worst than an Xvid clip encoded at 1200 kbps expecially in dark areas. Is that possible?
This is a screenshot, Xvid on left, Nero AVC on right, notice the right bottom corner of the frame, and how it is much more detailed in Xvid.
http://www.motoriweb.com/Public/screen.png
bye,

giandrea

[edit] added image

[netchris]

Giandrea : Yes nero avc has a problem with dark areas, its like wrong use of psuchovisual (though anomalies in black are still there even when no psycho are used).
Its an excellent progress to be able to play nero's files with ffdshow!
Well done to the coders.
It aint yet perfect but very close to it :)

[Leo 69]

Something is definetely wrong with b-frames in ffdshow guys,I'm sorry
It freezes ALL the time!
If you don't believe me here's my sample.Simply play it first in ShowTime and then in any directshow based player using ffdshow decoder.I guess you'll see the difference:rolleyes:

http://cp.people.overclockers.ru/cgi-bin/dl.pl?id=4231&filename=U571.mp4

Cheers

[kurt]

@ leo: i think u are not the only one who have problems with this....
but give the devs some time to fix that ;)

[giandrea]

Leo: I just compiled mplayer from CVS, and your clip plays smoothly! :confused: What are you using for playback? Anyway i suppose that it is a 12.5 seconds clip.
bye,

giandrea

[plonk420]

Originally posted by celtic_druid
26 eps though probably means Star Trek and last time I checked Paramount didn't like people downloading their stuff.

sounds like you speak from experience ;)

[Leo 69]

@ giandrea

I'm using ffdshow-20050105 from celtic-druid & either MPC or BSPlayer.
The playback is really choppy :( I very doubt that smth is messed up
in OS,well I know how to keep it in repair..
Anyway could you upload your Mplayer build somewhere and give a link? That would be cool:)

Cheers mate :)

[celtic_druid]

Two factors though if you are playing mp4's. What splitter are you using? Perhaps try an avi sample and see if you have the same issues.

[Leo 69]

@ celtic_druid

Those avis created by x264 play absolutely fine!!!
Besides I can't see the splitter while playing .mp4..Is this normal?

http://images.people.overclockers.ru/17522.png

I guess not :confused: Help!

[Taurus]

On a Win98 machine using ffdshow-20050105 on mediaplayer classic every avi, h.264 crashes mediaplayer @ closing the program.
ff_acm.acm is located in C:\Windows\system and in ffdshow program dir.
It is called via system.ini.
Former ffdshow installations didn't show this behaviour(tested on ffdshow-20041012.exe and ffdshow-20040808.exe.
Disabling the call of ff_acm.acm in the system.ini solves the crashes.

Cheers

Taurus

[bond]

Originally posted by Leo 69
Besides I can't see the splitter while playing .mp4..Is this normal?yes, thats the difference between a parser and a splitter (you are using the nero parser, which is also a source filter at the same time)

[BlindWanderer]

I may just be crazy but SPP deblocking seems a bit heavy when used on internaly decoded video from x264 compaired to xvid. Haven't done much testing on this. (I was noticing major differences with the faces; was running SPP with processing strength at 128, soft threashold off, automatic QC turned off, and the presets bar full forward)

Noticed on the Last Samurai samples found in the sex264 thread.
http://forum.doom9.org/showthread.php?s=&threadid=87719&perpage=20&pagenumber=5

[akupenguin]

Originally posted by BlindWanderer
I may just be crazy but SPP deblocking seems a bit heavy when used on internaly decoded video from x264 compaired to xvid.
H.264 is internally deblocked. Hence the "deblock" parameters you get to set while encoding. If you deblock it again, of course it will end up blurry.
(Yes it's less flexible this way, but in-loop deblocking improves the quality of motion compensation, so less filtering is needed in total.)
(Note that ffdshow hasn't implemented H.264 in-loop deblocking for B-frames as required by the standard, so that doesn't reflect the true quality.)

[hellfred]

Originally posted by akupenguin
(Note that ffdshow hasn't implemented H.264 in-loop deblocking for B-frames as required by the standard, so that doesn't reflect the true quality.) [/B]
So what soulutions do you suggest for a full featured playback?
Does one need to have a win32 system and neros dshow filters from latest Recode? And how to ensure that the filter is used in a dshow based player?

Hellfred

[akupenguin]

I hear that VSSH works.
But I don't want to resort to binary DLLs, so I'll just finish libavcodec. It shouldn't take too long.

[708145]

SPP strenght is relative to the average quant per frame.
maybe it doesn't understand h.264 quant range and filters like ASP quant 18 (very heavy) for AVC quant 18 (should be light filtering).

Just a guess.

bis besser,
T0B1A5

[hellfred]

Originally posted by akupenguin
I hear that VSSH works.
But I don't want to resort to binary DLLs, so I'll just finish libavcodec. It shouldn't take too long.
That would be great and highly appreciated.

Stefan

[Leak]

Originally posted by 708145
SPP strenght is relative to the average quant per frame.
maybe it doesn't understand h.264 quant range and filters like ASP quant 18 (very heavy) for AVC quant 18 (should be light filtering).

Whoa - that'd be a tad heavy... :D

By the way - will there be an update of SmoothD sometime in the future, or is it a bit dead for now? ;)

np: Pan Sonic - Vähentajä / Diminisher (Kesto (234.48:4) (Disc 1))

[708145]

Originally posted by Leak

By the way - will there be an update of SmoothD sometime in the future, or is it a bit dead for now? ;)


I'm doing one project at a time. But I'll debug SmoothD after the current one (ELDER).


bis besser,
T0B1A5

[Leak]

Originally posted by 708145
I'm doing one project at a time. But I'll debug SmoothD after the current one (ELDER).

Thanks for the info.

np: Pan Sonic - Virtamuuntaja / Current-Transformer (Kesto (234.48:4) (Disc 2))

[bond]

actually you might want to know that akupenguin (aka pengvado) also implemented weigthed prediction support to ffmpeg now too.
thanks a lot!!! :)

[Leo 69]

@ Bond

At last! Now we'll wait for celtic_druid's response (updated ffdshow) :D

[Franky]

Has anyone been able to download celticdruids latest build (20050111) yet, and if so, could that person please mirror it?

[BlindWanderer]

ffdshow build 20050111 crashes on raw video, problem not present in 20050105 (tried two different sources of different stream formats, wmv3 in mkv & mpeg2 in mpg). The crash was silent.

[Yong]

same here, but it don't want to play my video files at all...
crash ffdshow.ax...
I will send a bug report to milan.:)

[kurt]

it chrashes also by decoding xvid... like BlindWanderer said no problems with ffdshow 20050105

[Ark]

For me lastest build 20050111 is broken, it install correctly (altough a bit slow on the x264 dll file...) but i can't access any panel, a pop-up window of Rundll32 appears saying that an exception has occured on ffdshow.ax....

[celtic_druid]

11.01 was compiled with VC7, 05.01 was compiled with VC6 which I think explains the problem with RAW video as it seems to involve msvcr71.dll.

Might put up a new VC6 build, looks like libavcodec has been updated to.

No problems here decoding XviD by the way.

[LotharZ]

Originally posted by Ark
For me lastest build 20050111 is broken, it install correctly (altough a bit slow on the x264 dll file...) but i can't access any panel, a pop-up window of Rundll32 appears saying that an exception has occured on ffdshow.ax....

Same problem here, cant open any video format or config panel.

[kurt]

Originally posted by celtic_druid
No problems here decoding XviD by the way.

you are right - problem here is if i enable "picture properties" (like in previous build) then zoomplayer crashes...

[celtic_druid]

No problem here with picture properties enabled, although I am running a newer compile. Also no issues when enabling for RAW video.

Will put it up later when I am done compiling libavcodec.dll, etc. and have built the installer.

[Leak]

Originally posted by celtic_druid
No problem here with picture properties enabled, although I am running a newer compile. Also no issues when enabling for RAW video.

Will put it up later when I am done compiling libavcodec.dll, etc. and have built the installer.

By the way, celticdruid - exactly what connection does your server have? Cable?

Whenever I downloaded something from it in the last few days, I was lucky if I got more than 1 or 2 kB/sec - when I could connect at all... :(

(I hope you don't take this as a complaint; I'm just curious, and for getting the files eventually there's download managers... :))

[CruNcher]

New ffdshow crashes here directly after trying to load a video (mpc closes) hmm it crashes imidietly becouse the OSD is active, disableing it i can at least watch the video but when i try to access the config panel, it's crashing again :P

[celtic_druid]

Works fine here Win2k SP4 with an Athlon 64 3500+. OSD works fine. Config works fine.

Connection is only 512 ADSL and I agree it is way too busy, way more than I ever really intended. I have even gotten over 9,000 hits from people using google to search for pirate movies to download.

I only really set the server up for testing php scripts, etc.

[billou2k]

Cruncher: Got the same problem here with 11/01 ffdshow version. It doesnt crash with every codec but xvid is one of them. Going back to 05/01 version works fine.

[Leo 69]

ffdshow is now updated (20050113),weighted prediction seems to be supported but still there remains a BIG problem with b-frames in Recode's .mp4 :(

[CruNcher]

huh borks again ? the 5th jan build displayed b-frames in recode encodes nicely and yes i used 11th jan build trying the 13th now geez changes fast :D

B-frame are still working weighted prediction works too very nice but every b-frame now with activated OSD the OSD gets curupted looks funny :D

[Bluedan]

Opens multiple instances of ffdshow-video when I try to playback an matroska file containing DivX video with Core Media Player. With XVid it's the same. File plays fine though. Didn't alter any settings in ffdshow.

In the first place crashed player until I disabled raw video support.
No difference when using gabest mkv splitter or alternative from Haali.

[Leo 69]

@ CruNcher

What splitter are you using for playback? :confused:

[hellfred]

Originally posted by Bluedan
Opens multiple instances of ffdshow-video when I try to playback an matroska file containing DivX video with Core Media Player. With XVid it's the same. File plays fine though. Didn't alter any settings in ffdshow.

In the first place crashed player until I disabled raw video support.
No difference when using gabest mkv splitter or alternative from Haali.
Just a wild guess, two instances of ffdsow, one for decoding video, and one for audio? Test: Strip the audio out of one of your clips and retry. If ffdshows numer of instances wents down to one, you know what is up.

Hellfred

EDIT: If you have an avi-source, I would suggest VirtualDubMod to create a new file of just some hundred frames. Set video to direct stream copy and disable audio track via Stream -> Stream List. File should be ready to test just few minutes.

[filewalker]

Originally posted by Bluedan
Opens multiple instances of ffdshow-video when I try to playback an matroska file containing DivX video with Core Media Player. With XVid it's the same. File plays fine though. Didn't alter any settings in ffdshow.

Make sure that "only one" is set in:
ffdshow configuration-->Codecs (at the botton)

Cu

[celtic_druid]

Hmmm, looks like my compile just missed out on:
"hack to set correct frame timestamps for b frames with nero splitter"

[BlindWanderer]

I assume that fast rendering for the OSD doesn't keep a backup of the original frame; using the modified frame with the decoder. While having it turned on I noticed block movement containing OSD text. Could we get a tooltip about this for the check box?

[Yong]

Special SSE version, download here (http://cutka.szm.sk/files/ffdshow-20050112.exe)
:) :) :)
Anyone here may try to playing your lovely DVD with ffdshow(Decoding MPEG1/2 with libavcodec can't wotk, please change the MPEG1/2 decoder to libmpeg2),
see if it can work or not:p

[Bluedan]

Originally posted by filewalker
Make sure that "only one" is set in:
ffdshow configuration-->Codecs (at the botton)

Cu

A clean uninstall/install solved it. Sorry for the waste of letter soup here. :p

I have some kind of choppy playback now but A/V stays in synch. CPU consumption is ~35%.

Decoding AVC(in mp4) created with Recode.

:confused:

[Koti]

Just a big "thank you" to all the ffdshow dev's and compilers

ffdshow-20050117.exe (celtic druid's) plays (using nero's parser/splitter) recode created avc.mp4 files (all options enabled in nero encoding tools) perfectly!!!!! :D :D :D :D

[IgorC]

can anybody upload ffdshow last build to somewhere like http://www.rapidshare.de/en/main.html ?

i´m trying to get it 3´d time from http://celticdruid.no-ip.com/xvid/ffdshow/

Palikrovol thanks (se lo agradesco) i´ve got it in less than 10 min

[708145]

Another mirror is here (http://www.iti.uni-stuttgart.de/~bergmats/video_mirror/videotools/).

the latest build will be up as soon as I have it. CD is sloow.

bis besser,
Tobias

[Palikrovol]

Originally posted by IgorC
can anybody upload ffdshow last build to somewhere like http://www.rapidshare.de/en/main.html ?

i´m trying to get it 3´d time from http://celticdruid.no-ip.com/xvid/ffdshow/

Done.

Get it here http://rapidshare.de/files-en/371342/ffdshow-20050117.exe.html

regards

[CruNcher]

OSD B-frame problems are gone :)
and jep playback is very smooth with recode .mp4 now :)

[moonman]

You can download it from my mirror too.
You find the link in this article:
http://svartling.hopto.org/index.php?q=node/view/421


I hope the download link works for those that not have a streamload account.

[Sharktooth]

mirrored here too: http://ebola.gamersrevolt.it/celticdruid/ffdshow/

[Sirber]

I will mirror it too, once my server come back from the sig fault :angry:

[Sharktooth]

Mine is quite stable and fast :)
EDIT: I PMed C_D and offered him fast host for his binaries...

[Neo Neko]

Just use this URL from now on and everything should be fine. http://celticdruid.no-ip.com.nyud.net:8090/xvid/

[drjimmy526]

After switching from the 1/5 to the 1/17 build, I am getting an exception error AFTER closing MPC or Zoom Player when playing a clip that utilizes VSFilter to display subs (I'm not using ffdshow subtitle renderer). Anyone experiencing this?

[Yong]

The latest build of ffdshow-20050117.exe have problem with avisynth filter,
it crash for me(MS visual C++ runtime error) if i enter avisynth command the avisynth filter tab...
This problem was not appear at SSE version of ffdshow-20050112.exe.

[yaz]

Originally posted by Sharktooth
I PMed C_D and offered him fast host for his binaries... sharky ! u're already on my 'bless-forever' list but now u're the winning candidate for the upmost position, definitely :-)(especially if u'd mirror the mencoder builds too :-)))
the bests
y

[bill_baroud]

just a little question, it had been some times since i updated ffdshow, and i found there is a "ff_rar.dll" included ....
so the question is : why ffdshow use rar (de)compression ? for compressed vobsub subtitle ?

[Sharktooth]

Originally posted by Neo Neko
Just use this URL from now on and everything should be fine. http://celticdruid.no-ip.com.nyud.net:8090/xvid/
Internal server error when trying to download...

@yaz: well, i will mirror everything but only if i'm able to grab it from C_D's web... :)

[esby]

I made up a mirror on my website too:

http://esby.free.fr/CelticDruid/


esby

[Sharktooth]

Ok, i just talked with C_D and i will host a fileserver with all the new stuff from C_D.
All downloads from the C_D's website will be redirected to the new fileserver and it will be updated every 24h.
When it's ready you will see a major speed improvement:)

[rushin_911]

hi,
I would like to report what seems to be a bug in ffdshow version 20050117 which i downloaded from http://www.iti.uni-stuttgart.de/~bergmats/video_mirror/videotools/
In Warpsharp when choosing aWarpsharp with the blurmode set on high quality 3-pass and all the other parameters left on default the video stream's color turns to gray scale regardless of what chroma mode is set

[Sirber]

Originally posted by Sirber
I will mirror it too, once my server come back from the sig fault :angry: Mirorred in my download section.

[QQ]

is it only me, or is SPP deblocking terribly slow? if i allow auto strength, then as soon as it hits "best" or "next to best" movie starts lagging badly - and all this on 3.0ghz P4! :(

[DeeGee]

Yes, unfortunetly SPP deblocking is very slow. I usually use it at force and on "level" 2 (on the slider). And on very high resolution files I need to put it off, because of the humongous cpu requirements.

[akupenguin]

Originally posted by QQ
is it only me, or is SPP deblocking terribly slow? if i allow auto strength, then as soon as it hits "best" or "next to best" movie starts lagging badly SPP is very slow at high strengths, but if it lags on auto then that's a bug. I don't know about ffdshow, but in MPlayer, "auto" means the highest setting your cpu can keep up with (continuously adjusted), and it will never lag unless the fastest setting is still too slow.

[Yong]

Originally posted by rushin_911
hi,
I would like to report what seems to be a bug in ffdshow version 20050117 which i downloaded from http://www.iti.uni-stuttgart.de/~bergmats/video_mirror/videotools/
In Warpsharp when choosing aWarpsharp with the blurmode set on high quality 3-pass and all the other parameters left on default the video stream's color turns to gray scale regardless of what chroma mode is set

This problem i already reported to milan, hope this bug will fix at next release.;)

[rushin_911]

Originally posted by Yong
This problem i already reported to milan, hope this bug will fix at next release.;)
oh ok thx :)

[Irwin]

Why in OSD "Frame Type" is no QPel or GMC info? - It's only I,P or B - In 2004 versions ffdshow osd showing this info correct.

[Elias]

When will the ffdshow filter contain an mp4 splitter?

[QQ]

does it do better post-processing than the other two methods? what settings should i use for the other two, and should i use them together, or not?

[yaz]

@elias
ffdshow is (simply speaking) a 'codec pack' w/postprocessing. afaik, it does not contain any splitter.
the bests
y

[Elias]

Originally posted by yaz
@elias
ffdshow is (simply speaking) a 'codec pack' w/postprocessing. afaik, it does not contain any splitter.
the bests
y I just want to decode mp4 files without the need of installing 3ivx, nero or any other application than ffdshow.

[kurt]

you can decode easily avc-files with celtic_druids latest ffdshow build (without nero). enable only h264 in codec-list.

http://www.aziendeassociate.it/cd.asp?dir=/ffdshow

[Elias]

Originally posted by kurt
you can decode easily avc-files with celtic_druids latest ffdshow build (without nero). enable only h264 in codec-list.

http://www.aziendeassociate.it/cd.asp?dir=/ffdshow It is enabled, but what about ASP? I'm talking full mp4 compatability, not just AVC.

[celtic_druid]

It can decode ASP to, but it won't work for either in an mp4 container by itself.

[Elias]

Originally posted by celtic_druid
It can decode ASP to, but it won't work for either in an mp4 container by itself. I know, that's why I'm wondering when the ffdshow will have a built in mp4 splitter, or any other way to decode mp4 ASP by itself.

[Tommy Carrot]

Originally posted by Elias
I know, that's why I'm wondering when the ffdshow will have a built in mp4 splitter
Probably never, it's not ffdshow's job, as it's an encoder/decoder filter, nothing more. In my opinion adding even more features would make it too complex to maintain by only 1 developer.

[Elias]

Originally posted by Tommy Carrot
Probably never, it's not ffdshow's task, it's an encoder/decoder filter, nothing more. In my opinion even more features would make it too complex to maintain by only 1 developer. So I guess that mp4 won't have a decent way to decode without Nero+ffdshow, ever?

[bond]

Elias, use search and read the mp4 faq :rolleyes:

[stephanV]

Originally posted by Elias
So I guess that mp4 won't have a decent way to decode without Nero+ffdshow, ever?

well, write your own mp4 splitter :p

[Neo Neko]

Originally posted by Tommy Carrot
Probably never, it's not ffdshow's job, as it's an encoder/decoder filter, nothing more. In my opinion adding even more features would make it too complex to maintain by only 1 developer.

Not to make light of what millan is doing. While he may be the only ffdshow developer much of the code ffdshow uses was not written by him. Alot of it came from FFMpeg. FFMpeg also can read and write a ton of formats. Millan started with the video part for the original ffdshow. Since then he has added much of the audio portion as well. The next logical step is to isolate the format reading/writing capabilities and wrap then in directshow as well.

I want to repeat it. FFDshow is awesome software. But most FFDshow updates have more to do with FFMPEG updates than they do with ffdshow code. ;) It is only a matter of time. But patience is still required.

[RadicalEd]

The issue keeps popping up again and I must have mentioned it a dozen times by now, but once more: an ffdshow implementing lavf would kick ass :D

[Inc]

I installed CelticDruids latest ffdshow compile and still there's the bug in MakeAvis "avis" decoding under vfw environments.

Also the filtering in ffdshow doesnt seem to be aplied in vfw environments even if set in ffdhow vfw configuration.

another point, sorry if that already has been an issue in here, ... I cant set the priority slider in ffdshow anymore and its multiple ffdshow filter in the ds chain configuration :(

Thanks in advance and thanks for the compiles! :)

[midiboy]

Hi everyone !

I don´t know how old this bug in ffdshow is, at least a few years :D

The workaround, to manually set a registry key to put the video decoder in "PAL" mode is a nuisance if opening files from harddisc (for DVD´s it can be automated with AnyDVD etc.)

Now, the reason why I am bringing this up here now is ...

This bug has been in each and every ffdshow version for years except seemingly in the "ffdshow-20050105.exe" version. All the later versions have the problem again.

When I tried the version from January 5th 2005, I thought ... wow, they finally got rid of this bug ! Unforunately, the next version that came out didn´t work without setting those reg keys anymore !

So something had been done right with this one version but seemingly got lost again later on !

It would really be wonderful if this very old bug could be solved once and for all !

Thank you !
Bye,
Alex

[therealjoeblow]

Originally posted by Yong
The latest build of ffdshow-20050117.exe have problem with avisynth filter,
it crash for me(MS visual C++ runtime error) if i enter avisynth command the avisynth filter tab...
This problem was not appear at SSE version of ffdshow-20050112.exe.

I'm having problems with avisynth support too. Can you point me to a download location for ffdshow-20050112.exe please? C_D's site and all others I've found only have ffdshow-20050105.exe, ffdshow-20050113.exe and ffdshow-20050117.exe, no 12th.

[kurt]

Originally posted by Yong
Special SSE version, download here (http://cutka.szm.sk/files/ffdshow-20050112.exe)
:) :) :)
....

you mean that?

[Sirber]

Isn't the latest automatic CPU extention detection?

[celtic_druid]

20050122 AVISynth is working fine, at least here. I made sure of it before I put it up.

[Yong]

It is nice to stick with latest build by Celtic druid.

Here is some part of replies by Milan:

1.
> I tested it on Pentium 4 2GHz and for 704x288 AVI
> SPP at quality level 2
> consumed about 80% of CPU and playback was smooth.
>
> Automatic quality control is quite broken now, try
> to disable it. And probably
> it won't never work correctly with SPP, because
> there is a huge difference is
> processing time when SPP disabled and enabled even
> with lowest quality.

2.
aWarpSharp is known problem, but since I don't have source code of that
filter
it's very hard for me to fix it.

3.
DVD playback is still my priority for now, together with fixing bugs.

BTW I was thinking about dropping support for some encoder libraries.
For
example x264 has its own VFW interface now and XviD encoder interface
in ffdshow
lacks a lot of features compared to official XviD. I still want to
support
interesting libraries which don't have their own Windows interface such
as
theora or mpeg2enc, but I simply don't have time to duplicate others
work. And
of course, libavcodec is the main target, fortunately libavcodec
development is
almost stalled :)

I tried with the latest build by C_D, Avisynth filter in ffdshow also work fine for me ;)

[rushin_911]

2.
aWarpSharp is known problem, but since I don't have source code of that
filter
it's very hard for me to fix it.


aWarpSharp in the October 2004 version works perfectly... The lack of support for x264 bframes forced me to update to the new version (for anyone who would ask why I don't just stick with it ;))

[Yong]

Originally posted by rushin_911
aWarpSharp in the October 2004 version works perfectly... The lack of support for x264 bframes forced me to update to the new version (for anyone who would ask why I don't just stick with it ;))
I don't know why the aWarpsharp is so important for you,
But i think the ffdshow original warpsharp function might be same as aWarpsharp.

By the way, old version of ffdshow is very buggy,
start from the 20041012 version, i sent over 19 of bug reports to Milan already,
I think there will be more than 30 bugs inside the old version of ffdshow waiting you to trigger it:D

[iron2000]

The 20041012 works best for me so far.

The newer versions cause laggy playback for some Xvid files like the Talses of Phantasia subbed anime.

I use post post-procrssing and sharpen.

[acidsex]

Is there any eta on when ffdshow will support 2 pass encoding for x264?

[Sharktooth]

use x.264vfw instead of ffdshow.
its much better and updated everyday (or so).

[celtic_druid]

Originally posted by acidsex
Is there any eta on when ffdshow will support 2 pass encoding for x264?
Yeah, it has been and gone. ffdshow has supported 2 pass for I think over 1 month. I agree though, stick with x264vfw.

[hellfred]

Originally posted by acidsex
Is there any eta on when ffdshow will support 2 pass encoding for x264?
It is already possible, though not very straiggth forward. It was descriped in one of the threads here, but i cannot remember where exactly. If i remember right, do a normal one pass encode and trigger to write a stats file. Then run the second one by providing target Bitrate and enforce the use of the stats file.

Hellfred

[acidsex]

I wouldnt mind using x264vfw/ffdshow but I still cant figure out a way to mux my h.264 stream and he-aac into one file. i can no longer find the muxer I used for the Ateme beta so am I SOL or is there another way?

edit: the reason I tried ffdshow x264 is because my NLE (Vegas 5) seems much slower when exporting x264vfw then it does with ffdshow.

[rushin_911]

Originally posted by Yong
I don't know why the aWarpsharp is so important for you,
But i think the ffdshow original warpsharp function might be same as aWarpsharp.

Well for me with anime the combination of SPP deblocking and aWarpSharp produces very nice results :) It isn't something very important, but it would be nice if the bug were to be fixed :)

Anywayz thx :)

[celtic_druid]

Seems to be fixed now. Well at least it isn't producing B&W output.

[MSlv]

where can I get the latest ffdshow from? I couldn't find a link in the whole topic...

[celtic_druid]

I haven't put a new build up yet (the one where awarpsharp seems to be fixed).

[Yong]

Originally posted by iron2000
The 20041012 works best for me so far....
Are you sure?
Bug in old version of ffdshow:
-Postprocessing filter crash if "right half only" is selected
-Resample filter crash f set the value higher or lower than 44100Hz
-ffdshow audio decoder crash if use the following output sample format: 16/24/32bit integer
-ffdshow audio decoder crash with some WMA file.
-crash if visualization: motion vector is selected.
-Tremor decoder have distorsion while playing.
-many bugs in Mixer filter
-bug in DScaler filter
-Raw RGB24 decoder bug
-libavcodec with swap the uv if decoding AVIzlib
-crash if decoding RV10
-If activating the [Quantizer noise shaping] mode will crash ffdshow
-bugs inside the FIR filter, Deinterlacing filter, Shapen filter, Avisynth filter, Resize & aspect setting, crop filter
-crash if decoding MPEG1/2 video
-crash if use some filter with another filter...
-many small bugs inside the ffvfw...

[Yong]

Originally posted by celtic_druid
Seems to be fixed now. Well at least it isn't producing B&W output.
It producing Blue&Green output on my computer with old version...:p

[MSlv]

Originally posted by celtic_druid
I haven't put a new build up yet (the one where awarpsharp seems to be fixed).
No, I couldn't find a link to your ffdshow builds. But someone sent me a PM with the link already.

[therealjoeblow]

Originally posted by kurt
you mean that?

Thanks.

[ChronoReverse]

I'm curious if anyone else is having a similar problem with the builds of ffdshow after the "20050105" build. When I do a lot of search jumps (i.e., I click around on the search bar and move around) ffdshow occasionally crashes (it locks up and I have to kill the process).

It didn't do that with the older build so I wonder if it was something changed.

[Yong]

Originally posted by ChronoReverse
I'm curious if anyone else is having a similar problem with the builds of ffdshow after the "20050105" build. When I do a lot of search jumps (i.e., I click around on the search bar and move around) ffdshow occasionally crashes (it locks up and I have to kill the process).

It didn't do that with the older build so I wonder if it was something changed.

Sometime could be a splitter, or paser filter problem...
what container you use? What video clips are you playing?

[niamh]

I can't get ffvdub to preview real time in the 20050122 build, and ffavisynth won't work for me, I reverted to 0117, and that works fine...

[ChronoReverse]

Originally posted by Yong
Sometime could be a splitter, or paser filter problem...
what container you use? What video clips are you playing?

I've tried with both AVI, Matrovska and MP4 (codecs, xvid, wmv3, xvid respectively) and it occurs with the latest builds.

Build "20050105" doesn't have this problem (but since it doesn't like to decode x64 stuff very well I switched to a newer build. Do you think I could install "20050105" and copy the ff_x264.dll file over?

Also maybe related is that the seeking isn't as snappy with the newer builds too. It takes a few more seconds to resync than the "20050105" build.

[celtic_druid]

ff_x264.dll is only used for encoding.

[ChronoReverse]

Oops, ffdshow uses libavcodec doesn't it? Completely forgot. Oh well, I'll live with it unless it's discovered to be a bug XD

[drjimmy526]

I haven't had any crashes, but I noticed too that the seeking is much slower with builds later than 20050105.

[Yong]

Recently Milan make quite alot of change ffdshow structure to allow it for DVD decoding,
may be somethings is borked,
eg, now the ffdshow libavcodec was unable to play MPEG1 video smoothly...
and some old build was unable to decode MPEG1/2 video.

ffdshow crash for me
(with unknown reason, mostly avisynth filter problem)
around 5 times per day:p
Thanks to the ffdshow developers,
now the recent build is very stable;)

[timeismoney]

Now play DVD via libmpeg2 is OK, seeking is fine. But may fail in Nav, maybe milan is working on it :p

[ChronoReverse]

Might as well report this (even though it may be a bug with something else or even my own mistake).

x264 encode with 1 Bframe (revision 109).
Muxed into mp4 using mp4creator (optimized and interleaved with mp3 audio track).
Moonlight splitter.
Tested with ZoomPlayer and Media Player Classic.

Attempting to use the seek bar results in a crash where the audio continues to play.

Doesn't occur with the AVI version.
Doesn't occur with XviD mp4 version.

http://img200.exs.cx/img200/2848/error5hf.gif

[Axed]

Hey just a small problem with 22/01/05 version of ffdshow. If i set mp2 audio to use libmad, it crashes with an error. 17/01/05 worked fine using libmad.

Im using Windows 98SE.

MPLAYERC caused an invalid page fault in
module MSVCR71.DLL at 017f:7c3429c1.

[yaz]

@chronoreverse
moonlight splitter has some (serious) seeking issues. this was reported in many different threads
the bests
y

[celtic_druid]

@Axed, no problems here, so it could be a 98 issue, although I also I am using a newer compile, MSVC7.1 again though. 17-01 would have been MSVC6/ICL7 I think.

Anyway as far as I can recall libmad code in ffdshow hasn't changed in a fair while, so you can just use libmad from 17-01 with 22-01 or 01-02.

[vlada]

Hi,
I'd like to report some problems with FFDShow 2005-01-12 (SSE). I already noticed the issues in previous builds.
I checked, I'd like to have a tray icon displayed. It works, but after starting a new movie, a new icon will appear. Then after some time, the whole taskbar is full of FFDShow icons. They dissapear after closing my media player (tested with ViPlay and MPC).
In MPC I added FFDShow as RAW filter, so I have 2 icons for each movie. I believe that it only should be loaded once, but I'm not sure if it isn't rather a bug in MPC.
Also I noticed some problems playing MPEG-1 video with LIBAV codec. Sometimes it works, but usually I only get black screen. Decoding using LIBMPEG2 works fine. I didn't try decoding MPEG-2. Btw. How do you guys make FFDShow to decode DVDs? What player do you use and how do you set it up?
Vlada

[Yong]

@vlada:
Milan said the MPEG1 decoding with libmpeg2/libavcodec problems was fixed :D
The date of mail reply:
Mon, 31 Jan 2005 19:51:59 +0100

But still couldn't find any new ffdshow build in the Celtic_Druid website:p

Hi,
I'd like to report some problems with FFDShow 2005-01-12 (SSE). I already noticed the issues in previous builds.
I checked, I'd like to have a tray icon displayed. It works, but after starting a new movie, a new icon will appear. Then after some time, the whole taskbar is full of FFDShow icons. They dissapear after closing my media player (tested with ViPlay and MPC).
I also have this problem too:p
I'll try to check this problem

BTW,
Milan said his already added support for decoding MKV warpped FLAC format with Haali matroska splitter:D

[kurt]

Originally posted by Yong
Milan said his already added support for decoding MKV warpped FLAC format with Haali matroska splitter:D

nice to hear! so lets wait for for a new c_d build :p

[celtic_druid]

Build has been up for a while now (no cvs changes since though). Just have to wait for someone to mirror it.

Didn't do too much testing, but as far as I could tell.
AVISynth still works fine in ffdshow.
MAKEAVIS now works fine via VFW.
awarpsharp working fine.
Didn't check MPEG1 or 2.

[Sharktooth]

New build is up in the main mirror.

[ChronoReverse]

Originally posted by yaz
@chronoreverse
moonlight splitter has some (serious) seeking issues. this was reported in many different threads
the bests
y

Only with H264 eh? Too bad the 3ivx doesn't work with AVC.

Oh well, I'll live with avi and mkv for now

[igor1st]

Originally posted by celtic_druid
Build has been up for a while now (no cvs changes since though). Didn't check MPEG1 or 2.
All builds newer then 20050113 crashes on DVD menus. But decoding mpeg2 with libmpeg2 works without problems.

[Leak]

Originally posted by vlada
I checked, I'd like to have a tray icon displayed. It works, but after starting a new movie, a new icon will appear. Then after some time, the whole taskbar is full of FFDShow icons. They dissapear after closing my media player (tested with ViPlay and MPC).

Make sure you have "Make DirectShow graph available to graphedit" unchecked on the "Info & debug" page if you're using MPC, as it does the same - and if it is done twice the graph isn't freed twice until the program is closed, leading to tray-icon buildup because the old graph lingers aroudn, or even worse resulting in 2 graphs playing simultaneously as it was in my case...

np: Starchboy - Rangoon (Clicks & Cuts 4)

[Axed]

Little extra to add to my problems with the new ffdshow compiles, i cant use anything past 17/01/05 because it just wont run at anywhere near the proper speed. Video stutters using libavcodec or libmpeg (for mpegs of course).

Before anyone says it. No, upgrading from Windows 98 is not an option because of hardware and software issues that cannot be resolved until i can afford replacements (not cheap).

[iron2000]

Originally posted by Yong
Are you sure?
Bug in old version of ffdshow:
-Postprocessing filter crash if "right half only" is selected
-Resample filter crash f set the value higher or lower than 44100Hz
-ffdshow audio decoder crash if use the following output sample format: 16/24/32bit integer
-ffdshow audio decoder crash with some WMA file.
-crash if visualization: motion vector is selected.
-Tremor decoder have distorsion while playing.
-many bugs in Mixer filter
-bug in DScaler filter
-Raw RGB24 decoder bug
-libavcodec with swap the uv if decoding AVIzlib
-crash if decoding RV10
-If activating the [Quantizer noise shaping] mode will crash ffdshow
-bugs inside the FIR filter, Deinterlacing filter, Shapen filter, Avisynth filter, Resize & aspect setting, crop filter
-crash if decoding MPEG1/2 video
-crash if use some filter with another filter...
-many small bugs inside the ffvfw...


Hmm..

So these causes those errors I got in BSplayer?
I get an error that closes the player when I use the same instance to play a series of a few video files.

Anyway I'm happy if the file plays with no lag with Auto Quality Control full screen postprocessing and xsharpen.

I only use those 2.
Most new builds this year lags with those 2 on, especially when playing Xvid files.

Is there an option to not install the audio decoder component?
I'm using Fraunhofer's decoder as it sounds better to me.
Not using the audio decoder for any format so the errors there don't affect me.

[niamh]

I couldn't access the vfw config in the 2005-02-01 version of fdshow ("msvcr71 cannot be found" of all things, Or plain nothing happens), until I copied said msvcr71 from the ffdshow folder to system32. 2005-01-17 worked fine with the bundled dll.

[oddball]

The audio decoding portion of ffdshow really needs work. I cannot get it to playback 5.1 AAC to 5.1 AC3 SPDIF output without it stuttering and causing video playback to stutter at the same time and really really slow playback too.

Having to use AC3Filter for that instead.

[Yong]

Originally posted by iron2000
Hmm..

So these causes those errors I got in BSplayer?
I get an error that closes the player when I use the same instance to play a series of a few video files.

Anyway I'm happy if the file plays with no lag with Auto Quality Control full screen postprocessing and xsharpen.

I only use those 2.
Most new builds this year lags with those 2 on, especially when playing Xvid files.

Explained by Milan, why the recent build will lag with auto quality control or SPP:
There were some new regressions introduced by latest changes mostly
related
to frame skipping and automatic postprocessing strenght. ffdshow now
behaves quite badly when CPU usage it too high even for short period
of time - it skips a lot of frames.

Is there an option to not install the audio decoder component?
I'm using Fraunhofer's decoder as it sounds better to me.
Not using the audio decoder for any format so the errors there don't affect me.

During installation of ffdshow, don't check all check box for audio part.:)

[Bluedan]

Just to add: ffdshow-20050201 build also consumes to many CPU cycles when decoding DivX 3 (M$MPEG hack) clip with 640x352 resolution via libavcodec.
On a PIII@1GHz video stutters and goes up to 100% CPU usage.
With Nero Video decoder which is at hand on this machine it only takes 45% max.
No postprocessing in ffdshow used.

I wonder if latest additions to libavcodec caused this.

[celtic_druid]

Just tried a 512x384 DivX3/mp3 encode here and if anything Nero's decoder uses more CPU, although ffdshow seems to fluctuate more. With Nero's decoder CPU usage was 14-17% and ffdshow was 9-20%.

Athlon-64 3500+, Win2k SP4. ffdshow-20050204.

[marcellus]

Hi, I just tried ffdshow-20050204 and comparing to ffdshow-20050201 is really slower. I use it as my tv capture codec (encoding to mpeg2) and now, with the same settings and all it drops a lot of frames, so it's unusable. With ffdshow-20050201 the CPU varies between 79% and 97-100% but now is permanently at 100%. I use as filters crop, resize&borders, deinterlace with kerneldeint and denoise3d so one of these or the mpeg2 encoding itself eats a lot more CPU. I'm on AthlonXP.
Back to ffdshow-20050201 for now...

[namchik]

There's a bug I had also noticed with older versions of FFDShow. I've got a DivX 3 movie with resolution 512 x 277 (It wasn't me who had encoded it :) )... So when FFDshow-20050204 is chosen to handle div3, simply clicking on this file makes Explorer crash. When FFDShow is disabled for div3, the file is normally played with Divx 5 decoder and Nero Video Decoder as well.
I know that 277 is out of DivX 3 standard, but those 2 decoders could normally play the movie...

[Bluedan]

@celtic_druid

Weird. I got these files from Mosu's sample page (http://www.bunkus.org/videotools/mkvtoolnix/samples/) linked from Matroska site.
I don't think there's something special about their creation.
Well, I'm often mistaken, but I cannot recall having done anything wrong with codec installation.
Cannot test other PCs ATM.

[yaz]

got the same prob as oddball has. no audio w/the latest builds (aac, ac3, mp3 tested). audio in 0122 stiil worx fine.
Originally posted by Yong

During installation of ffdshow, don't check all check box for audio part.:) what do u mean by this? i don't tick anything during the installation (as i can do it later on:-)
thx
y

[Yong]

Originally posted by yaz
got the same prob as oddball has. no audio w/the latest builds (aac, ac3, mp3 tested). audio in 0122 stiil worx fine.
what do u mean by this? i don't tick anything during the installation (as i can do it later on:-)
thx
y
Decoding ffdshow with DVD?
May be you need to write a bug report to Milan ;)
(i couldn't test it because i don't have DVD drive :()

what do u mean by this? i don't tick anything during the installation (as i can do it later on:-)
thx
y
Sure you can :)

[yaz]

Originally posted by Yong
Decoding ffdshow with DVD? sure, i've made it many times but now ffdshow decoding is somewhat slower, so i observe some jerkiness on playback.
the bests
y

[gunnerhkjp]

Hi.

I'm trying to use the perspective correction filter in ffdshow, but I get many black lines in the video. Is this filter a work in progress?

I use this filter to compensate for a projecter that needs to be placed slightly to the left of the screen centre.

Also, it seems if I use the "resize&aspect setting" filter and the perspective correction filter together, the media player will crash.
I have confirmed this with both winamp and media player classic.

I would also appreciate greatly if anyone can recommend a way to do perspective correction with or without ffdshow without the artifacts showing on the screen.

I'm using ffdshow version Feb 1 2005 18:12:38

[Didée]

No, perspective correction ever has worked like that. You can check "source rectangle", then the process is inverted (image not squeezed inwards, but torn outwards). This method is free of artefacts. You'll loose a little of the picture, but that can be compensated by adding some black borders beforehand, eg. by resizing to a bigger resolution with resize method "none".

[gunnerhkjp]

Originally posted by Didée
No, perspective correction ever has worked like that. You can check "source rectangle", then the process is inverted (image not squeezed inwards, but torn outwards). This method is free of artefacts. You'll loose a little of the picture, but that can be compensated by adding some black borders beforehand, eg. by resizing to a bigger resolution with resize method "none".

Thanks a lot. That did the trick.
It is amazing how versatile ffdshow is.

[kurt]

is there a sse version of ffdshow from 20050201 or later?
i would need it for proper playback of avc with my secondary pc (pIII, 1000mhz, 256MB) in the living room... :)

[Sirber]

Latest FFDShow are in "autodetect" mode. See codec info :)

[gunnerhkjp]

Hello again.

After experimenting with perspective correction filter (which works very well after I activated source rectangle) I decided change back to the original setting to watch on my computer monitor.

I turned off perspective correction and turned on resize&aspect setting (to resize the video to my LCD screen size) but now the filter seems to be turned off no matter what I do.

Is this a common occurance?

[gunnerhkjp]

Hi.

I solved the above by uninstalling and installing again.

[xneo]

ffdshow-20050216 was uploaded here. (http://www.aziendeassociate.it/cd.asp?dir=/ffdshow)
and... MPC,too.
These files build by Celtic Druid.:p

[igor1st]

20050216 build still crushs on DVD-menus.
Milan is still working on DVD support or no? Too much time has passed from 20050113 build...

[midiboy]

By the way .... (DVD menus) ...

just found a bug in ffdshow that completely prevents navigating DVD´s.

I am using the raw video processor as a post processing filter for resizing DVD content. Some DVD menues are completely broken, meaning, there is no way to navigate them either through mouse or cursor keys.

This happens with NVIDIA, Dscaler and Elecard decoders and all renderers (in ZoomPlayer). I was able to narrow it down to the use of ffdshow resize. The algorithm used and the resize values do not seem to make any difference (using 2x input rez, lanczos 2)

Some DVD menues are fine still but some are completely broken. One example of a DVD with broken menues is the DVE disc (Digital Video essentials)

There is a thread in the AVS forum about this already ... see here (http://www.avsforum.com/avs-vb/showthread.php?s=&postid=5197986#post5197986)

Would it be possible to fix this ?

Thanks a lot,
Alex

[marcellus]

I downloaded ffdshow 20050216 and guess what: is slow. I think I managed to find the cause: is resize module. Even if it's check box isn't checked, is enough to set borders at anything else than 0 - the resize module enters in action and it slows everything down. I'm curious if anybody else finds the same symptoms.

[xneo]

@marcellus
I don't have that problem. not slow,,, everything OK.:p
athlon64 3000+ xpProSP2

#edit
oops..:eek:
really slow...
before i did not check filters

[tmpy222]

Yes, I also find 20050216 version is quite slower than 20050201 version. It's either resize or subtitle filter that is slowing things down.

Thanks.

[Yong]

I found the 20050216 build of ffdshow,
postprocessing filter option:
level fix, doesn't work anymore,
compiler issue?

[loni_blues]

The 20050216 seems to be missing msvcr71.dll. I cannot find the file anywhere in my system.

[hhsg]

Hi !

I have a problem with ffdshow 20050216 when i seek/ff in MPC, no problem with nero decoder

Error:

Assertion failed:
File: libavcodec\h264.c
Line: 2190
Expression: pic->data[0]


Thanks.

[bond]

Originally posted by hhsg
I have a problem with ffdshow 20050216 when i seek/ff in MPC, no problem with nero decoder

Error:

Assertion failed:
File: libavcodec\h264.c
Line: 2190
Expression: pic->data[0] what did you try to play?

[hhsg]

My clip encoded by VDM + X264VFW rev133

my setting:
-640x480
-2 pass @1000kbps
-h264 default but 5 ref frame

[bond]

can you upload a small sample that shows that behaviour somewhere plz :)

[timeismoney]

I think CVS 20050222 has fixed it

I met the issue when using 20050216, too. Now I compiled 20050222, it's all OK when seeking AVC media.

[saratoga]

Is there a compiled copy of 20050222?

I'm running the Feb-17 build and it crashes horribly (shows first frame and then stops playing, if I try and seek WMP just crashes) with all the AVC samples on http://www.videosoftinc.com/videoclips.html

[bond]

Originally posted by saratoga
I'm running the Feb-17 build and it crashes horribly (shows first frame and then stops playing, if I try and seek WMP just crashes) with all the AVC samples on http://www.videosoftinc.com/videoclips.htmlthese are all not compliant as they use some sort of packed bitstream

read here (http://forum.doom9.org/showthread.php?s=&threadid=90282) how to get rid of it

[Palikrovol]

In the FFdshow webpage http://ffdshow.sourceforge.net/tikiwiki/ can be found daily builds
http://ffdshow.sourceforge.net/tikiwiki/tiki-index.php?page=Getting+ffdshow

[Sharktooth]

Originally posted by Palikrovol
In the FFdshow webpage http://ffdshow.sourceforge.net/tikiwiki/ can be found daily builds
http://ffdshow.sourceforge.net/tikiwiki/tiki-index.php?page=Getting+ffdshow
Nice!

[Soulhunter]

http://cutka.szm.sk/files/ffdshow-20050222-sse.exe

Many small cleanups, nothing grounbreaking today:

more static const ints
changed some enums to static const int
REG_OP_N and REG_OP_S simplifying
updated x264
don't honour changes to "allow always extensible" during playback
possiblity to get list of options for few audio decoder parameters
removed i64 and u64 macros
updated swscaler
updated libavcodec
headers
renamed Debug configuration to Debug ICL in dsp files
removed Release configuration from dsp projects
optional high quality YV12->RGB conversion via Avisynth routines
fixed sometimes wrong timestamps generation


Thanks... http://img85.exs.cx/img85/6388/xyxthumbs1kf.gif (http://www.imageshack.us)


Bye

[marcellus]

This build (ffdshow-20050222-sse) isn't slow anymore for me, when I activate resize/borders.:cool::)

[oddball]

What I would like to see in the audio options of ffdshow is the option to send only 5.1 AAC/Ogg etc to the AC3 to SPDIF output. At present if you tick AC3 it sends 'ALL' audio to it. This is not the ideal for me. Why add extra CPU use to audio streams that do not need it (I.E. converting to AC3 on the fly means CPU extra CPU usage)?

Ideally the option to send to AC3 should be on a dropdown on the codecs menu of ffdshow so you can select which codecs to send to AC3 and which type (I.E. option to set 2.0 as untouched and 5.1 as converted).

[oddball]

(ffdshow-20050222-sse) This version is very slow when I activate AC3 and resample to 48000. Previous builds did not use as much CPU or jerk like this version does.

[hhsg]

Originally posted by bond
what did you try to play?

Playing is OK , just crash when seeking (H264 clip) in Media Player Classic


Error in ffdshow20050216:
Assertion failed:
File: libavcodec\h264.c
Line: 2190
Expression: pic->data[0]


PS: 20050222_sse : still crash :(

[Irwin]

Maybe someone resolve this very big bug:
http://forum.doom9.org/showthread.php?s=&threadid=90440

[Sharktooth]

Remember there are daily builds here: http://ffdshow.sourceforge.net/tikiwiki/tiki-index.php?page=Getting+ffdshow

[RadicalEd]

Not for those of us with Athlons :/

[SeeMoreDigital]

Originally posted by oddball
What I would like to see in the audio options of ffdshow is the option to send only 5.1 AAC/Ogg etc to the AC3 to SPDIF output. At present if you tick AC3 it sends 'ALL' audio to it. This is not the ideal for me. Why add extra CPU use to audio streams that do not need it (I.E. converting to AC3 on the fly means CPU extra CPU usage)?

Ideally the option to send to AC3 should be on a dropdown on the codecs menu of ffdshow so you can select which codecs to send to AC3 and which type (I.E. option to set 2.0 as untouched and 5.1 as converted). What I would like, is for FFdshow to offer a totally separate "AC3 Transcoding" application.

And in order to keep the application "very simple" to use (and small in build size) I think it should offer very basic AC3 transcoding options. So no need to include stuff like, channel mapping, noise reduction, mixers, crossovers, separate output and equaliser level adjustment, etc, etc... Like I say... the "lighter" the better.

As usual, I whipped up a prototype GUI. How about something like this: -

http://img239.exs.cx/img239/9096/audiofilterproposal3jx.gif


Cheers

[marcellus]

Originally posted by RadicalEd
Not for those of us with Athlons :/
On my AthlonXP is working fine. Or maybe do you mean non SSE Athlons?

[llamameat]

are there any versions of ffdshow that don't shift colors slightly towards green when using sinc resize? I recall this bug having been around for as long as ffdshow, any reason it hasn't been resolved yet?

[DeathTheSheep]

The most recent ffdshow does not decode aac audio streams muxed into AVIs. When opened with graphedit, a message exclaiming an undecipherable stream ensued.

Upon reinstall, the occurance again transpired. However, attempts to reinstall a prior version of ffdshow (dated in January) proved to be successful in playing the muxed aac-he stream.

Nearly every video I now create is using AAC and x264 in an AVI, and the lack of ability to decode is a terrible thing, especially since I need the most recent ffdshow package to encode as well.

I believe this is a bug, and a terrible one at that.

[Bluedan]

Why don't you shift to matroska container instead?
AAC as well as x264 in AVI is a hack anyway I believe.

[oddball]

Video playback on any version of ffdshow since the Athos builds has never been very smooth playback wise with me. I get more frame jerks and judders using the liba decoder than the XviD decoder. I have everything set to XviD decoder in ffdshow because of this.

[yaz]

Originally posted by DeathTheSheep
The most recent ffdshow does not decode aac audio streams ... read this (http://forum.doom9.org/showthread.php?s=&threadid=90478) it helped me.
the bests
y

[Didée]

Originally posted by oddball
Video playback on any version of ffdshow since the Athos builds has never been very smooth playback wise with me. I get more frame jerks and judders using the liba decoder than the XviD decoder. I have everything set to XviD decoder in ffdshow because of this.
That's the sad truth. But btw, even using XviD decoder (versions), I'm not fully sure about past-1.0.3 builds. Every now and then, there are jerks. Mostly minor, but they _are_ there. And it's not a setup problem - older encodings done with pre- and past-1.0.0 alpha builds play smooth as silk. Makes me think of gremlins sometimes.

[Sirber]

never had that problem :confused:

[Sirber]

Originally posted by SeeMoreDigital
What I would like, is for FFdshow to offer a totally separate "AC3 Transcoding" application.

And in order to keep the application "very simple" to use (and small in build size) I think it should offer very basic AC3 transcoding options. So no need to include stuff like, channel mapping, noise reduction, mixers, crossovers, separate output and equaliser level adjustment, etc, etc... Like I say... the "lighter" the better.

As usual, I whipped up a prototype GUI. How about something like this: -

http://img239.exs.cx/img239/9096/audiofilterproposal3jx.gif


Cheers Could be cool to have also speaker setup (IE 2.0, 5.1) :)

[Esc]

Possible bug found.
When installing ffdshow-20050301-sse.exe I have unchecked DivX 5 from the list of supported codecs. But when I opened Video Decoder Configuration window I found it was assigned 'libavcodec' value anyway.

[marcellus]

Great news! The resize/borders module was redesigned and now one can choose wether he wants inside (like ffdshow used to do) or outside (like AviSynth) borders. And the size of borders can be set in pixels not only in percents!
Thanks, Milan!

[Sirber]

Originally posted by Esc
Possible bug found.
When installing ffdshow-20050301-sse.exe I have unchecked DivX 5 from the list of supported codecs. But when I opened Video Decoder Configuration window I found it was assigned 'libavcodec' value anyway. Where can I download?

[MarkCoolio]

@Sirber:

here: http://ffdshow.sourceforge.net/tikiwiki/tiki-index.php?page=Getting+ffdshow

[Sirber]

Thanks! :D

[tmpy222]

Originally posted by Didée
That's the sad truth. But btw, even using XviD decoder (versions), I'm not fully sure about past-1.0.3 builds. Every now and then, there are jerks. Mostly minor, but they _are_ there. And it's not a setup problem - older encodings done with pre- and past-1.0.0 alpha builds play smooth as silk. Makes me think of gremlins sometimes.


If the playback isn't smooth, turn on "use overlay mixer" in output.

[DeathTheSheep]

Hm... That's funny... after I rip them out of the AVI, the AAC streams play fine with ffdshow. Somehow, just inside the new version, they won't get out of the AVI.

[tmpy222]

Hi,
Latest 20050302-sse version is having problem adding inside vertical borders (actually they are horizontal borders. that's another story.) when using percentage...
When switching to pixels, it works fine.

Thanks.

[Elias]

Will there ever be a bitrate mode in 2pass MPEG-4 mode encoding for ffdshow? Oh and when is the next official release?

[filewalker]

Originally posted by DeathTheSheep
The most recent ffdshow does not decode aac audio streams muxed into AVIs. When opened with graphedit, a message exclaiming an undecipherable stream ensued.
Try the 03.03.2005 build...I had problems to decode AAC from MKV, too (pin connection error in Graphedit). This build solved my problem and now ffdshow can decode both AVC and AAC very well. :)

Cu filewalker

[lithoc]

Wow!!
There's fixes in libavcodec

"H.264: decode arbitrary frame orders and allow B-frames as references"

Can't wait for new ffdshow
:cool:

[DeathTheSheep]

Try the 03.03.2005 build
Why thank you! :)

Well, that sure did solve my problem. Paging through the list of vfw codecs, I happened upon Theora by accident and wondered "They still have that hunk of junk?"

Woah, theora's looking a heck of a lot better than it did 2 years ago. Man, the speed is more than 50x more than before, the compression sure is a heck of a lot better than I expected, and the smoothness/sharpness meter was just what I was looking for.

Too dang bad there's no ffdshow/dshow based decoder for Windows CE, Microsoft Pocket PC, or Windows Mobile. Dang, I'm not sure a heck of a lot of people are looking to a Theora decoder, but there's a decoder for everything else (MPEGs, ASP, H261/3, Divx3, WMV, Realvideo--the decoder sucks though, upcoming VP6/7, heck even stuff like VP3 and MJPEG are implemented).

Wouldn't it be conveniant if there was an ffdshow type decoder for these emerging OS's? Albeit impossible, but even videolan was ported at one point (very old version).

Cheers.

[Sirber]

What's sad with videolan is no realvideo support :(

[lithoc]

Originally posted by Sirber
What's sad with videolan is no realvideo support :(

That's because VLC entirely only depend on libavcodec for codec support

But libavcodec do have realvideo support but it's very old version

:)

[Sirber]

That's cool in a way (libavcodec), but my main anime database is in RMVB... :(

[Elias]

Originally posted by Sirber
That's cool in a way (libavcodec), but my main anime database is in RMVB... :( Well suite your self for choosing a proprietary codec like real :D Stick to MPEG-4 or MPEG-2, or read the g00z capture bible :)

[Sharktooth]

Real is really good for anime. He made his choice and i think he wont turn back. Some time ago i choosen VP6 + Vorbis for backing up my movies (even 1 crappy HD WMV DVD i could only watch on my PC) and i was more than happy i could use xbox media center to play them back.
Dont think open standards are the best for everything. There could be proprietary codecs that do the job better (look at how light is VP6 to decode for example...).

[Sirber]

:stupid: :goodpost:

:D

[Elias]

Originally posted by Sharktooth
Real is really good for anime. He made his choice and i think he wont turn back. Some time ago i choosen VP6 + Vorbis for backing up my movies (even 1 crappy HD WMV DVD i could only watch on my PC) and i was more than happy i could use xbox media center to play them back.
Dont think open standards are the best for everything. There could be proprietary codecs that do the job better (look at how light is VP6 to decode for example...). Well, I'd never choose a proprietary codec. I don't trust them for one second. By the way, check this: http://slashdot.org/comments.pl?sid=89851&cid=7759569

[Sharktooth]

LOL, that's why i use Realalternative :D

[Sirber]

RealAlternative is crap. Better use Satsuki Decoder Pack.

Anyway, let's not pollute ffdshow thread :devil:

[MSlv]

I can't play x264 in mp4 container. I used AutoAC to do a short music video and I can't play the final .mp4 file. I tried videolan but it plays jerky. I can only play in Windows media player 9, but not from the begining, I have to move the slider to the left to have video. I used mkvmerge, but it showed an error saying that the video doesn't begin with a keyframe and then something about b-frames.

Does ffdshow decode x264 in mp4 container or not?

[Elias]

Originally posted by MSlv
I can't play x264 in mp4 container. I used AutoAC to do a short music video and I can't play the final .mp4 file. I tried videolan but it plays jerky. I can only play in Windows media player 9, but not from the begining, I have to move the slider to the left to have video. I used mkvmerge, but it showed an error saying that the video doesn't begin with a keyframe and then something about b-frames.

Does ffdshow decode x264 in mp4 container or not? I think so, it should because ffdshow decodes h.264 or are you using an old version of ffdshow? Maybe x264 isn't MPEG-4 ISO compliant?

[MSlv]

I/m using the latest ffdshow (2005-03-03)

[mezzanine]

Originally posted by MSlv
I/m using the latest ffdshow (2005-03-03)

Try Osmo4
http://www.aziendeassociate.it/cd.asp?dir=/gpac

[MSlv]

it's missing a dll - libmmd.dll.

[bond]

regarding videolan:
1) dont use the latest official version, its totally outdated
2) if you use a videolan nightly build, i assume that it is still not updated to the latest ffmpeg avc decoder version

x264/ffmpeg is developing so fast, its hard to follow even for people who know the stuff :D

[Elias]

Originally posted by bond
regarding videolan:
1) dont use the latest official version, its totally outdated
2) if you use a videolan nightly build, i assume that it is still not updated to the latest ffmpeg avc decoder version

x264/ffmpeg is developing so fast, its hard to follow even for people who know the stuff :D Well that's a good thing :) Speaking of the videolan nightly build, is it possible to find changelogs for it?

[celtic_druid]

Originally posted by MSlv
it's missing a dll - libmmd.dll.

Yeah, I was checking that out the other day and figured out that the reason was that I had obviously compiled js32.dll with ICL. None of the actual gpac files should require libmmd, but most depend on js32.dll. I probably should put a new compile up anyway.

[oddball]

Anyone know why AC3Filter can pass 44Khz AAC to SPDIF whilst ffdshow's has to be upsampled to 48Khz?

[bob0r]

Originally posted by celtic_druid
Yeah, I was checking that out the other day and figured out that the reason was that I had obviously compiled js32.dll with ICL. None of the actual gpac files should require libmmd, but most depend on js32.dll. I probably should put a new compile up anyway.

You should indeed :D

A good ffdshow version without any extra files needed, that's what we need, and that's what i been waiting for, for almost a month :eek:

I want to put in on my website, but there seems to be a curse on me when it comes to ffdshow :devil:

[celtic_druid]

New version of GPAC already went up yesterday.
http://www.aziendeassociate.it/cd.asp?dir=/gpac/dev

[bob0r]

But we (I) need a good new version of ffdshow, thats what this thread is about :p

[celtic_druid]

Might be the title of the thread but...

mezzanine said "Try Osmo4"
MSlv, "it's (Osmo4) missing a dll - libmmd.dll."
me, "I know I will do a new compile (of gpac)"
bob0r, "You should indeed (do a new gpac compile)"

ffdshow hasn't really been updated since Milan's last build. However if I have time I will put up a new build anyway.

[Yong]

Here's reason why Milan haven't put up a new build:
I'm very busy these days and not only by developing ffdshow, where
I'm doing quite big internal changes to allow more flexible filters design
- for example to have subtitles from two displayed at once. It sounds simple,
but to do this properly I had to extensively modify some parts of ffdshow which are there from the beginning and weren't designed with such flexibility in mind.
good news,
ffdshow will be able to decode TTA audio format (http://www.true-audio.com/) at next build, i think:)

[Palikrovol]

problems with build 20050312 decoding H.264 (Nero Recode).
This build can't decode it.

I'm using MPC and when opening the movie, it keeps in 00:00 minutes (i have press 'play' :)). This happens with some movies.

With this one http://multimediacom.free.fr/Video/NDAVC-600.mp4 (a sample encoded by Sagittaire for the coded comparison in the thread "x264 vs VP70 vs VP62 vs RV10") there is no video at all, only audio.

If i use build 20050303, there is no problem.

Btw. What are the differences between this builds?

[bob0r]

I can confirm what Palikrovol said, no video visible, even though ffdshow says its decoding video.

... as for AAC+SBR+PS playback, no more libmmd.dll needed, good work! :D

Version used:
http://celticdruid.no-ip.com/xvid/ > ffdshow > ffdshow-20050312.exe
http://www.aziendeassociate.it/./cd//ffdshow/ffdshow-20050312.exe

[Sharktooth]

replace libavcodec with this one: http://www.aziendeassociate.com/libavcodec.bpyramid.7z

[Xebec]

Originally posted by Sharktooth
replace libavcodec with this one: http://www.aziendeassociate.com/libavcodec.bpyramid.7z

Does not work for me. Crashes right in the beginning. Nero decoder works though.

[kurt]

on my machine it works with sharktooths hint....:D

[Ark]

It's strange:

- a b-pyramid encode doesn't work (with patched ffdshow)
- no b-pyramid but 16 ref frames doesn't work too (but hasn't this been fixed some rev. ago?)

Obviously the two together doesn't work...

Playback start correctly but as soon as player encounter one b-pyramid predicted frame, crash (MPC one of the lastest versionl, don't remember right now).

[iron2000]

Can there be a ffdshow lite that contains only video decoders for well-known/most used video codecs + the sharphen, postprocessing..etc?

ffdshow installs at 8.1MB.
Is this the usual size?

[Ark]

No, it should be around 3mb (if you mean the installer, not the decompressed folder). Where did you get it from?

And ffdshow was decoder only a looong time ago, when there was its encoding counterpart (can't remember the name..), now merged with it, so now it's only in this merged form.

[vlada]

What I would really appreciate would be a documentation for FFDShow. There are so many functions/switches that it would be really useful. For example I have no idea what "heuristic search" in subtitles filter does. I couldn't find any information about it. Also I have no idea how to set up correctly noise reducion, what is the difference between implemented postprocessing algorithms, I have no info about built-in deinterlace filters etc.
Also I asked in another thread if there is a way to cnotrol saturation, brightnes, gama or aspect ratio using Girder. Most of the time I only can switch on/off some filters. Fortunately I can set up subtitles' size/position which is a very important future for me. But being able to adjust colors would be even better.

Vlada

[iron2000]

Originally posted by Ark
No, it should be around 3mb (if you mean the installer, not the decompressed folder). Where did you get it from?

And ffdshow was decoder only a looong time ago, when there was its encoding counterpart (can't remember the name..), now merged with it, so now it's only in this merged form.


For the 8.1MB, its the size gotten from the Add/Remove control panel.
Not the installer size.

[bond]

milan finally updated the ffdshow cvs to the latest ffmpeg sources, i hope we can see a compile soon :)

[Hiro2k]

Maybe mplayer can finally catch up. :P

[rushin_911]

Originally posted by bond
milan finally updated the ffdshow cvs to the latest ffmpeg sources, i hope we can see a compile soon :)

I hope so too :)

[easyfab]

For testing here is the cvs latest libavcodec.dll (athlon-xp optimized)
http://rapidshare.de/files/936283/libavcodec.dll.html
(sorry for the host i don't know free better one)

For me it doesn't bring much more .

[Ark]

Originally posted by iron2000
For the 8.1MB, its the size gotten from the Add/Remove control panel.
Not the installer size.

Ok, 8mb is about the size of the decompressed ffdshow folder, so it's correct.

[rushin_911]

Originally posted by easyfab
For testing here is the cvs latest libavcodec.dll (athlon-xp optimized)
http://rapidshare.de/files/936283/libavcodec.dll.html
(sorry for the host i don't know free better one)

For me it doesn't bring much more .

Any possibility of a P4 build :)? (Not Prescott)

[easyfab]

Originally posted by rushin_911
Any possibility of a P4 build :)? (Not Prescott)

Ok here it is with p4 cflags
http://rapidshare.de/files/952291/libavcodec.dll.html

[Palikrovol]

can someone, please, mirror them? I can't download from rapidshare :(

Thanks

[kurt]

@Palikrovol: u have to klick "free" at the bottom of the page and then you have to wait ~60 seconds... finally there will appear a download link (again at the bottom) ...

[Palikrovol]

Originally posted by kurt
@Palikrovol: u have to klick "free" at the bottom of the page and then you have to wait ~60 seconds... finally there will appear a download link (again at the bottom) ...

:D Yes i know. But when i press the download link, the first page appears again :confused:

[kurt]

did you try "save as"...?

[Palikrovol]

Originally posted by kurt
did you try "save as"...?

Yes, it saves the html page.

[rushin_911]

Originally posted by easyfab
Ok here it is with p4 cflags
http://rapidshare.de/files/952291/libavcodec.dll.html
Thank you :)

[wata]

is there DXVA for ffdshow?

i am able to decode a high resolution divx5 clip using the provide divx player but doesn't have the cpu power when set ffdshow to decode it

there is also a high resolution xvid clip which ffshow don't have cpu power to decode but after changing the fourcc to dx50 i am able to play back without problem

i am using ati radeon with accelerated post-processing turn on
so i am assume this is the dxva

[IgorC]

When I do Divx3 FFDSHOW encoding at following settings VD just closes.
FFDSHOW last version 12 march. AVS prepared in GK 0.35 . Others videocodec with the same *.avs just work fine (x264, VP7, Xvid). It just crash during the 1-st pass

http://multimediacom.free.fr/Video/FFDSHOW.PNG

[PlazzTT]

Small bug for me in the March 12th build of the decoder...

The changes I make in the "Aspect ratio" area of the "Resize & Aspect" part of the config seem to be applied whether I have the "Resize & Aspect" box ticked or not?

[QQ]

I have some tough problems.. ever since version 200305XX or something like that, if I try to play file using ffdshow and output set to overlay mixer (using MPC in this case), all I get is gray screen. Now if I go to video driver setup, and check the saturation slider, I see (attached pic). Why does ffdshow mess up my saturation slider, thus messing up whole video? :(

ty

http://dreamy.ws/images/mpc/overlay2.gif

[oddball]

Is there a way for ffdshow to crop black borders from uncropped encodes? CUrrently I use DiVXG400 to do it as it can crop into the overlay. But it would be great if ffdshow had this ability.

[BlindWanderer]

Originally posted by oddball
Is there a way for ffdshow to crop black borders from uncropped encodes? CUrrently I use DiVXG400 to do it as it can crop into the overlay. But it would be great if ffdshow had this ability.

you mean to crop the frames automaticly? It would be an interesting feature with "Allow output format changes during playback" turned on. But i'm thinking of the decoder.

[oddball]

It's something I have asked for for some time. However I was told it's not possible which is somewhat odd since DiVXG400 can do it no problemo.

[vlada]

I believe it cannot be so hard to crop black borders from movie if FFDShow can add them.
But I have another 3 problems. The first one is displaying of itallic subtitles. The last letter is always cropped. Why?
The other problem I already mentioned before. I use tray icons for matroska splitter and for audio/video ffdshow decoder. So I have 3 icons there. If I stop a movie a start another one, without closing the player, I get another 3 tray icons. Btw. I unchecked "Make DS graph available to graphedit". What else could be the problem?
And the last thing is a problem with embedded subtitles and Haali's matroska splitter. FFDShow is trying to display the embeded subtitles end last used external once at the same time. Also switching subtitles in the splitter causes playback to pause, but this is not a big problem.
Vlada

[namchik]

Why the size of latest (???) ffdshow 2005.03.28 is so small compared to previous builds (1.8 Mb against 3+ Mb)... What features are removed? :confused:

[ernstblaauw]

I installed the latest build of celtic_druid (20050312) to use MakeAVIS with DVD2SVCD. But when I open a fake AVI made by MakeAVIS, DVD2SVCD sais in the preview window: 'could not decompress frame 0' (this is an Avisynth error). I never earlier used ffdshow, but a guy in another forum told me that this problem was introduced at the moment ffvfw.dll was integrated in ffdshow.
But maybe there is a workaround or solution available. Does anyone here uses DVD2SVCD with MakeAVIS? Is a solution available, or is this a known bug?

[timeismoney]

Originally posted by namchik
Why the size of latest (???) ffdshow 2005.03.28 is so small compared to previous builds (1.8 Mb against 3+ Mb)... What features are removed? :confused: Nothing removed, I compiled it via VC2003 in Release Win32 mode, not use ICL, so its size is small.

Here is a new version as ffdshow-20050405, of course unofficial version but fully cvs compiled, on default libfaad2 is removed, instead of realaac, I don't know if it can really fully replace libfaad2, but test some mp4/aac, sounds not bad. The tta decode function is pretty, but also need a splitter.

ftp://stormdl:stormdl@dl.yesure.com:3721/ffdshow/ffdshow-20050405.exe

[namchik]

timeismoney
but the size of installed ffdshow also differs: only 4.26 Mb .. previous builds were about 8 megs :confused:

[timeismoney]

Originally posted by namchik
timeismoney
but the size of installed ffdshow also differs: only 4.26 Mb .. previous builds were about 8 megs :confused: Yes, if i use a different compiler, the size will be increased, based on vs2003 itself, the performance maybe not as good as ICL, but the compatibility is not bad.

[samab]

timeismoney,

It's great that you make available your compiled versions for the public. One thing I like to add (from a brand new forum member) is that when your compiled versions hit the various codec websites (like free-codecs.com), those websites announced it as anonymously compiled with the big careful sign in capital letters.

Even though both you and Celtic compile "unofficial versions", the public treats the two versions very differently --- because they can't put a face on your anonymous versions.

I saw the 03-28 version of free-codecs.com and I decided against downloading it because I didn't know its origins. But now I know you compiled it and a quick search on this forum would immediately confirm that you are legit.

It's as simple as telling free-codecs.com that you compiled it and they checked your "credentials" by doing a quick search here in this forum.

[samab]

Milan is doing major changes to the source code, so he decided to suspend doing daily builds for a short while.

http://sourceforge.net/tracker/index.php?func=detail&aid=1170802&group_id=53761&atid=471490

[CruNcher]

Originally posted by timeismoney
Nothing removed, I compiled it via VC2003 in Release Win32 mode, not use ICL, so its size is small.

Here is a new version as ffdshow-20050405, of course unofficial version but fully cvs compiled, on default libfaad2 is removed, instead of realaac, I don't know if it can really fully replace libfaad2, but test some mp4/aac, sounds not bad. The tta decode function is pretty, but also need a splitter.

ftp://stormdl:stormdl@dl.yesure.com:3721/ffdshow/ffdshow-20050405.exe

Why do you replaced libfaad2 with Reals Decoder that one doesn't support parametric stereo, or does it?

[timeismoney]

Originally posted by CruNcher
Why do you replaced libfaad2 with Reals Decoder that one doesn't support parametric stereo, or does it? So milan tell it as:

And about realaac: I'm not sure, but for few sample files I tried (2 and 6 channel, LC and HE) it worked correctly and fast.
I already removed libfaad2 from installer (but I didn't enfore use of realaac if ff_libfaad2.dll is found)

Now the original nsis script of the installer has marked libfaad2 out of the package, and I tested some aac audio files or mp4 files I can find, realaac lib can support them correctly. But I don't know if my test is enough, so please try it, the report will be helpful to milan.

Originally posted by samab
timeismoney,

It's great that you make available your compiled versions for the public. One thing I like to add (from a brand new forum member) is that when your compiled versions hit the various codec websites (like free-codecs.com), those websites announced it as anonymously compiled with the big careful sign in capital letters.

Even though both you and Celtic compile "unofficial versions", the public treats the two versions very differently --- because they can't put a face on your anonymous versions.

I saw the 03-28 version of free-codecs.com and I decided against downloading it because I didn't know its origins. But now I know you compiled it and a quick search on this forum would immediately confirm that you are legit.

It's as simple as telling free-codecs.com that you compiled it and they checked your "credentials" by doing a quick search here in this forum. Thank you

[STTS]

Hi all.
Recently I discover the same feature (precision frame interpolation based on motion vector info) in hi-end LOEWE TV. TV owner suggest me to try WinDVD 6.0 with enabled DNM. I am be very impressed - DVD video look so smooth, so natural motion , I start to reviw all my DVD collection. Of course, it requires more CPU power (>>1.2 Hhz).
I am not familiar with mpeg1(2,4) codek internals but *feel* that if we have some motion information discovered during encoding and already stored in video stream. Is it hard to interpolate frame not onliy in I, P, B time, but between, in eventual time ? I think that any player can easy check for output device vertical blank interrupt and pass info to decoder. Decoder will use two nearest frames plus motion info, and produce frame that percision correlate with vb interrupt.
Sorry if it OT, but ffdshow seem to be best place for such features.

[namchik]

timeismoney
i can not download your new compiled build of ffdshow from your ftp... Maybe some other place?... Thanks.

[timeismoney]

Originally posted by namchik
timeismoney
i can not download your new compiled build of ffdshow from your ftp... Maybe some other place?... Thanks. Try here:

http://www.sbtop.com/UploadSoft/ffdshow-20050406.exe

Just updated, seems good. But as it is an unstable cvs compiled version, so use it as your own risk, thanks.

And I think more report will be helpful for milan.

[namchik]

timeismoney
Thanks... but this link is too slow... even on my dialup :(

[planet1]

http://ffdshow.sourceforge.net/tikiwiki/tiki-read_article.php?articleId=18


... :rolleyes:

[bond]

Originally posted by planet1
[B]http://ffdshow.sourceforge.net/tikiwiki/tiki-read_article.php?articleId=18that was the build where only the libav.dll was updated afaik and yes it was borked

[samab]

Originally posted by planet1
http://ffdshow.sourceforge.net/tikiwiki/tiki-read_article.php?articleId=18


... :rolleyes:

This is why I posted the above comments on why timeismoney should put a "face" on his compiled versions (i.e. telling milan and free-codecs.com who he is). At the very least, I think that you should name the file as ffdshow-20050406-vc2003. Celtic compiles various xvid versions with ICL 7 and ICL 8.1 --- and he names them differently.

It would be so much easier for everybody if there is more communication --- however I do understand that this is an open source project with volunteers contributing their free time. Milan should have said that he decided to stop daily build for a short while because of major code changes --- on his website, instead of burying it deep in some bug report comments.

[timeismoney]

To all points, I just compiled it for simply test as I think it is a great job by milan, to see many new functions have been there and no chance to use make me be anxious, so I simply build it for myself (a lazy woman so use the easiest VC71) and put it on my homepage with marked "Unofficial", "cvs build" & "VC71". I never announced it to be universal public until it has been transshipmented to other website like freecodecs.com by some friends, and someone here asked something for it. So I'd do something to cancel it, and tell milan my sorry

[samab]

Originally posted by timeismoney
So I'd do something to cancel it, and tell milan my sorry

You shouldn't feel sorry for it. You are nice enough to provide us with your build.

This is just a classic lack of communication by various volunteers.

I have said it twice before that it would be much easier for the rest of us if milan just announced on his website that he is temporarily suspending his daily build because of big code changes. Milan didn't explain his reasons to the public and we EAGERLY grab either celtic's 0312 release or your 0328 release without knowing that the source code is broken because of major code changes (and then of course all three of you gets the blame for unexplained crashes).

And milan has already stated that it's as much his fault of not writing down simple instructions on how to compile this open source project.

We appreciate your kindness in providing us with your builds.

[vlada]

Hello,
I have problems with MJPEG and DV encoders in FFDShow. I tried encoding to DV from VD (720x576@25, no audio) with latest build (20050406) and with some older ones. But the result is completely broken. It even sometimes crashes VD or player (FFDShow is used for decoding). Is this a known bug?
And now to MJPEG - it has a very poor performance. My computer (Athlon XP2000+) needs 80%-100% of CPU power for realtime capture in 768x576@25. It even sometimes drops frames. I remember I was able to capture full PAL video using Pegasus PicVideo MJPEG on Duron 600 MHz. This is a huge difference in performance. Could the MJPEG in FFDSHow be improved?
Thank you,
Vlada

[namchik]

I also noticed poor performance of MJPEG encoder in ffdshow 03.28... in milan's 03.03 it's ok and no dropped frames with my TV-tuner capture. By the way, milan explains this:The whole package is build in Visual Studio .NET 2003 with Microsoft compiler in Release target. While this isn't bad choice for ffdshow.ax itself and several other libraries included in VS.NET solution, it's inapropriate for building libavcodec and libmplayer.dll. Those two libraries are essential for good performance of ffdshow - libavcodec for decoding/encoding most of video and audio formats and libmplayer for software scaling, many colorspaces conversions and fast mp3 decoder. When those two libraries are build by Microsoft's compiler, all hand optimized code developed to effectively use MMX, 3DNow and SSE extensions is skipped resulting in much worse performance.

[timeismoney]

Originally posted by namchik
I also noticed poor performance of MJPEG encoder in ffdshow 03.28... in milan's 03.03 it's ok and no dropped frames with my TV-tuner capture. By the way, milan explains this: Thanks all who care it, I've contacted milan and luckily heard that two ffdshow builds in the daily builds section will be released soon: a SSE ICL build as usual and a build using uClibc++ as STL library.

[vlada]

namchik> I know about this incorrectly compiled build, but I'm not sure if this is my problem. I noticed the poor performance of MJPEG even in the last build (which should be O.K., I believe) and also in all previous builds I tested. So it seems to me as a problem with optimization of the codec.
Also could anyone else test the DV codec, please? Decoding works correctly, but the encoder is probably corrupted.
Regards,
Vlada

[madman1980]

I have two questions.

Is it possible to include all the different compiles that I see (sse, regular, sse2 etc) in one installation file? And just let it detect CPU itself. That would make things a lot easier for most people.

Will there be a compile for Windowx XP x64 soon?

[Terka]

which filter should i use to achieve high quality for interlaced video playback at 50fps?

[magicclue]

http://100fps.com/

scroll down a little...

[bob0r]

http://mirror02.x264.nl/ffdshow_reg_error.jpg

Celtric Druid, can this be fixed? If so please put a new version on http://celticdruid.no-ip.com/xvid if you could.

This error has been confirmed by more people, running installer again will "solve" the problem.

[Vitos]

Originally posted by bob0r
Celtric Druid, can this be fixed? If so please put a new version on http://celticdruid.no-ip.com/xvid if you could.
This error has been confirmed by more people, running installer again will "solve" the problem. [/B]

Where did you read that's Celtic Druid's compile? You can't find it on his page. It's unofficial compile (you can read about it earlier in this thread) and isn't recommended at all.

[bob0r]

Originally posted by Vitos
Where did you read that's Celtic Druid's compile? You can't find it on his page. It's unofficial compile (you can read about it earlier in this thread) and isn't recommended at all.

Thats what i figured, since it was not on his main website.
A well know doom9 mod gave it to me :o

I got it from http://esby.free.fr/CelticDruid/ffdshow-20050406.exe
I figured it was his.... i PMed him also, so we can sort it out, thanks for the reply.

What is the best version you recommend now?

[namchik]

i use this:
http://cutka.szm.sk/files/ffdshow-20050303-sse.exe
works fine...

[celtic_druid]

The filesize is way too small to be my compile.

Try here: http://m17n.cool.ne.jp/freeware/mpc/

[namchik]

2 celtic_druid
in ffdshow-20050406 (http://m17n.cool.ne.jp/freeware/mpc/1/ffdshow-20050406.exe) there's a problem with russian fonts in the installer when choosing the components (though GNU license is OK)

Also i capture video with tv-tuner using ffdshow encoder (mjpeg). Builds after 20050303 are very CPU consuming during real-time capture... about 95-100%. With older builds CPU load was about 75%

ps. Sorry for my bad english :rolleyes:

[ExtraEye]

what's the difference between milan's ffdshow build to celtic druid's ffdshow build?

[namchik]

2 celtic_druid
i restarted my PC, now cpu load is normal... my apologies

and problem with russian font still exists, but i'll ask other russians, coz maybe something is wrong with my "MustDie" =)

[celtic_druid]

Other than adding msvcrt7.1 to the installer I didn't touch it so any problems with it you would need to contact Milan. Unless it has something to do with me compiling the nsi script?

The difference is that Milan's build is now over one month old and there have been lots of changes since then.

[bob0r]

@celtic_druid

Thanks for clearing everything up.

I have put your "new" version on http://x264.nl which should work again (when your DNS is updated)

Keep up the great work!

[ExtraEye]

is ffdshow in trouble too?
http://ffmpeg.sourceforge.net/
http://www.mplayerhq.hu/homepage/index.html

i sure hope things woun't change...

[esby]

Well with the european patent problem, every software that deal with patented material will be in trouble... meaning allmost every piece of software, including the open source ones.

And since we are in the multimedia field, everything is patented or something near that.

esby

[Liisachan]

I once asked Matroska devs why ffdshow in their packs were so old. and this (http://lists.matroska.org/pipermail/matroska-devel/2004-July/001828.html) is their reply:

The big problem is, all newer versions of ffdshow are coming ( which i
find VERY annoying to be honest ) with FFvfw bundled, and for licensing
reasons we definitely decided againt distributing a working MPEG4
encoder in our packs. Some of our team are family men or have a job, we
cant risk to have the MPEG-LA standing outside our houses and trying to
sue us for millions of US$'s of lost revenues, because matroska pack is
distributed several 10.000 times per months.


In this theory, I will be sued too eventually for distro'ing celtic_druid's builds on http://m17n.cool.ne.jp/freeware/mpc/

Software patents are getting really weird :(
libdca is another example of many
http://www.via.ecp.fr/via/ml/libdca-devel/2004-12/msg00001.html

I hope at least people can distribute source codes freely, as they are just plain texts after all.

[Mug Funky]

it's good to be careful, but (no offense to matroska project which i think is a good idea, albeit full off politics) i doubt anybody is using mkv for commercial purposes.

[iron2000]

Decided to try Celtic's 20050406 build.

The size is bloated to 10.43MB from the already big 8.1MB!!
Whats going on?

[PatchWorKs]

A possible solution ?

Decode -> all formats
Encode -> just free-patented ones

This can convince devs to migrate towards Theora/Dirac/SNOW/anyotherfree that is not so bad... ;)

[celtic_druid]

The installer is actually smaller than my last release.
5,537,792 ffdshow.ax
2,426,880 libavcodec.dll
536,576 ff_libfaad2.dll
385,024 ff_x264.dll
368,128 libmplayer.dll
348,160 msvcr71.dll
208,896 ff_libdts.dll
192,512 ff_theora.dll
188,416 TomsMoComp_ff.dll
159,744 ff_libmad.dll
159,232 ff_realaac.dll
131,072 ff_samplerate.dll
112,640 libmpeg2_ff.dll
95,744 ffvdub.vdf
91,648 ff_unrar.dll
76,288 ff_liba52.dll
74,240 ff_tremor.dll
35,840 ff_wmv9.dll
25,600 makeAVIS.exe
7,680 ffavisynth.dll
5,632 ff_vfw.dll
4,608 FLT_ffdshow.dll
4,096 ff_acm.acm

If I were to make it SSE2 only then I could shrink it a fair bit I would imagine. The way that it is at the moment it should work on all CPU's and the ICL parts should make use of SSE/SSE2, etc. if available, at least if you have an intel CPU. Second largest file is libavcodec.dll though which is just a generic gcc compile.

If you are worried about the size, stick with the MSVC built version that is floating around.

[esby]

Encode -> just free-patented ones

Well that's a nice goal, but not very useful, since you probably won't be able to use any of the codecs available...
Mpeg4 is patented, wmv & qt are also...,and I am not talking of possible sub licences affecting the whole.
And of course, even the codecs you are naming might be patented...

esby

[Liisachan]

Originally posted by PatchWorKs
A possible solution ?

Decode -> all formats
Encode -> just free-patented ones

Maybe not. as libdca (decoder for DTS) devs say: "The patent might be challengeable (at least for the decoder side) because it doesn't describe any invention as such but more a format of data ... However as everybody knows, it is a lengthy and expensive process to try to challenge (even a bogus) patent so ..." :(

This can convince devs to migrate towards Theora/Dirac/SNOW/anyotherfree that is not so bad... ;) That's exactly what I'm thinking, too. And let's not use MP4 or WMV or RV, let's prefer OggFile/OGM/MKV/NUT etc...as long as they can do the same job.

"The sad reality of free software is that it may take drastic measures such as these to make the four freedoms important enough to value."
(Your software rights or the best tools: often a sad choice (http://trends.newsforge.com/article.pl?sid=05/01/05/1842204))

[planet1]

prefer OGM over MP4 - now whats wrong with that ??? :rolleyes:

[Liisachan]

I just said patent-free things are preferable "as long as they can do the same job" that patented software can. Nothing more, nothing less.

In reality, I don't think (for instance) Ogg Theora can do what XviD can atm, but history may repeat itself. Think about MP3 vs Vorbis, or GIF vs PNG. Sometimes free software is, or will be, better in quality too, not just being open-sourced and patent-free.

[bond]

Originally posted by Liisachan
In reality, I don't think (for instance) Ogg Theora can do what XviD can atm, but history may repeat itself. Think about MP3 vs Vorbis, or GIF vs PNG. Sometimes free software is, or will be, better in quality too, not just being open-sourced and patent-free. [/B] you are mixing up formats with implementations of formats
i am sure you can write a vorbis encoder that gives worse quality than a mp3 encoder or a theora encoder that is better than a mpeg-4 asp encoder
apart from that theora has lots of features where its superior to mpeg-4 asp (which for example xvid is following)
apart from that theora uses patented technologies. the important point imho is not whether a technology is patented but whether its license free or not

[vlada]

OGM is dead the future is Matroska (MKV).
Btw. which video codecs are patent free? I suppose OGG Theora, but what about Dirac or Snow?

[opsis81]

The MP4 container is patent free.

celtic_druid,it would be really great if you could compile a special build with the video decoder and the audio decoder only.You could even remove many audio and video filters from them.All these filters confuse not experienced users and can cause major playback problems when used incorrect.
It would also be great if it could be registered together with the original build(by changing the original clsid,the name of the decoders,the merit of the decoders etc)

[Liisachan]

Originally posted by bond
the important point imho is not whether a technology is patented but whether its license free or not That's your opinion. I respect what you think and feel. But I'm not you. My position is that software patents should be banned entirely. You don't have to agree with me at all. I'm not trying to persuade anyone either. Just a random thought.

[vlada]

Liisachan> I don't know too much about SW patents, but I expect it should allow you to patent some algorithms like bubble sort, for example. Isn't MPEG compression rather a technology then a certain algorithm? Couln't it then be patented as every other technology?
Btw. I believe you should have the right to patent a technology, but only for short time (10 years or so). You only should be able to patent some inovative things and not anything what has already been used before. But if you would cancel patents completely, it would slow down technologic progress, because researchers wouldn't be able to earn a lot of money. But I do believe that patents need to be reduced.

[ExtraEye]

Originally posted by Liisachan

"The sad reality of free software is that it may take drastic measures such as these to make the four freedoms important enough to value."
(Your software rights or the best tools: often a sad choice (http://trends.newsforge.com/article.pl?sid=05/01/05/1842204)) [/B]

That's so true!
I'm kind of mixed up though...
Which video codecs/audio codects/formats/etc are patented?
I also didn't really understand what the new law the wanted to make was...

[Doom9]

the important point imho is not whether a technology is patented but whether its license free or notThat that is what really matters. If you take Theora for example, it's not patent free, and On2 does not indemnify you against patent claims by third parties. Since there are so many patents, unless you do specific (expensive) research, you cannot really know whether other codecs like Snow or Dirac would violate any patents either.

I fully agree though that software patents should not be granted and that the patent system worldwide needs a serious overhaul. Pretty much every patent law states that only stuff that represents a considerable and non obvious achievement can be patented.. but how can the lawyers that work at the patent office figure out if something is patent worthy or not? You need people working in the field in question to determine that.. and they are more likely to work in the private sector because it pays more..

And interestingly enough, we've seen in the news that people from patent heavyweights like Cisco or IBM dare to ask for the sense of patents altogether and some think patents have done more to slow down progress than to speed it up. But now we're completely off topic.

[esby]

I found a minor typo in the installer.
I don't think that 'Dobly surround downmix' actually exists in the audio filters selection

esby

[vlada]

Hello all,
I did a testing of different deinterlace algorithms and I wanted to include FFDShow too. I used Milan's version from 2005/03/03. Unfortunatelly I found out, that most of the deinterlace filters are broken. TomsMoComp, Cubic and Kernel seem to work correctly but
1) DgBob is depending on a threshold either interpolating everything or nothing.
2) The frame doubler isn't working at all
3) DScaler filters were not working correctly in VirtualDub. Maybe because of RGB color space?

I would really like to see a fast filter producing double framerate. The DScaler Greedy 2-Frame has pretty low quality, the internal framedoubler doesen't work as I already mentioned. The DScaler 2 frame deinterlacer is similiar to KernelDeint, which is pretty good, but is too slow. I only get around 30-40 fps on my Athlon XP 2000+.

The situation in filters preserving the original framerate is slightly better. Although blending is terrible and interpolation isn't a good solution as well, TomsMoComp is a decent filter. It produces pretty good pirture in realtime taking only 60-70% CPU together with MPEG-2 decoder. So this is my choice for 25 fps deinterlacing. But the playback isn't always smooth.

What about implementing a simple 50 fps BOB? No analysis just plain bicubic interpolation of both fields. Then a simple analysis could be added for faster processors. Shouldn't the internal filter work like this, if it worked?

And I also have one idea. This thread is getting really huge and there are many unsolved questions and problems. I was thinking about founding a FFDShow forum with several categories (bug reports, development, support etc.) I would be able to set it up for free but I only get 5 MB for database. I don't think it is enough, but I have no experiences with it.

And now the last thing. If you are interested in result of the deinterlacers test I did, you can check the pictures at http://jech2.webz.cz/gallery/index.php. Unfortunately my coments are written in Czech, but you can still see the results yourself. For me the winner was Smart Deinterlace for VirtualDub, on second place finished Deinterlace MAP and DgBob, which had some problems in one of the tests.

Best regards,
Vlada

[oddball]

ffdshow audio won't decode 5.1 HE-AAC to SPDIF output! :confused:

[kurt]

you can transcode your AAC-Stream to AC3, look at the picture...
http://img230.echo.cx/img230/8384/ffdshow7xd.th.jpg (http://img230.echo.cx/my.php?image=ffdshow7xd.jpg)

[oddball]

I know that. But it transcodes *EVERYTHING* including MP3 audio which I don't want to transcode a second time (It's already been transcoded from AC3/DTS once. I don't want any further degradation of sound). Plus it makes the volume of transcoded MP3 far too loud. I want the option in the codecs page just like you have a drop down for AC3 that says SPDIF. We need one for AAC and Ogg also!

EDIT: Actually I'd like to see a transcode drop down SPDIF option on all codecs that support multi channel surround sound.

[celtic_druid]

You have a reciever that handles vorbis and 5.1 aac?

[LigH]

He probably rather wants a drop-down option for all other multi-channel capable formats (except AC3 itself) separately, if they shall be transcoded to AC3 and passed to S/P-DIF.

[oddball]

Originally posted by LigH
He probably rather wants a drop-down option for all other multi-channel capable formats (except AC3 itself) separately, if they shall be transcoded to AC3 and passed to S/P-DIF.

Exactly. At present I feed 5.1 HE-AAC amd 5.1 Ogg Vorbis to AC3Filter via the PCM option to transcode. BUT! It transcodes *ALL* PCM including MP3 which I do not want it to do as that resamples the MP3 thus lowering the quality one stage further. I want ffdshow to handle MP3 independently. But if I enable PCM transcoding in AC3Filter it takes over all PCM so ffdshow gets bypassed for MP3 playback. Valex has said he is studying the Vorbis API so native support for HE-AAC and Ogg Vorbis may come in a later version of AC3Filter.
ffdshow currently supports those codecs via Tremor's filter. However ffdshow does not allow transcoding those to 5.1 AC3 SPDIF output unless you enable it for ALL PCM, including MP3 (Just like AC3Filter actually). Which is not what I want.

[LoKi128]

How about two options for output codec: two-channel and multi-channel. All two channel streams will be output as PCM (or whatever option in the dropdown) and multi-channel streams get a different dropdown, like mixdown to stereo PCM, AC3 transcode, or something else.

AFAIK, the only way to get a digital multi-channel signal into a receiver is AC3 or DTS. Are there any receivers that understand multi-chan PCM, like the datastream of a 6-channel WAV or something?

[WimW]

Hi all,

I'm trying to get a decent capture with a Compro card. I used to use the mpeg2 that comes with it, but now I've got a new HD, so I'm looking for a lossless solution. Unfortunately, 720x576 huffyuv won't work. I think it's a throughput problem, cause the CPU is not maxed out.

So I tried a lot of other codecs, but none seem to be fast enough on my Athlon XP 1700+. I'm not buying any new hardware for this. I can use 576x576 huffyuv and correct AR when resizing, or use PicVideo MJPEG.

Anyway, here's my question: why is ffdshow huffyuv even much slower than ccesp huffyuv? I thought maybe there's some problem with my ffdshow installation. If I can fix that, then maybe one of the other lossless codecs would work...

Thanks,

Wim

[madman1980]

Does anybody know how the normalize function works? Does it find the peak of the whole file and adjust playback volume level from that - or does it change the volume as it goes along? (I obviously don't want the latter for neither music nor movies).

[eb]

I think that in 1 pass it sets at start max and than setting it lower when meets case were 100% is crossed.
In 2 pass it is looking for maximum and then setting corections accordingly.

[Irwin]

Hi
OffTopic

Is possible to release ffdshow for pocketpc platform?

[Episode]

I don't think it's possible to port ffdshow for pocketpc. However, Videolan is available for pretty much every platform (including pocketpc/wince), so you might wanna try that. Iirc it uses same decoding library as ffdshow (libavcodec), so it should be as fast as ffdshow.

[BoNz1]

celtic_druid, would it be possible to make a new ffdshow build? The latest one I have is more than a month old :( and there have been a lot of speedups and improvements in libavcodec recently. I have been trying (unsuccessfully) to make my own libavcodec.dll but I can't. Thanks :).

[celtic_druid]

If I am not too hung over tomorrow I will rey to put up a new compile.

[bob0r]

Originally posted by celtic_druid
If I am not too hung over tomorrow I will rey to put up a new compile.

Thats the spirit!! :cool:
I'll test and put it online when it arrives, x264 is the best! :D

[BoNz1]

Well looks like it didn't happen today. Hehe, I hope you had a blast :). Cheers.

[celtic_druid]

Actually that message was from when I got home at apparently 5am and yeah, it didn't happen yesterday (Sun), but I just did a compile then.

[Liisachan]

and I've mirrored it already
http://m17n.cool.ne.jp/freeware/mpc/
http://subforge.s116.xrea.com/freeware/mpc/

[bob0r]

Where did you get it from?
I usually check
http://celticdruid.no-ip.com/xvid > ffdshow
( http://www.aziendeassociate.it/cd.asp?dir=/ffdshow )

Edit:
Oh hehehe, i refresh and it's there :D
( I'd still like to know where you get it from though :p )

Ah Sharktooth, where did you get it from then?
And i am testing it, when my revision 236 x264 encode is done :cool:

[Sharktooth]

i've just updated the main mirror.

[celtic_druid]

Liisachan, Sharktooth and a few others have access to where I upload stuff.

[bob0r]

@celtic_druid
ok

@celtic_druid
ffdshow-20050510.exe A7E69FB0 (May 10 2005)

[Library] libavcodec.dll
[Version] 0.4.9-pre1, build 4755 (May 23 2005 14:28:03)


Were you able to grab some files from the future or did you name it wrong?

[Sharktooth]

maybe ffdshow source is from 2005.5.10 and then he got the 1005.5.23 sources of libavcodec....

[Liisachan]

@bob0r: please read "About this page" (http://m17n.cool.ne.jp/freeware/mpc/#about).
celtic_druid doesn't want people to download the files directly from his server (which resulted in troubles because of too high traffic). This is where "mirror sites" come in. Mirror sites get the files directly from celtic_druid and disto'ing them. The traffic and/or server load can get really high. That account of mine on cool.ne.jp was once suspended :( Back then, prolly my web hosting provider got angry because I used too much bandwidth to host a lot of popular, big files (aka MPC).

[Sharktooth]

well... i dont want to publish my mirror statistics... i do it for your heart :D

Ok, maybe just this number: 276.674 hits

[esby]

I did a manual update of my mirror this morning.
(So new build being there too)

Since we are in the stat parts...

http://esby.free.Fr/perso/usage.png
(I started mirroring in january)

esby

[708145]

I'm really surprised there are so many video enthusiasts!

276K downloads ; ~3 versions per month => 100K compressionists!

bis besser,
Tobias

[BoNz1]

Originally posted by celtic_druid
Actually that message was from when I got home at apparently 5am and yeah, it didn't happen yesterday (Sun), but I just did a compile then.

Wicked, many thanks :).

[celtic_druid]

ffdshow-20050510.exe is what the NSI script spat out. I just neglected to rename it. Probably more correct would have been 20050521 or so as I think that is the last time that the cvs was updated before compiling.

[Liisachan]

hey isn't libmpeg2_ff.dll buggy this time?
Test it by playing MPEG-2/VOB after configuring: Codecs->MPEG2->libmpeg2.
To me libavcodec does work while libmpeg2 doesn't, it outputs strange reversed-color, embossed video. Anyone can reproduce?

[DeeGee]

Yep, seems so. Very badly messed up image. Green/black blocks with some color. Motion vectors seem to still work, but as they just move around green/black it isn't very pretty :).

[Yong]

same here with mpeg1-video decoding problem with libmpeg2,
output even worst than mpeg2 decoding:rolleyes:

[LotharZ]

Same problem here with mpeg1 and mpeg2 running libmpeg2, and with libavcodec get some crazy fps (go from 0 to 60 fps and weird movement).

[Taurus]

Confirmed.
Same on my machine.

Cheers

Taurus

[Yong]

Originally posted by LotharZ
Same problem here with mpeg1 and mpeg2 running libmpeg2, and with libavcodec get some crazy fps (go from 0 to 60 fps and weird movement).
iirc the fps problem only occur in mpeg1(ES doesn't have fps problem) video with libavcodec decoding, this is a known problem.

[edit] btw decoding hd quicktime h264/avc video also broked,
output video distorted.

[madman1980]

So I should stick to ffdshow-20050406.exe for now to avoid problems?

Btw, is there a 64bit compile?

[celtic_druid]

http://celticdruid.no-ip.com/test/libmpeg2_ff.7z

Ok, libmpeg2 and ICL7 don't seem to get along. Above is recompiled with gcc and appears to work ok here.

[Taurus]

Originally posted by celtic_druid
http://celticdruid.no-ip.com/test/libmpeg2_ff.7z

Ok, libmpeg2 and ICL7 don't seem to get along. Above is recompiled with gcc and appears to work ok here.

Libmpeg2 is working (again) on mpeg1 + 2 @ my win98 and winXp machines.
Thank you for your fast responding "CD". Maybe you should rename your anonymous to " Lightning Fast CD" :D

Cheers

Taurus

P.S.: Downloading the file was hell. Only with the help of an agressive download manager I was able to get the file.
But anyway a big THANKS.

[Sharktooth]

Mirrored here: http://www.webalice.it/f.corriga/misc/libmpeg2_ff.7z

[celtic_druid]

Well there is only 10-20kB upload on that line.

[bob0r]

@celtic_druid

Maybe compile a new ffdshow overal, when all complaints are resolved.
People don't like to download seperate files.

We only want the best (especially when there is standalonbe no x264 decoder yet)

;)

[celtic_druid]

Well today is Friday, so if I don't have much to do here can knock off work early and do a new compile.

[Liisachan]

Thanks as always :)

well, i got like 10 reports about some other troubles with this new build. i cannot reproduce the pb myself (it works ok for me), but i can sum up the feedback:

(1) The installer complains 'ffdshow.ax' is missing, or "Error while registering ffdshow.ax" even if you try to install this version right after rebooting your OS.

seeminly, this might happen when you uninstall the previous version before installing this version, and according to some reports, the workaround in that case is: first re-install a random old version, and then overwrite it by this version.

Did anyone experience this problem? Or can anyone guess what is wrong?

(2) Generally, i got a vague impression that this version works better on Win2K and a bit less reliable on WinXP--especially it installs but doesn't work on x64Edition.

[MacAddict]

Loaded latest build on 3 XP machines with SP2 and not had any problems with the installer. I uninstalled the previous builds first.

The libmpeg2 build seems to work great so far but need to test MPEG2 a little more.

[celtic_druid]

msvcrt71.dll is a dependancy of ffdshow.ax and was not included in the installer. Some of the previous installers did include it which is why installing over an old version would work. Really people should have the file in they system32 folder though.

So previous builds worked ok on XP64?

[Liisachan]

Are you sure? I installed your ffdshow-20050510.exe into a newly created folder, and I got ffdshow.ax 1.0.1.1 dated 2005-05-23 04:04 UTC there. I believe this one is included in the package.

About 64-Bit Editions, I doubt it. It's very new and only a few use it. I think you can ignore it for now.

[Leak]

Originally posted by Liisachan
Are you sure? I installed your ffdshow-20050510.exe into a newly created folder, and I got ffdshow.ax 1.0.1.1 dated 2005-05-23 04:04 UTC there. I believe this one is included in the package.

Actually, he meant that msvcrt71.dll, which ffdshow.ax depends on, wasn't included in the installer. If this gets removed when you uninstall an old build installing the new build must fail, as loading ffdshow.ax obviously won't work.

np: Apollo 440 - Carrera Rapida (Theme From Rapid Racer) (Electro Glide In Blue)

[LigH]

About "Friday - early off":

One half of Germany - mainly southern and western federal states, which are more related to the catholic church, especially Bavaria - had a leisure holiday yesterday, and most people tried to expand it to a "long weekend" (Thu => Sun). Other federal states in northern and eastern Germany, which are more related to the protestantic church, have less leisure holidays and just a normal week of work. Isn't that "fair"?! :rolleyes:

But anyway, way too many people in Germany don't need to care about holidays, without a job. :angry:

[Liisachan]

@Leak: thanks! That explains everything :)

EDIT: You mean msvcr71.dll? (-t ?)

[subliminal]

yes that´s what he was saying

[LigH]

celtic-druid wrote: msvcrt71.dll - therefore, Liisachan asked back...

[celtic_druid]

Same ****, different bucket. New installer includes it anyway.

[Sharktooth]

mirrored

[Liisachan]

upped
http://m17n.cool.ne.jp/freeware/mpc/
http://subforge.s116.xrea.com/freeware/mpc/

[bob0r]

Same 4 stars!
Mirrored!
Upped!
Tested x264... OK!

*bob0r buys celtic_druid some ice cream

Good work! :eek: (bloody 35 Degrees here in Amsterdam) :confused:

[LigH]

Got it. :p

(Sorry, just a small joke... ;))
__

And a queer thought, but who knows...: Because several people complained that H.264 videos look strange, because they left post-processing activated (which is obviously much more useful for MPEG-ASP video). Which way might be most useful for switching off PP on H.264 videos? Maybe auto-applied profiles depending on the codec?

[Vincent Vega]

after installing latest ffdshow with vdub plug-in i get the following error upon exiting vdub. "VirtualDubMod.exe - Application Error: the instruction at ... referenced memory at ... memory could not be read. click ok to terminate the program."
both addresses are different every time it happens. you can reproduce that: load divx+ac3 avi and use video direct stream copy, ac3 -> mp3 (lame).
i think i also got the same kind of error in zoomplayer while configuring customized playback to use ffdshow. though this one i couldn't reproduce later.

[LigH]

You want to convert an AC3 audio stream inside VirtualDubMod into MP3?

And you can use ffdshow as ACM decoder?

[Vincent Vega]

Originally posted by LigH
You want to convert an AC3 audio stream inside VirtualDubMod into MP3?

And you can use ffdshow as ACM decoder?

no. i'm saying that doing something totally unrelated to ffdshow in vdub still causes this error on exit. probably because vdub loads this new plugin. i never had this error before installing it.

[Vincent Vega]

after installing latest ffdshow with vdub plug-in i get the following error upon exiting vdub. "VirtualDubMod.exe - Application Error: the instruction at ... referenced memory at ... memory could not be read. click ok to terminate the program."
both addresses are different every time it happens. you can reproduce that: load divx+ac3 avi and use video direct stream copy, ac3 -> mp3 (lame).

...doing something totally unrelated to ffdshow in vdub still causes this error on exit. probably because vdub loads this new plugin. i never had this error before installing it.


anyone?

[TheShadowRunner]

About SVQ3 in 20050527 :

my test MOV (Video:SVQ3/ Audio:AAC)
http://pdl3.warnerbros.com/thematrix/us/med/trailer_final_480_dl.zip

IN FFDshow version :
20040312 : works absolutely perfect!
all build in between : libavcodec.dll crash
20040418 to 20050527: works but frame rate doesn't resolve (2fps or something)

Is it possible for you libavcodec gurus to "take" the SVQ3 code from 20040312 and somehow put it in the next FFDshow build?

Later,

TSR

[bob0r]

http://forum.doom9.org/showthread.php?p=664353#post664353

ffdshow(latest build) crash when decoding x264 encoded with --8x8dct --analyse i8x8 enable.
moonlight h264 decoder cannot decode it(blank output)...
only the latest CVS mplayer is able to play it back


Yup, i can cornfirm when you enable --8x8dct the latest ffdshow by celtic_druid does infact crash.


So hopefully celtic_druid can fix both TheShadowRunner's and our problem, and put a fresh new all working ffdshow build online!

[celtic_druid]

I'll see about putting up a new build, maybe tomorrow. I did put up some files (ffdshow.ax, ffx264 and libavcodec) for testing on I think the 2nd.

Should warn everyone that I might not be around next month to compile anything.

[Revgen]

So you're going on vacation?

[celtic_druid]

Yep, looks like I'll be going to the UK, France and Spain.

[JoeBG]

Yep, looks like I'll be going to the UK, France and Spain.

Have a nice holiday!!! If cold UK why not cold Germany?

[celtic_druid]

Well I only found out that I might be going 3 days ago and I won't really have time to go everywhere. It is only a short trip. Never been to Germany though so I guess I should go sometime. I did visit Hahndorf (oldest German settlement in Australia) last month though.

[Mug Funky]

have a good trip, celtic_druid!

though, if you're trying to escape the melbourne winter, the UK isn't the best place to go :)

[Zero1]

Everyday is winter here in the UK :(

Anyhow, certainly looking forward to a FFDshow update ;)

[celtic_druid]

I like it cold. Besides the last time I was over there in January and it really wasn't that cold and it should be warmer now.

Oh yeah, new ffdshow build should be up later.

[708145]

Well I only found out that I might be going 3 days ago and I won't really have time to go everywhere. It is only a short trip. Never been to Germany though so I guess I should go sometime. I did visit Hahndorf (oldest German settlement in Australia) last month though.

Hahndorf :rolleyes: I hope they didn't scare you too much.
It has about the same amount in common with Germany as IE has with a decent browser ;)

I wish you a nice holiday here in .eu anyway. May the source be with you.

bis besser,
Tobias

[celtic_druid]

New build is up.

[LigH]

And where? Italy / France / Japan 1/2 / x264.nl ?

[TheShadowRunner]

Where is it up? :)
Also, did you change anything related to SVQ3 ?
See you,

TSR

[celtic_druid]

http://mirror05.x264.nl/celtic_druid/ffdshow-20050607.exe

[Sharktooth]

main mirror is updated too :)
http://www.aziendeassociate.it/cd.asp?dir=/ffdshow

[LigH]

Great, guys! :thanks:

Small typo: "disabdled" for "Raw video" in the DS filter "Codecs" config.
__

Aah! Post-processing option to configure H.264 deblocking! Nice! :D

[maska]

1. No "Create shortcuts" option in the new build installer? :confused:

2. It's still an old build name on the new installer.

[ChronoReverse]

Is this the build that can decode HP x264?

*downloads to find out

[Liisachan]

http://subforge.s116.xrea.com/freeware/mpc/
http://m17n.cool.ne.jp/freeware/mpc/

[Sharktooth]

Is this the build that can decode HP x264?

*downloads to find out
yep..

[G-Slide]

hi all ! i have a problem wth zoom player and MPC : i play ND AVC files with 2 audio tracks (franche/english) and subtitles... with click right, i can pick one of the trtack but nothing happens wwhen i click on one to change the band... can you help me ? (it works on nero showtime)

[celtic_druid]

Doesn't sound like an ffdshow issue.

[TheShadowRunner]

no changes about SVQ3 but thank you for the new build neverthless ;)
Celtic, would you know where i can contact the libavcodec developers to tell them about the SVQ3 problem?
Thank you,

TSR

[celtic_druid]

So mplayer, ffplay, etc. also have the same issues? Check before contacting them.

Best place would I guess be their mailing list.

[LigH]

@ G-Slide:

Separating audio streams is the duty of the splitter, not of the decoder.

Maybe haali's splitter helps (MP4 support may have to be enabled).

[Shirokuu]

Liisachan, i tried installing your mirrored version hosted on http://subforge.s116.xrea.com/freeware/mpc/ but the NSIS installer prompted me with an error:

"The installer you are trying to use is corrupted or incomplete. This could be the result of a damaged disk, a failed download or a virus"

I didn't have a program installed to check the hash, but i suggest you refresh the uploaded file. It may have been damaged. The mirror posted by Celtic Druid (hosted on http://mirror05.x264.nl/celtic_drui...ow-20050607.exe) worked just fine. I hope to have informed you adequately.

[TheShadowRunner]

So mplayer, ffplay, etc. also have the same issues? Check before contacting them.

Best place would I guess be their mailing list.

Hmm how could i check this?
I only have one PC with WinXP..
Later,

TheShadowRunner

Edit: Ok, I tested with http://ftp5.mplayerhq.hu/mplayer/releases/win32-beta/MPlayer-mingw32-1.0pre7.zip and it works just fine! so maybe it's an issue with ffdshow after all..

Here's the DOS output
C:\Unzipped>c:\unzipped\mplayer\mplayer c:\unzipped\a.mov
MPlayer 1.0pre7-3.4.2 (C) 2000-2005 MPlayer Team
CPU: Intel Celeron 2/Pentium III Tualatin (Family: 6, Stepping: 1)
Detected cache-line size is 32 bytes
CPUflags: MMX: 1 MMX2: 1 3DNow: 0 3DNow2: 0 SSE: 0 SSE2: 0
Compiled with runtime CPU detection - WARNING - this is not optimal!
To get best performance, recompile MPlayer with --disable-runtime-cpudetection.

c:/windows/fonts/arial.ttf doesn't look like a font description, ignoring.
Cannot load font: c:/windows/fonts/arial.ttf
Playing c:\unzipped\a.mov.
QuickTime/MOV file format detected.
--------------
MOV track #0: 596 chunks, 3626 samples
MOV: Found unknown movie atom SMI (21)!
Image size: 480 x 260 (24 bpp)
Display size: 480 x 260
Fourcc: SVQ3 Codec: 'Sorenson Video 3'
--------------
MOV track #1: 741 chunks, 6506 samples
Audio bits: 16 chans: 2 rate: 44100
Audio extra header: len=111 fcc=0x77617665
MOV: Found MPEG4 audio Elementary Stream Descriptor atom (51)!
Fourcc: mp4a
--------------
MOV: longest streams: A: #1 (6506 samples) V: #0 (3626 samples)
VIDEO: [SVQ3] 480x260 24bpp 24.000 fps 0.0 kbps ( 0.0 kbyte/s)
Clip info:
name: Matrix Reloaded Trailer
author: www.thematrix.com
copyright: 2000 Warner Bros. All rights reserved.
==========================================================================
Opening audio decoder: [faad] AAC (MPEG2/4 Advanced Audio Coding)
AUDIO: 44100 Hz, 2 ch, s16le, 192.0 kbit/13.61% (ratio: 24000->176400)
Selected audio codec: [faad] afm:faad (FAAD AAC (MPEG2/MPEG4 Audio) decoder)
==========================================================================
==========================================================================
Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
Selected video codec: [ffsvq3] vfm:ffmpeg (FFmpeg Sorenson Video v3 (SVQ3))
==========================================================================
Checking audio filter chain for 44100Hz/2ch/s16le -> 44100Hz/2ch/s16le...
AF_pre: 44100Hz/2ch/s16le
AO: [dsound] 44100Hz 2ch s16le (2 bps)
Building audio filter chain for 44100Hz/2ch/s16le -> 44100Hz/2ch/s16le...
Starting playback...
VDec: vo config request - 480 x 260 (preferred csp: Planar YV12)
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [directx] 480x260 => 480x260 Planar YV12

[Liisachan]

Liisachan, i tried installing your mirrored version hosted on http://subforge.s116.xrea.com/freeware/mpc/ but the NSIS installer prompted me with an error

Compared it with the file on the Italian server.
CRC and MD5 match.
Probably it's your downlaod. Please re-download it and tell me the result.

[Shirokuu]

Liisachan, me bad. ^_^;

I tried three times to download ffdshow from your mirror, and the installer kept complaining about broken installer data. Then i got the bright idea to clear my cache and try again. What do you know? It worked...

My download probably got corrupted upon the first download. Not realizing Internet Explorer didn't download a fresh file but instead retrieved the file from my cached internet items stored locally on my own computer, I kept ending up with the same file.

Anyway... With all this cleared up: thanks for your quick reply!

[ExtraEye]

A problem on the new version,ffdshow-20050607, while mpg files are played pink and green lines are displayed.
Problem stops when mpeg1 codec is changed to libmpeg2 from libavcodec.
Also, in post-processing I think there's a problem too.
When automatic quality control is checked the preset power changes all the time from the best you chose to the one before it(if you choose seven it jumps back and forth between 7 and 6 and if you chose 4 it jumps from 4 to 3 and back all the time). - this haven't occured on the last version.

I would like to suggest a nice feature you could consider adding.
Some files are interlaced while others aren't and so I use the deinterlacing filter only on files that need it by specifying them spesificaly on the "Image Settings" tab.
Although that is a great feature, it would be better if I could just add directly to the deinterlacing filter tab the names of the files that need to be deinterlaced(in to some box or something alike).
This way I woun't have to render a special preset for every file that needs to be deinterlaced.

Another suggestion is adding an "else if" part to the options on "resize & aspect" tab.
What I mean is including a feature that resizes by 2, files that have some defined resolution(or between resolutions) while resizing by 3 a different resolution(or between other resolutions).
for example if video1's resolution is smaller than321X181 make it bigger 3 times more, else, if it is smaller than 681X481 make it only twice as big.
I know this can get endless, but just thought you may like it and find a way to make it happen.
I may sound lazy but note these are just a suggestions so don't be mad if you don't like it. I love the job you do on FFDshow.

[DeathTheSheep]

Hm... How would I be able to access the "direct asf output" that ffdshow's WMV9 pass-through codec produces? There isn't an option in the output tab... Where do the frames go when this option is checked?

[Shirokuu]

A problem on the new version,ffdshow-20050607, while mpg files are played pink and green lines are displayed.
Problem stops when mpeg1 codec is changed to libmpeg2 from libavcodec.
Also, in post-processing I think there's a problem too.
When automatic quality control is checked the preset power changes all the time from the best you chose to the one before it(if you choose seven it jumps back and forth between 7 and 6 and if you chose 4 it jumps from 4 to 3 and back all the time). - this haven't occured on the last version.

I think this is a feature not a bug. I have seen the quality slider jump left to right during playback in other versions as well. I reason it's exactly what 'automatic quality control' should do. It throttles the postprocessing power down when the framerate drops below a set mark and throttles it back up when scene complexity is less. It's especially noticable when you select 'spp' with 'automatic quality control' enabled. (Well it's on my not-too-fast computer.)

[ExtraEye]

I know what it's supposed to do
but if I disable the automatic quality control the video doesn't get stuck.
Also note that if I select 7 and it ALWAYS falls down to 6 and back, when I select level 6 shouldn't it stay stable on 6? not to mention, shouldn't it stay stable on 3??
I hope I got my point through, I am pretty sure this is a problem in the build but for now use this feature while automatic quality control is unchecked.
I'm using mplayer's postprocessing and the latter never had this problem for me.

[Shirokuu]

I wasn't fully able to reproduce your problem with mpeg-1 playback. What i DID notice was that libavcodec needs a bit more processor power to decode a mpeg-1 stream than the libmpeg2 decoder library. Postprocessing with the automatic control setting enabled with libavcodec couldn't keep up with the framerate so the slider jumped left to right from 5 to 7, while postprocessing with libmpeg2 was able to keep the slider almost all the time on the 7th setting. Apparently libavcodec takes up so much processing time there is not enough left to dering chroma and luma (on my computer).

I could't reproduce the green lines while using libavcodec though. I used an mpeg1 compliant stream.

But the strange thing is that the 'automatic quality control' seems to overrule the slider in a sense that it even keeps downthrottling the quality settings even when it's set to level 2 or 3. It probably has to do with the postprocessor, because it also occurs in both libavcodec as well as libmpeg2.

[ExtraEye]

yes well switching over to libmpeg2 did solve the problem for me.
I don't really know much about how ffdshow works but I still think the automatic quality control feature is broken.
Like you said, it changes the level of the quality without any good reason, always.
It's a shame because this feature really helps alot on spp deblocking.

[Shirokuu]

I remember there was one build where 'automatic quality control' kept the framerate for 'SPP deblocking' around 25 fps for proper viewing. On my machine the only way to use SPP while mainting a decent framerate is to disable deringing both luma and chroma manually.

[Leak]

I remember there was one build where 'automatic quality control' kept the framerate for 'SPP deblocking' around 25 fps for proper viewing. On my machine the only way to use SPP while mainting a decent framerate is to disable deringing both luma and chroma manually.

Ummm... the "quality" setting for SPP doesn't work like for the other postprocessing methods; while the same slider is used and the checkboxes get checked accordingly, all the slider does is set the amount of processing the SPP deblocker does; each notch up makes it calculate twice as much.

The only deringing option would be the "soft threshold" setting...

np: Chris De Luca And Peabird - Deadly Wiz Da Disko (Deadly Wiz Da Disko)

[Shirokuu]

Ummm okay, you were right. The slider controls the amount of times the DCT blocks get smoothed and averaged. And i remember it doubles with every tick.

:stupid:

[bob0r]

1: start\run\programs\ffdshow\Video decoder configuration\Codecs\Raw video\disabdled (found by: LigH)
2: run ffdshow-20050607.exe\language-ok\"title of ffdshow installer": ffdshow-20050510 (found by: maska)

It seems celtic_druid indeed can use his vacation (but the first is probably not his fault, hehehe)

I myself use ffdshow only to decode H.264 video.
I tested some x264 video i encoded with 8x8dct and they all work great.
Please note for proper .mp4 splitting of the x264 8x8dct file you need an older version of MatroskaSplitter.exe (if you use this), The build from 14 april 2005 i have put online too.

Great works and i hope in the future they will fix all your mpeg1/svq3/typo and other problems!

:thanks:

[celtic_druid]

2 comes from me changing the name of the exe after compiling rather than changing the var in the nsi script.

I actually did a new installer, with the correct title, but then it isn't worth updating all the mirrors just for that is it?

[Liisachan]

I actually did a new installer, with the correct title, but then it isn't worth updating all the mirrors just for that is it? Upto you ;)

[bob0r]

2 comes from me changing the name of the exe after compiling rather than changing the var in the nsi script.

I actually did a new installer, with the correct title, but then it isn't worth updating all the mirrors just for that is it?

Not worth it, as long as you aware of it, and won't do it again! bad celtic_druid, bad!

I am sure more ffdshow builds will come, we shall see the typos fixed then.

People who have real issues with ffdshow are probably thinking: WHAT THE "4 stars" are they talking about! :D

[maska]

Don't forget that option to create shortcuts in Start Menu is also missing in the last build. ;) :D :)

[dinolib2]

Do you know if H.264 deblocking in postprocessing options works fine?
Shold it improve SSIM? I've made few tests and it seems to lost 0,1 points.
Is it right?
I asked because with XviD postprocessing gains some points.

thanks

[Sharktooth]

h.264 has it's own deblocking (inloop filter), so enabling postprocessing is not properly a good idea.

[DeathTheSheep]

Ah. Very interesting indeed.

As for the direct .asf output, how is this possible? I mean, with WMV9 encoding, there is an option for direct .asf output, but where do the frames go? The output tab has no .asf option!

[Ajaja2005]

I try to encode mpeg1/2/4 video with ffdshow-20050607 in "one pass - constant bitrate" mode (for ex. preset "Libavcodec DVD PAL")" (TV-tuner capture). If "Maximun bitrate" in "Ratecontrol" is not equal 0, than ratecontrol don't work :(
Bug?

[dinolib2]

h.264 has it's own deblocking (inloop filter), so enabling postprocessing is not properly a good idea.

ok it's true... why, then, a specific H.264 deblocking option in the latest ffdshow?

[Sharktooth]

It's a sort of choice for enabling or disabling inloop filter when playing back h.264.
However the default is ok.

[bob0r]

Don't forget that option to create shortcuts in Start Menu is also missing in the last build. ;) :D :)

Did you mean ffdshow-20050607.exe ?
Because its working fine here, maybe you hit NEXT too fast twice.
Try uninstall ffdshow and delete the directory ffdshow from program files.
Also manually delete the ffdshow dir/shortcuts from start menu (to see it does work just fine)

If it indeed does not work, provide more information like your Operation System and other system info.

[LotharZ]

Selecting "Info & Debug" option with a movie playing, ffdshow freezes (and the whole movie obv).

Anyone with same problem?

Running Win2003+ffdshow-20050607

[Liisachan]

no problem on my Win2k
http://subforge.s116.xrea.com/ffdshow.png

[iron2000]

ffdshow-20050607 causes slow playback on a Xvid video I have.
The characters are like moving at half speed but sound goes fine.
(Picture slower than sound)

But it plays another Xvid video without smudging.

Xvid seems hard to please.

[maska]

Did you mean ffdshow-20050607.exe ?
Because its working fine here, maybe you hit NEXT too fast twice.
Try uninstall ffdshow and delete the directory ffdshow from program files.
Also manually delete the ffdshow dir/shortcuts from start menu (to see it does work just fine)Oops! :o I see only now the new build has modified the option of creating shortcuts, it's the last step to do before clicking Install.

[Sharktooth]

patched the cabac typo in ffdshow libavcodec: http://www.webalice.it/f.corriga/misc/libavcodec_ff.7z (it fixes the x264 rev254 playback)

[bob0r]

@celtic_druid
libavcodec needs updating again after x264 revision 254, where there was a typo in cabac. Please stick around when libavcodec is updated.
Sharktooth's .dll file works here.
Don't know if that can be included or you have to wait for an official libavcodec update.

@ffdshow developer (milan)
When i uninstall the latest ffdshow by celtic_druid, ff_realaac.dll does not get delete from the ffdshow dir.
Also start\programs\ffdshow is not removed.

Also this typo "start\run\programs\ffdshow\Video decoder configuration\Codecs\Raw video\disabdled" (found by: LigH)


Good luck and please give us x264 playback pleasure!!

[celtic_druid]

Sure it can be included. Just thought it would be quicker to have a working version of libavcodec available until I update ffdshow and easier for me if someone else did it.

[bob0r]

Sure it can be included. Just thought it would be quicker to have a working version of libavcodec available until I update ffdshow and easier for me if someone else did it.

Ok cool when will the new version be online? ;)

[Sharktooth]

i updated libavcodec dll as you can see... so, relax... :)

[Mc Onyx]

I have a question about encoding with X264 in FFDshow, is there a way to encode 2pass, because in settings i only have option to choose 1pass quantizer, or 1pass constant bitrate, there is no 2pass option, like in Xvid...
This is the first time i tried to use FFDshow for encoding, rather than decoding, thank you...

[G-Slide]

ligh> thx, haali's splitter works !!! :thanks:

I've discovered ffdshow+ avisynth 2 days ago... and i use this call :
Import("C:\Program Files\AviSynth 2.5\plugins\LimitedSharpen2.avs")
LimitedSharpen2(ss_x=1.0,ss_y=1.0,Smode=4,aSharpS=0.4,aWThresh=0.99)

it's slow (3400+ my computer), but it's weird cause it's faster on AVC videos than DVD's... :eek: it was supposed to be contrary... anyway, do you know a good call to use on DVD and .mp4 avc ?

[bob0r]

i updated libavcodec dll as you can see... so, relax... :)

I can't relax, i hate installing 400 files just to watch/test a video.
For testing the file is cool, but people want to test x264, not ffdshow :P

The more people are interested the less chance x264 will DIE :scared:

;)

[bob0r]

https://sourceforge.net/tracker/?func=detail&atid=471489&aid=1217036&group_id=53761
Resolution: Fixed

https://sourceforge.net/tracker/?func=detail&atid=471489&aid=1217034&group_id=53761
Resolution: Fixed

https://sourceforge.net/tracker/?func=detail&atid=471489&aid=1217032&group_id=53761
Resolution: Fixed

I did bug report via: http://ffdshow.sf.net/
quote of one of the emails by Milan Cutka (milan_cutka):
Thank you for reporting this. Two bugs in four lines of nsis
script - what a shame :(

:D i guess celtic_druid isn't the only one needing a vacation, hehehe

Good work ffdshow team (Milan) glad to see the bugs gets fixed soon, hope celtic_druid can do one last compile before he leaves ;)

[celtic_druid]

I wouldn't say that I need a vacation.
Apparently this weekend is a long weekend to. No one ever tells me these things.

Yeah, I just did a cvs checkout so I have the updated nsi script.

[Sharktooth]

since sourceforge bugtraking system gone nutz heres the ffmpeg and x264 cabac fixes for ffdshow:

http://www.webalice.it/f.corriga/misc/ffdshow/ff_cabac_fix.diff
http://www.webalice.it/f.corriga/misc/ffdshow/ff_x264_cabac_fix.diff

[Sharktooth]

New FFDShow build is up.

[Liisachan]

Thanks as always, celtic_druid :)

http://m17n.cool.ne.jp/freeware/mpc/
http://subforge.s116.xrea.com/freeware/mpc/

[bob0r]

Good work celtic_druid, i hope we can use this version for a while now :D

Also i noticed:
ffdshow-20050527.exe 2.30 MB (2,421,859 bytes)
ffdshow-20050607.exe 2.84 MB (2,981,024 bytes) <-- quite a lot bigger
ffdshow-20050609.exe 2.37 MB (2,485,305 bytes)

So now we got 3 ffdshow fixes (+ maybe more in latest cvs) libavcodec.dll updated AND we have the right size file again.
Also i could enjoy 2 hours of the sun today, what more could a man want? :eek:


P.S.
mirrored \/

[celtic_druid]

20050527, ffdshow.ax = MSVC7.1
20050607, ffdshow.ax = ICL8.1
20050609, ffdshow.ax = MSVC7.1

This is the reason for the size difference.

[TheShadowRunner]

I just though of an enhancement for the ffdshow Audio codec that would be awesome : decoding QDM2 audio via the QT "hack" that both mplayer and VLC use : placing qtmlclient.dll and quicktime.qts in the windows\system folder.

If this is doable then we could have FFDshow video dec to decode SVQ3 and the FFDshow Audio dec decode QDM2 and adios ressource-hungry-slow-as-hell Quicktime Player!!

Well, 0.5US$
Later,

TSR

[DeathTheSheep]

Yeah, QT sure is one heclc of a slow loader. I think that idea is definatly useable! I'm surprised that no one had thought of this before. :D

[Revgen]

I went to bob0r's site and tried to download ffdshow-20050609.exe, but I ended up downloading ffdshow-20050610.exe instead.

What's new in this version?

[Didée]

Just to have it mentioned:

ffdshow's Huffyuv coder seems to have the same buffer problem/bug that older versions of original Huffyuv codec had - exception if a compressed frame gets bigger than the input frame.

Today I had the codec die on decoding an extremely complex frame (much detail plus tons of mpeg artefacts) it had codec before.

Unless this gets fixed, ffd-Huffyuv is not safe to use.

[celtic_druid]

No idea what ffdshow-20050610 is.

http://mirror05.x264.nl/celtic_druid/ffdshow-20050611icl.exe
http://mirror05.x264.nl/celtic_druid/ffdshow-20050611msvc71.exe

ffdshow with ICL8.1 came out even bigger this time. Not sure what kind of a speed difference there is between ffdshow.ax ICL8.1 and MSVC7.1.

[Liisachan]

fix for that huffyuv thing?
mirrored anyway...
http://subforge.s116.xrea.com/freeware/mpc/
http://m17n.cool.ne.jp/freeware/mpc/
ffdshow-20050611.exe = ffdshow-20050611icl.exe
and
ffdshow-20050611msvc71.exe

FYI, betanews (randomly?) picked up 20050609 yesterday
http://fileforum.betanews.com/detail/FFDShow_MPEG4_Video_Decoder/1054056131/1

[celtic_druid]

Once again. Is the problem present in mencoder or ffmpeg? Because it is possible that it is a libavcodec bug or it could be just ffdshow.

[Liisachan]

Um, I just wondered "Whoa, yet another new build...! What was changed? Maybe a fix for that huffyuv thing posted above?"
I guess the answer is no and it's not like I personally need that fix either.

Thanks again for your dedicated work.

[LigH]

If you would like to check for yourself, you may use any source (even a looped still) and gradually add more and more random noise:
LoadPlugin("BlockBuster_2.5_WarpEnterprises.dll")

function SquareNoise(clip c, int dev)
{
return BlockBuster(c, method="noise", detail_min=1, detail_max=100, variance=dev*dev/10)
}

Animate(v, 0, v.Framecount-1, "SquareNoise", 0, 100)
I once used such a function for codec quality comparisons.

[IgorC]

FFDSHOW(11 of june) Audio Decoder has a bug with reproducing 3gp HE-AAC2 .Audio codec http://home.cwru.edu/~bes7/db_EnhAACPlus_1.2.zip . It's not correct just rename 3gp to mp4 but foobar plays fine.
First time it plays just fine but if press STOP in Media Player Classic and return to play it sound changes from stereo 44.1 khz to mono 22Khz and SBR part of HE-AAC isn't decoded.

Audio sample with clear Stereo sound http://sr2.mytempdir.com/50107

Graphedit doesn't show nothing strange :
sample.mp4 (Audio out) -> In FFdhsow audio decoder Out -> Audio input pin (rendered) Default direct show sound device

Bug appears only during playback of audio files. As sound track to video file it plays fine.

[iNFO-DVD]

Does anyone ever get a problem with NERO whilst ffdshow is installed?

If I'm writing a DVD (data) with one or more AVI files whilst ffdshow is installed NERO either locks up or crashes.

It's to do with the preview that NERO insists on doing when you select an AVI via the 'ADD' button.

My friend also gets the same problem. Is there any fix for this?

[celtic_druid]

Maybe you should try faad instead of realaac.

[iron2000]

Is the postprocessing faulty again?
I always use the same video to test the codec.

The video and audio is unsynced when postprocessing is on.
It played well on the 27-May-2005 build though.

Info on video from GSpot:
http://img226.echo.cx/img226/1285/vidprop9sx.png


Hmm..

I tried playing it with Koepi Xvid latest beta codec.
Plays fine without postprocessing but unsynced with post processing.

Is it a file problem?

[Tuesday]

is anyone else having trouble with the www.aziendeassociate.it site?

Every file i get off it seems to be b0rked :S

Can anyone provide a mirror please ?

[Sharktooth]

Yup, i mantain the site and it's constantly overloaded...:(
Yesterday 540 concurrent connections was reached...

[iron2000]

Heres a mirror for ffdshow:
http://m17n.cool.ne.jp/freeware/mpc/
theres links for other mirrors in there too.

[Revgen]

I go to bob0r's site (http://x264.nl) if I can't get it from Sharktooth's server.

[Rash]

If I'm writing a DVD (data) with one or more AVI files whilst ffdshow is installed NERO either locks up or crashes.
Does that happen with any ffdshow version or just this 11th June one? Because I have never had this problem. But I didn't test with the latest ffdshow release yet.

[Yong]

the jun 11 2005 version of ffdshow seems have problem with YUY2 colospace with "resize & aspect" filter.
ive tried play XviD(XviD dshow decoder) and MPEG 1 video(ms mpeg 1 decoder),
then i force ffdshow decode raw video by using YUY2 colorspace,
only the luma method:simple/warped resize dun have problem.
heres the pic output by ffdshow:
http://img117.echo.cx/img117/4044/yuy24wf.png

[wata]

i have a avi with a corrupt frame 0
all version after ffdshow-20050406.exe crash the player with this fault

Faulting application mplayerc.exe, version 6.4.8.4, faulting module libavcodec.dll, version 0.0.0.0, fault address 0x001fe3ef.

[vortex_hl]

@celtic_druid
icl9.0 released.
will you compile ffdshow with icl9.0? what are you (and milan) thinking about it?

[celtic_druid]

http://www.intel.com/software/products/compilers/cwin/whatsnew.htm
Not sure that there would be much of a difference other than for dual core CPU's.

[LigH]

The new "Cinderella" MPEG4-AVC-HP trailer from Ateme (http://forum.doom9.org/showthread.php?p=672441#post672441) does not play in ffdshow or Mplayer. It seems that they used features which are not yet supported in libavcodec:

MPlayer dev-CVS-050609-00:54-3.4.4 (C) 2000-2005 MPlayer Team
CPU: Advanced Micro Devices Duron Spitfire (Family: 6, Stepping: 1)
Detected cache-line size is 64 bytes
CPUflags: MMX: 1 MMX2: 1 3DNow: 1 3DNow2: 1 SSE: 0 SSE2: 0
Compiled for x86 CPU with extensions: MMX MMX2 3DNow 3DNowEx


85 audio & 199 video codecs
Playing cinderella-ateme-3000k-hp.mp4.
ISO: Unknown File Type Major Brand: avc1
QuickTime/MOV file format detected.
--------------
Warning! pts=1 length=0
MOV track #0: 1 chunks, 0 samples
Generic track - not completely understood! (id: 0)
--------------
Warning! pts=1 length=0
MOV track #1: 1 chunks, 0 samples
Generic track - not completely understood! (id: 1)
--------------
Warning! pts=1484398397 length=0
MOV track #2: 72 chunks, 3559 samples
MOV: AVC decoder configuration record atom (49)!
MOV: Found unknown movie atom btrt (20)!
Image size: 1280 x 720 (24 bpp)
Display size: 1280 x 720
Fourcc: avc1 Codec: 'AVC Coding'
--------------
MOV: longest streams: A: #-1 (0 samples) V: #2 (3559 samples)
VIDEO: [avc1] 1280x720 24bpp 23.976 fps 0.0 kbps ( 0.0 kbyte/s)
==========================================================================
Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
Selected video codec: [ffh264] vfm:ffmpeg (FFmpeg H.264)
==========================================================================
Audio: no sound
Starting playback...
[h264 @ 00A3A468]custom scaling matrix not implemented
[h264 @ 00A3A468]non existing SPS referenced
Error while decoding frame!

P.S.: I still didn't find out for sure where to report this "issue" (not quite a bug, right?). Mplayer and FFmpeg are related to the libavcodec library, but who/where are the "core developers"?

[Cyberace]

didn't find out for sure where to report this "issue" (not quite a bug, right?)... ...who/where are the "core developers"?FFmpeg bug-reports:
http://ffmpeg.sourceforge.net/ffmpeg-bugreport.html

MPlayer bug-reports:
http://www.mplayerhq.hu/DOCS/HTML/en/bugreports.html
http://bugzilla.mplayerhq.hu

(I believe the "core developers" are the same as for x264 codec (http://www.videolan.org/x264.html) and also FFmpeg lead-developers (http://sourceforge.net/project/memberlist.php?group_id=16082))

[Sharktooth]

SPS is not mandatory for AVC decoders, so ffmpeg should simply ignore it (or add support for SPS), so this is a bug.
For what concerns the custom matrices they are in the HP specs, so they should be implemented to be compliant. This is a missing feature but since the decoder is not compliant to the standard it can be considered a bug.

[LigH]

I was just unsure: Both ffmpeg and Mplayer are based on libavcodec, but there is no special libavcodec buglist.

I sent a mail to the ffmpeg developer mailing list, that seemed most suitable to me.

[Inventive Software]

libavcodec is part of the FFMPEG project. So if you want to bitch about libavcodec, talk to the guys at FFMPEG.

[Rash]

http://www.intel.com/software/products/compilers/cwin/whatsnew.htm
Not sure that there would be much of a difference other than for dual core CPU's.
Well, it supports the 64-bits instructions, doesn't it? But I don't think ffdshow's case is as simple as compiling it in 64-bits. :cool:

[multiblitz]

Well, does this mean that compiling ffdshow simply with Intel V.9 compiler gives us a multi-threaded ffdshow ? To good to be true !!!

[LigH]

I don't think so. Multiple threads must be handled by one "base" thread, must be synchronized. This requires code in the source. A compiler might optimize multi-threading for optimal performance on real multi-processor or multi-core systems, but the thread forking and control must be present in the source already.

Multi-Threading routines will still work even with only one CPU, anyway - but only time-sliced.

[madman1980]

What's up with the SPP Deblocking in Postprocessing anyway? I thought that when i upgraded to A64 3800+ I could run it, but still as slow as ever. No one can use this thing. :angry:

[Shinobu]

SPP Deblocking is ok 720*576 on my athlon 2000+ (set up 2 (H+V luma), more than 2 is ways to blurry) and mpc take about 85% cpu with it

++

[madman1980]

Ok, that worked better. All the six boxes were marked by default. Still a little slow though, especially with soft treshold.

[bob0r]

New ffdshow build needed :eek:
lossless playback! http://forum.doom9.org/showthread.php?t=95962
:thanks: :sly:

[celtic_druid]

Don't think that libavcodec in ffdshow has been updated yet.

[Egh]

Bug or feature?

(finally able to post in the forum after initial 5 days:)

So -- is this bug or feature?
In ffdshow one can enable visualisation of motion vectors and quantizers. If libavcodec is used, everything is ok. If xvid is used, only quantizers will be shown.

[Egh]

2 celtic_druid:

also, could you please explain what presicely "Fix luminance level" option does?

[bob0r]

Don't think that libavcodec in ffdshow has been updated yet.

Check again please, sorry i was a bit hasty :o

[celtic_druid]

http://mirror05.x264.nl/celtic_druid/force.php?file=./ffdshow-20050619.exe

[LigH]

Yes! You guys are tight at the heartbeat of development!

[Liisachan]

When I first downloaded it from the above URL, the filesize was 2480344, but it installed fine. I downloaded the same file several times by several methods (to make sure I'll mirror the correct file), and I think the original filesize is 2480343. I have no idea why the filesize was different when I first downloaded it...

[celtic_druid]

2,480,343 bytes.

[Liisachan]

k, thanks.

[moonman]

I have mirrored it here:
http://svartling.hopto.org/index.php?q=node/view/1039

[bob0r]

Good work.
The lossless encodes i did all work!

Mirrored also, ofcourse :cool:

[FredThompson]

@celtic_druid,

How about creating a shortcut to the localized help file during installation?

[LigH]

And since x264 now supports custom scale/quant matrices, I hope that the ffmpeg team will update their routines as well; MP4 files created by x264 with "--cqm jvt", for example, play well with the Ateme beta decoder filters.
__

I just found a strange bug in the FourCC registration on my system!

Although I registered ffdshow's VfW interface to decode FFV1, applications like VDub(Mod) tell me they cannot find a decompressor for 'ffds' or 'FFV1'. Reason:

[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Drivers32]
vidc.ffds = ":ff_vfw.dll"

Please note the leading ':' colon !

I'm not sure if this was a mistake which was introduced in earlier versions. So I uninstalled and re-installed ffdshow (2005-06-19), and watched the registry -- the colon did not appear again. Who knows...

So if someone else wonders why he cannot decode a video encoded by ffdshow's own VfW codecs, he might want to look at this registry entry.

[Sharktooth]

sorry for the delay, i've been busy but finally i mirrored ffdshow too.

[Egh]

AVIsynth routine bugs (?)

If avisynth script is used inside ffdshow while playing video, then unchecking box "add ffdshow as video source" is followed by an instant playback stop and ffdshow+mpc shutdown. Quite inconvinient, could a fool-proof protection be implemented here?

Also, if "Avisynth" box in ffdshow is enabled (despite box "add ffdshow as video source" being checked!), and mpc is restarted by user, then _randomly_ ffdshow refuses to start (the moment mpc window appears after double-clicking on a video file it closes again). If this happens, the workaround is to go directly to ffdshow's configuration dialog and uncheck Avisynth box. Then you may double-click the same file, it starts nicely. And the very same avisynth script works again after being enabled (so it's certainly not a particular script problem). I'd thought that happens cause during initiation of videoplayback ffdshow doesn't always provide usable videosource to avisynth but tries to use it (prooly leads to internal exception of some sort, but no messages are shown). Maybe caused by increased CPU load too (several filters enabled).

Also, requesting info on filters priveleges in ffdshow. For example, if i have aWarpsharp, bicubic resizer and avisynth script simultaneously enabled, is resizer the last one to be executed ?

[vinetu]

Egh,
Right now ,there is some incompatibility between Avisynth and winXP Service Pack 2 ,thanks to m$...
- if something is wrong with the *.AVS script the host application will just disappear instead of showing the error message .

Try to put in ffdshow some "wrong proof" commands like this:

code:

ShowSMPTE(25)

and see what happens ...

[Shirokuu]

I encountered a bug in the june 19th FFDSHOW compile. Playing a particular MKV (XVID/VORBIS) with bsplay or zoom player works perfectly. But as soon as i browse the playing video to any spot on the timeline the movie starts to stutter. Audio and video run out of sync every few sync. The strange thing is that OSD shows that bitrate is dropping in a slow steady pace from 2000 down to null. While playing the same MKV from the start gives a nice almost steady bitrate of around 1900~2000.

I don't know for sure if the problem is related to FFDSHOW, since I use the latest Haali splitter and the OGG directshow filter to decode the audio.

EDIT: OOPS, problem fixed. Using FFDSHOW to decode the Vorbis audio stream as well fixed the problem. I will leave the post here if people encounter the same problem.

[Rash]

Hey guys. I just noticed the "2/2 quadro" preset on the Mixer (audio filter) doesn't relay the LFE channel to the front R, L anymore. And it used to. Does anyone know why? I used to like it. :(

No more LFE relay to the .0 presets » http://img147.echo.cx/my.php?image=ffdshowaudio6yq.png

ffdshow release ffdshow-20050619.exe

[pankov]

I'm having problems with the latest build (20050619).
The playback of DTS wav files using libdts as a decoder is horrible on my NF3 onboard ALC850 soundcard. The sound is VEEEERY loud and corrupt althogh I can guess what's been played so it do decode it correctly. So this makes me think that the problem is in ffdshow itself not libdts. When I use the 20050303 version everything is perfect.
can someone confirm this or this is only with my setup?

[pankov]

I found what was causing the problem:
32 bit floating point output
I knew that I have heard the same horrible sound before and after changing different settings I found it
:)
The problem was solved when I disabled it.

I suppose that it was intentional to change the priority of the different output sample formats and make the 32bit float the default one in the new version. But I don't think it's a good idea because as the note bellow it sais there can be driver incompatabilities.

[Defiler]

I've never read through the ffdshow source code, but I'm wondering how hard it would be to split the decoder functions and postprocessing functions into two separate modules.
At this point, it's extremely difficult for testers to help pinpoint the cause of a particular playback problem, because ffdshow is so feature-rich.
If there were a little checkbox called "Connect to post-processing module?", you could disable all of those code paths and easily verify that changes to one of the post-processing features hadn't broken something else.

Just a thought.

[ExtraEye]

but if you just disable all the features wouldn't it work just as well? I mean, what's the point of seperating it?

[Defiler]

You'd think so, but take a look at how many posts there are in this thread along the lines of "New version improves feature X, but now I can't play files encoded with codec Y."

Sometimes life is simple enough that your code can look like:
if(blue_widget_checked) { tweakMagicKnob(z); }

Other times, for various reasons, you might be stuck doing:
z = 0
tweakMagicKnob(z);
In theory, both should do "nothing".. but if a bug has just crept into "tweakMagicKnob", any process that passes through it may be affected.

There's a reason that all the video card drivers in the world don't come in one big file called "video_card_drivers.exe". Heh.

[DeeGee]

Was someone from the x.264 team doing custom quantizer support for h.264 in libavcodec? I'd like to use ffdshow to watch all my encodings...

[LigH]

Look into the libavcodec CVS (http://www1.mplayerhq.hu/cgi-bin/cvsweb.cgi/ffmpeg/libavcodec/?cvsroot=FFMpeg&sortby=date) for yourself if someone committed an according patch recently: As far as I can see while writing this reply - not yet...

[foxyshadis]

I encountered a bug in the june 19th FFDSHOW compile. Playing a particular MKV (XVID/VORBIS) with bsplay or zoom player works perfectly. But as soon as i browse the playing video to any spot on the timeline the movie starts to stutter. Audio and video run out of sync every few sync. The strange thing is that OSD shows that bitrate is dropping in a slow steady pace from 2000 down to null. While playing the same MKV from the start gives a nice almost steady bitrate of around 1900~2000.

I don't know for sure if the problem is related to FFDSHOW, since I use the latest Haali splitter and the OGG directshow filter to decode the audio.

EDIT: OOPS, problem fixed. Using FFDSHOW to decode the Vorbis audio stream as well fixed the problem. I will leave the post here if people encounter the same problem.
I had this problem with the radlight vorbis codec in the 6/11 and 6/19 release too, even when I didn't seek, always at the same points. (Drove me insane when I couldn't get it to use ffdshow or any other vorbis decoder without uninstalling.) It occured only in mkvs (several versions of haali splitter), ogms played fine. Apparently some incompatibility creeped in.

[nanoflower]

I've seen something like that (stuttering picture) recently with the recent celtic_druid build. What I see is when I have two copies of MPC open watching two different videos they both start to stutter using ffdshow. It seemed that the problem was with the libavcodec. If I changed the decoder for Xvid videos to be Xvid instead of libavcodec the problem went away when I restarted the clips.

[namchik]

Does any1 know when Milan is going to release his new builds?
his latest build of March, 3 was good and fast... but old =(

[Liisachan]

"Waiting for GCC 4.0 (http://ffdshow.sourceforge.net/tikiwiki/tiki-read_article.php?articleId=16&PHPSESSID=0a56077b3918dfedd0daff1c43d56009)" is related?


I'm waiting for GCC 4.0 to be released to make it one of compilers supported by ffdshow build process again.

[namchik]

+-- GCC 4.0 branch created ------+
| (Feb 25 2005) \
| v
GCC 4.1 Stage 1 (ends April 25 2005) GCC 4.0.0 release (April 20 2005)
| \
| v
GCC 4.1 Stage 2 (ends July 8 2005) GCC 4.0.1 release
|
v
GCC 4.1 Stage 3
|
:
v

http://gcc.gnu.org/develop.html#future
So GCC 4.0 wasn't released on April 20 ? =(

[Musafir_86]

-Does anyone know how well the support is it for Ogg Vorbis decoding in ffdshow? Is it better than CoreVorbis or any decoder based on libvorbis 1.1.1 (or at least same - but with lower CPU usage)? :confused:

-But..., are there any new DS decoder based on libvorbis 1.1.1? :stupid:

Thanks.

-Musafir_86.

[celtic_druid]

Last time I tried ffdshow compiled with gcc 4.0 it just crashed.

Radlight's filters I believe include a vorbis decoder.

By vorbis decoding in ffdshow do you mean libavcodec or tremor?

[Musafir_86]

-Of course I mean libavcodec; isn't Tremor dated back to year 2003 then? Also, it's based on fixed-point method, right?

-Musafir_86.

[Liisachan]

RadLight's filters and VSFilter don't mix well, and so not suitable for softsubbed OGMs, but RadLight's filters even can decode Ogg Theora via DS nicely.
Yet another Ogg DS decoder is by illiminable's
http://www.illiminable.com/ogg/

[madman1980]

Ffdshow really covers all my needs, but one really annoying thing is MPEG2 aspect ratio. It is almost always off when I play files in MPC. SVCD is not played with the right aspect ration, neither is PAL or NTSC vobs (dvd spec). I have to adjust it manually.

My solution has so far been to install Windvd in addition to ffdshow. But it is kind of a waste, since Ffdshow should be able to detect aspect ratio too, if it was just given some priority.

Any ideas if this will be fixed?

[eb]

use VLC player , then you are independent of M$ system, very important when playing back multiaudio files, MPC plays them all at the same time

[madman1980]

Sure, VLC is good. But i prefer MPC by far. FFdshow has made everything work except this, so I hope the developers can learn something from the WinDVD MPEG2 decoder with regards to defaulting to the correct aspec ratio.

[celtic_druid]

New build: http://mirror05.x264.nl/celtic_druid/force.php?file=./ffdshow-20050630.exe
ICL9

[Musafir_86]

@Celtic_Druid,

-Can you always post the changelog for ffdshow? Or at least bundle together with the binary package; this will help to know whether someone specific problem was solved or not.

Thanks.

-Musafir_86.

[celtic_druid]

Changelog?
http://cvs.sourceforge.net/viewcvs.py/ffdshow/ffdshow/?sortby=date#dirlist

[Musafir_86]

-Thanks, but wouldn't it be nicer if they come together, so one don't have to browse the CVS each time a build is up. If you don't have time, then it's okay...

Musafir_86.

[celtic_druid]

Well someone has to browse the cvs, because as far as I know there is no simple changelog, otherwise I would just copy and paste it.

[Liisachan]

@celtic_druid
I downloaded the file 3 times from the .nl server, 2 times from your server.
The file from the .nl server is always 1 byte larger (3192268 Bytes), it has extra 0x20 at the end of the file. The same happened to me for the pervious build. Maybe firefox is not good at downloading files??????

I mirrored the 'smaller' one (3192267 Bytes)

[celtic_druid]

Don't know. Maybe the server? ftp shows that the file on their is 267bytes, so it isn't an upload issue.

[celtic_druid]

Ok changelog since last release:

fixed ICL 9 compilation (ffdshow.ax compiled with ICL9 to)
updated libavcodec
faster h264 deblocking,
updated x264 - custom quantization matrices (not supported by libavcodec yet)
saving of JM quant tables
removed min_order and max_order,
GCC hash_map implementation, the old one was buggy
adaptive FFT length in convolver depending on impulse file number of samples
SSE2 avisynth yuy2 <-> yv12 conversion
SSE2 bitmap blending and soften
SSE2 colourize
SSE2 gradual denoise
SSE2 hue and saturation
SSE2 luminance offset and gain
SSE2 mplayer noise
SSE2 unsharp
SSE2 warpsharp bump
recreate configuration pages when changing presets
working on direct communication between filters and configuration pages
trying to fix HW overlay filter

taken from cvs log -rHEAD -d"lastbuild<thisbuild". Not quite as easy as a cut and paste.

[JoeBG]

Ok changelog since last release:

fixed ICL 9 compilation (ffdshow.ax compiled with ICL9 to)
updated libavcodec
faster h264 deblocking,
updated x264 - custom quantization matrices (not supported by libavcodec yet)
saving of JM quant tables
removed min_order and max_order,
GCC hash_map implementation, the old one was buggy
adaptive FFT length in convolver depending on impulse file number of samples
SSE2 avisynth yuy2 <-> yv12 conversion
SSE2 bitmap blending and soften
SSE2 colourize
SSE2 gradual denoise
SSE2 hue and saturation
SSE2 luminance offset and gain
SSE2 mplayer noise
SSE2 unsharp
SSE2 warpsharp bump
recreate configuration pages when changing presets
working on direct communication between filters and configuration pages
trying to fix HW overlay filter

taken from cvs log -rHEAD -d"lastbuild<thisbuild". Not quite as easy as a cut and paste.


Thanks a lot :)

When are you visiting europe? Didn´t you plan holidays? :)

[celtic_druid]

No, didn't plan anything, but yeah I leave Monday. So ~4 days 2 hours.


2005-06-19

another try on fixing b frame delay
common function to search for pin interface
fixed DwString::sEmptyBuffer initialization,
less explicit casting in binds and options lists
md5sum OSD item,
more comptrQ,
show coded frame size in OSD,
simple right-click menu in codecs list
updated x264 and libavcodec - lossless h.264
using function pointer in TomsMoComp
writing OSD info to file,
moving binds to Twindow (can't be compiled now)
working on direct communication between configuration pages and filters

2005-06-11

attempt to use libavcodec computed PTS when available
better unicode subtitles support
cabac fixes
don't create keyboard, remote and tray icon threads when in explorer.exe,
merged OSD and Info,
removed IDFF_max constant,
remove ff_realaac.dll and start menu shortcuts on unistall,
updated minilzo
updated vc6 project

2005-06-09

always display levels curve,
don't remove all characters surrounded by { },
fixes in separate luma and chroma resizer
new info panel,
makefiles update and italian setup translation by Gianluigi Tiesi
offer all font charsets (detected are displayed bold in combobox)
updated libavcodec,
updated x264,
zones support for x264,
working on full unicode subtitles support

2005-06-07

autoincrement build version
autoload preset on decoder match
included baseclasses can be compiled by GCC without modifications
include stripped baseclasses library
fixed GCC ffvdub compilation
ffdshow can be compiled by GCC: image filters are unoptimized, crashes immediatelly at start
QueryInterface -> comptrQ
support for reading theora from RadLight ogg splitter
support for cross-compilation by Gianluigi Tiesi
updated libavcodec
cosmetic,
trying to compute correct font width for italic subtitles,

[LigH]

I have some small issues as well: b0bors server tells the MIME type "application/force-download". That's some nonsense, in my opinion: *.exe files usually shall be sent using "application/octet-stream".

Most probably due to this strange MIME type, Opera 7.54 suggested to save this file with the extension "*.ace" (WinAce archive)?! Don't know why Opera once registered this extension to this MIME type - I had to edit this list entry...

[yaz]

Sure, VLC is good. But i prefer MPC by far. FFdshow has made everything work except this, so I hope the developers can learn something from the WinDVD MPEG2 decoder with regards to defaulting to the correct aspec ratio.
may i suppose u use mpc w/ffdshow for svcd/dvd playback ? if so, how do u manage to force ffdshow to be the mpeg2 decoder. i've tried tricking in mpc a/o in ffdshow but the only way is killing all the possible candidates (nero, windvd, pdvd, aso, aso)

as regards AR.
- u set 'use overlay mixer' in 'output' and 'overlay a.r.' to default in 'resize...', don't u ?
- does it work w/other stream types ? u may find some anamorphic example files here in the forums w/which u can test it.
just ask, because that part worx correctly for me.

thx
y

[Leak]

Well someone has to browse the cvs, because as far as I know there is no simple changelog, otherwise I would just copy and paste it.

Well, this could help with the recent changes:

http://cia.navi.cx/stats/project/ffdshow

np: Markus Guentner - Sleep Well (Audio Island)

[celtic_druid]

Speaking of leaks, I noticed earlier that there is a leak in subtitles reader. I'll see about putting up a new build.

[Sharktooth]

mirrored.

[iron2000]

Why does the last version install at 6.89mb and the new version install at 11.77mb?
Not that I'm pressed for space, just curious.
(A bit off 10mb is ok but don't go to 20+mb)

[Alvy]

Latest Ffdshow audio decoder doesn't work with reclock together. After downgrading to the last version it works again.

By the way thanks for this great program !!!!

[ViCroié]

Is this me, or ffdshow?

I'm using ffdshow-20050630 (celtic) and i wanted to use ffdshow to be able to deinterlace video from my TVcard (YUV2 Interlaced) in Media Player Classic.

I used the AVIsynth feature from ffdshow to do this. and deinterlacers like: TomsMoComp and FieldDeinterlce work. The problem comes when i use a BOB deinterlacer (wich is the best way of watching this footage imho). When i enable it it will only show 25 of the 50 FPS. When i do thesame thing in avisynth itself it works great.

[celtic_druid]

2005-07-01

fixed few possible crashes when processing small images
fixed memory leak in subtitles reader,
positional controls apply to embedded bitmap subtitles
Italian translation by Dario Brama
smoothing method and scale apply to embedded bitmap subtitles,
SSE2 32-bit to 16-bit integer samples conversion
SSE2 original and avih noise
SSE2 SPP store_slice

Probably the last build from me for a month.

[Sharktooth]

CD, enjoy your holydays and thanx for your builds :)

[yaz]

@celtic_druid
waving good-by is always sorrow, but u made it a bit easier ;) have a nice holiday !
thx a lot
y
ps. the climate is a bit 'armageddonian' here in europe (it's pouring like hell, actually), so don't forget your life-belts :D

[Yama4050242]

maybe a bug,
both ffdshow-20050630 ffdshow-20050701 crash on playback clip encoded with megui(x264 cli build 270) setting subpixelrefinement RDO and ME Algo Exhausitive

ffdshow-20050619 is fine for playback

[Leak]

I used the AVIsynth feature from ffdshow to do this. and deinterlacers like: TomsMoComp and FieldDeinterlce work. The problem comes when i use a BOB deinterlacer (wich is the best way of watching this footage imho). When i enable it it will only show 25 of the 50 FPS. When i do thesame thing in avisynth itself it works great.

I fear this is a limitation of ffdshow - it doesn't notice if you change the framerate, and thus doesn't propagate it down the DirectShow filter graph...

I took the liberty to file a bug on it... (http://sourceforge.net/tracker/index.php?func=detail&aid=1231101&group_id=53761&atid=471489) ;)

np: Luke Vibert - Gwithian (Lover's Acid)

[madman1980]

maybe a bug,
both ffdshow-20050630 ffdshow-20050701 crash on playback clip encoded with megui(x264 cli build 270) setting subpixelrefinement RDO and ME Algo Exhausitive

ffdshow-20050619 is fine for playback

Thanks for letting us know. I'll stay with the .19 version until there is one out that doesnt seem to introduce new problems.

[celtic_druid]

Just tested a clip with RDO and Exhausitive ME and it played back fine here.

[Yama4050242]

@celtic_druid
http://mail.e8007.com/viewsharenetstore.php?sessid=2765442c51873b4700&userid=snow4050242&filename=L3NhbXBsZS9zYW1wbGUucmFy&isdir=0

the above link is small parts from my crashed clip.the file sever is not stable, hope you can get my file

[celtic_druid]

Looks like it is probably an ICL9/VC6 issue. The build from the 19th was compiled with MVC7.1.

[bob0r]

@celtic_druid

I assume you got the test file?
Can you upload the video file to mirror05 (maybe latest ffdshow too)?

And before you go siesta, compile ffdshow with all possible tools and we will figure out which is best :sly:

[Palikrovol]

OT: CD if you come to Madrid, think here we are at 36 degrees at 15:42 and sometimes at 28-30 at 23:00-24:00 (Hotter if you go to the south, Sevilla for instance 38 degrees at 16:00) so do no bring too much clothes with you :).

Happy hollydays here in HOT Spain

E=MC^2
España = Mucho Calor^2 :)

[celtic_druid]

Yeah, I left the test file downloading over night and somehow managed to get it.
MSVC7.1 and ICL9 builds are up. ICL9 crashes on the sample, but maybe is faster?

Packed a couple of pairs of shorts and tshirts/singlets. So should be ok.

[wata]

the OSD remaining time doesn't work for all builds after ffdshow-20050527.exe
it turn to 00:00:00 before movie end.

[Dark_Angel_PT]

I have submitted this request at Sourceforge. What do you think of this?

http://sourceforge.net/tracker/index.php?func=detail&aid=1231905&group_id=53761&atid=471492

[Defiler]

Anyone feel like making a non-SSE/SSE2 build?
I'd do one myself, but I haven't had a lot of luck getting the ASM parts working in Visual Studio.

[LigH]

ffdshow by celtic-druid works well on my AMD Duron (which is not SSE capable).

[Defiler]

ffdshow by celtic-druid works well on my AMD Duron (which is not SSE capable).
Really? That's good news. Does it melt when you enable things like denoise3d, that have been SSE2-optimized?

[vlada]

I fear this is a limitation of ffdshow - it doesn't notice if you change the framerate, and thus doesn't propagate it down the DirectShow filter graph...

I took the liberty to file a bug on it... (http://sourceforge.net/tracker/index.php?func=detail&aid=1231101&group_id=53761&atid=471489) ;)

np: Luke Vibert - Gwithian (Lover's Acid)

I did some tests with FFDShow and it's deinterlace filters some time before. No one of them is able to produce double framerate (in player or in VirtualDub). Also the quality is rather poor and they aren't very fast as well. I tried some deinterlacers from DSclaer and they didn't work as expected neither. I would like to see at least one smart bob filter, which would be able to deinterlace to 50 fps realtime on 2,0 GHz CPU. But I don't know if I'm not asking too much.

If anyone's interested, I did a comparison of different deinterlace filters. It is located at http://jech2.webz.cz/gallery/index.php. I think I already posted the link here some months ago. My commentaries are written in czech, but the pictures say usually enough.

[LigH]

Does it melt when you enable things like denoise3d, that have been SSE2-optimized?
:D Melt - no. But it "crumbles", e.g. SPP deblocking is impossible. And -- Denoise3D? Where is that one hidden?!
__

Ah - found it... No, doesn't crash. I'd guess that SSE optimizations are used when possible, but probably either skipped or applied with the unpotimized version if not.

[Px]

And what about of ffdshow build for Windows x64?

[multiblitz]

One question: Why needs the latest celtruid version 10% more cpu load (Athlon 64 3500@2400mhz) that the normal sse2-build from Oktober 12th ?

[Isochroma]

Using MPC 6.4.8.4, all builds of ffdshow with matroska container. Aspect ratio embedded in matroska container for the video stream is totally ignored if the video stream is decoded by ffdshow (any supported codec). Using the real codec (ie. xvid, etc.) aspect is shown fine. Why?

[LigH]

Do you have "use overlay mixer" enabled in the "output" register?

[Isochroma]

Using all video rendering modes in MPC, same problem. Maybe it's MPC?

[LigH]

Sorry: I mean, in ffdshow!

Because the AR is correct with other decoders, the reason cannot be MPC, it must be that you did not set up ffdshow correctly.

[Foreigner999]

I get a lot of stuttering in a mkv with AVC1 video track and A_VORBIS audio track with one ssa track and one sub(i think) track.

FFdshow july 3, 2005.MSVC7.1

I fooled around a bit and it seems ffdshows audio decoder when set to tremor or libavcodec seems to make the video and audio stutter with OggDS 0.9.9.5. I checked the corevorbis project site and saw they released corevorbis 1.1 today so i tried that and upgraded OggDS to 0.9.9.6 and it seems to have fixed the suttering problem. I tried again to switch back to ffdshow's internal decoder for vobis (both settings) and it still stutters with OggDS 0.9.9.6. :o

Is this by design or is this a bug?

[bond]

I fooled around a bit and it seems ffdshows audio decoder when set to tremor or libavcodec seems to make the video and audio stutter with OggDS 0.9.9.5. I checked the corevorbis project site and saw they released corevorbis 1.1 today so i tried that and upgraded OggDS to 0.9.9.6 and it seems to have fixed the suttering problem. I tried again to switch back to ffdshow's internal decoder for vobis (both settings) and it still stutters with OggDS 0.9.9.6.oggds doesnt support mkv. how did you use it with mkv if not the vorbis decoder (which you said you didnt use)

[Liisachan]

Foreigner999:
This might be totally unrelated to your case, but fyi,
there is a known problem like that related to Ogg Vorbis audio: The jitter/stuttering problem may occur when the Audio output is Default WaveOut,
and problem is gone if the output is Default DirectSound.
(You can switch them in MPC at least.)

As another note, using OggDS 0996 for OGM is unadvisable. (There is a bug,
and only one chanel is decoded for stereo.) 0995 is much more stable and preferable for OGM. (*)

bond:
Unlike 0995, OggDS 0996 was divided into VorbisDS.ax and OggDS.ax.
VorbisDS.ax is a general Vorbis decoder, which works with MKV too.

(*) EDIT: This bug affects Vorbis in MKV too.

[LigH]

And what about CoreVorbis, or ffdshow (which I prefer)?

BTW: Brand new CoreVorbis version available, using current libraries.

[bond]

bond:
Unlike 0995, OggDS 0996 was divided into VorbisDS.ax and OggDS.ax.
VorbisDS.ax is a general Vorbis decoder, which works with MKV too.sure, but if you read his thread than you will see that he writes that he has problem with oggds when using matroska and ffdshow's vorbis decoder and corevorbis as vorbis decoder

and i wondered how this all should have something to do with oggds obviously

[Foreigner999]

sure, but if you read his thread than you will see that he writes that he has problem with oggds when using matroska and ffdshow's vorbis decoder and corevorbis as vorbis decoder

and i wondered how this all should have something to do with oggds obviously
Thanks Liisachan that seemed to have been the problem (isn't this long standing problem fixable? ).

Oh and bond the confusion is because i failed to mention in my original post that i was having crashes in TCMP latest version with a standard .ogg music when using a 192kbit .ogg with a libVorbis date of 20040920. Seems with OggDS 0.9.9.6 and Corvorbis 1.1 it causes a crash in OggDS by either stopping the music or letting it come to an end. Sorry im not all too coherent that late in the night.

EDIT: happens with ffdshows audio decoder as well.

[wata]

the OSD show remaining time function still faulty with the 03Jul version

all builds after ffdshow 27May2005 are faulty

counter become 00:00:00 halfway before movie END.

[vlada]

Foreigner999> Did you try OGG splitter from Radlight or Illiminable? The OggDS filters are know for causing a lot of problems and AFAIK, they are no more in development for a long time.

[WilliamC]

Think someone could start compiling these since Celtic Druid is on vacation?

[DeathTheSheep]

Has something happened to the old AAC decoder?

I heard rumors of there being a bit of turmoil in this area-- I've heard of the aac ip interest holder (VA liscensing) inciting a bit of rukous over compilations of libfaad (and those libraries like it).


Is this really true? I mean, there's nothing greiviously wrong with "realaac," but it doesn't recognize or decode raw AAC files like previous versions of ffdshow. And it certainly doesn't seem to support parametric stereo (so there is no stereo playback of files encoded in AACPlusv2).

I veiw this as quite a problem. Any suggestions? Or is it over for ffdshow in this dept.?

[Foreigner999]

Foreigner999> Did you try OGG splitter from Radlight or Illiminable? The OggDS filters are know for causing a lot of problems and AFAIK, they are no more in development for a long time.

Yes i just tried it now, Radlight Oggsplitters v1.0.0.2. With rads own decoder it seems to work fine with the .ogg file. With ffdshow it also seems to work fine with the same .ogg file. Although when it comes to Corevorbis latest release seeking is very slow and it seems that there is a problem seeking towards the end of the file because it turns to silence.

Is OGM and it's OGG counterpart cursed to never work right with all the features or did it's developers abandon the project? By the way any developement still underway on the radlight splitters/filters?

[bond]

Foreigner999 , if you think there is a bug than tell the devs about it (oggds is not developed anymore, but corevorbis, radlight and ffdshow are)

[Liisachan]

Technically this is not related to this thread, but FYI: RadLight splitter cannot call VSFilter (iirc) and that is critical for playing back softsubbed OGM. On the other hand, RadLight splitter can play OggFile (Ogg Theora) nicely.

[Oki]

I am playing arround with HE-AAC v2 (AKA aacPlus v2 or AAC+SBR+PS). I am testing ffdshow audio decoder with the official MPEG4audio SBR+PS conformance streams (ftp://mpaudconf:adif2mp4@ftp.iis.fhg.de/guests/mpeg4audio/incoming/testSequences/mpeg4audio-conformance/compressedMp4/new_sbr_ps/) from Fraunhofer IIS and they can not be played. I am using ffdshow-20050703 MSVC 7.1 and MatroskaSplitter 11/07/2005 for MP4 splitter. CoreAAC works.

Any idea?

[bond]

people also call it ps-aac ;)

btw its backwards compatible, every aac player can play ps-aac
must be another problem if ffdshow cant decode it. are you sure you have the (correct) parser installed?

[madman1980]

Silly question, I know, but I can't for the life of me figure it out now. I have a AC3 stream that I want to play with ffdshow audio on stereo speakers. I can't make it downmix to stereo, the voices are very, very low. Where do i set it to output in stereo?

[Palikrovol]

Silly question, I know, but I can't for the life of me figure it out now. I have a AC3 stream that I want to play with ffdshow audio on stereo speakers. I can't make it downmix to stereo, the voices are very, very low. Where do i set it to output in stereo?

In "FFDshow Audio decoder configuration" enable 'mixer' and use "2/0 - stereo" in 'output speaker configuration'.

You can also enable 'volume' to boost the volume.

hope it helps.

[madman1980]

That did the trick. Thanks!

[Oki]

people also call it ps-aac ;)

btw its backwards compatible, every aac player can play ps-aac
must be another problem if ffdshow cant decode it. are you sure you have the (correct) parser installed?
I have tried with Matroska and 3ivX D4 splitters with ffdshow audio decoder. 3ivX is the only one capable of playing the full lenght of the files when used with it's own AAC decoder. ffdshow can not render anything in any case and CoreAAC is hanging after playing a couple ot times one of those conformance files.

RealOne does not decode even the SBR part but plays the full lenght of the files.

Which parser do you think is the best for connecting ps-aac in MP4 container to ffdwhow?

[DeathTheSheep]

Heh, I just made an "AAC problems" post a few back, lol.

Has something happened to the old AAC decoder?

I heard rumors of there being a bit of turmoil in this area-- I've heard of the aac ip interest holder (VA liscensing) inciting a bit of rukous over compilations of libfaad (and those libraries like it).


Is this really true? I mean, there's nothing greiviously wrong with "realaac," but it doesn't recognize or decode raw AAC files like previous versions of ffdshow. And it certainly doesn't seem to support parametric stereo (so there is no stereo playback of files encoded in AACPlusv2).

I veiw this as quite a problem. Any suggestions? Or is it over for ffdshow in this dept.?

Apparently, the same thing happens with me. FFDshow simply wouldn't pick up the PS in the HE-AAC v2 stream.

So I renamed the file to a .3gp and disabled ffdshow's "realaac" AAC decoding. Nero's filter took over and it worked perfectly. So obviously it is a flaw in ffdshow.

[Sharktooth]

new ffdshow build by CD is available: http://www.aziendeassociate.it/cd/ffdshow/ffdshow-20050727.exe (right click -> save as..., if you have problems dowloading it use IE or wait some minutes hoping the server load will be lower...)

[Liisachan]

WB celtic :)

mirrored:
http://m17n.cool.ne.jp/freeware/mpc/
http://ffdshow.xrea.jp/

[esby]

mirrored too.

esby

[moonman]

Mirrored here too:
Link (http://svartling.hopto.org/index.php?q=2005/jul/ffdshow-20050727-released)

[multiblitz]

One question to the ffdshow-fan-group: I played recently a HDTV-copy of Star Wars II in avi-fomat using ffdshow. Runs beatifully, but when I try to switch in ZP the audio stream from english to german, the audio stream starts from the very beginning of the movie, while the picture goes on. As ZP does not remember which stream to use after I press stop, the movie always starts in english and the switching results always in an out-of-sync of sound and picture. Any ideas how to resolve this ? Is this a known ffdshow-bug ?

[celtic_druid]

I just tried a dual stream avi using Media Player Classic with ffdshow to decode both audio streams. Switched audio multiple times and it was fine.

[multiblitz]

Thanks for testing. What is your result with ZP ?

[celtic_druid]

Didn't try ZP as I don't have it on my system.

[Ajaja2005]

Where is ff_libfaad2.dll in build by CD? Is realaac better than faad2?

[Tuesday]

Hey, i use FFDSHOW to decode just about everything i watch on my media PC and i hate having to reconfigure the settings for different files.

The settings i reffer to are the post-processing options and the resizing options.

I would like to see these options have some more "intellegence" to them.

The problem i have is that i have a large collection of files, some backed up using pretty old codecs/techniques and thus have a fairly variable video quality/resolution. This means i use the post-processor filter quite often, but i hate having to tweak the stength of the filter for different sets of files ie, when watching a fairly poor quality files the PP would be used quite heavily, but if the next file viewed is a nice high quality one then the details are killed by the high PP settings set up for the previous file.

Would it be reasonable possible for the post-processor to detect the level of compression artifacts in the video stream and adjust the strength of its filtering automatically? i assume for Xvid/Divx this could be done by detecting the repeating block pattern or oversharpened ringing of edges.

Ranges for this would be nice would be nice to allow users to accomodate for their own tastes, ir some kind of generic quality factor for the video stream could be calculated and for set ranges of quality set strengths of PP could be used.

By the strength of the filtering i dont just be the various parts of post processing(horiz luma deblock etc) but actually the strength to which they are applied.


The second problem is that different backups have been compressed with different resolutions, not all of which scale up perfectly to fill the screen, meaning i use the resizing filter in FFDSHOW. This is all well and good until after having the filter setup to correct the aspect of an almost 1.33:1 file, but if the next is basically 2.35:1 or 1.78:1 then it will be incorrecly resized to 1.33:1.

What i would like is a more advanced version of the "resize if" tab, so you could specify "If the AR is within this range, resize to this AR, or if it is this range, resize to this AR" etc etc. 3 ranges would do it for me, one to correct almost 1.33:1 files, one for ~2.35:1 files and one for ~1.78:1 files, but it would be nice to be able to add as mant ranges as you like.

If anyone could implement these features i would be eternaly greatful as i think others may find them useful for the same reasons as me:) and it would mean less getting off the couch to play with FFDSHOW settings and sadly i am not a programmer so cannae do it myself :(

[ExtraEye]

a problem i have with this latest version is that the presets for image settings "forget" their loading conditions.
is just me?
also it seems like it either takes more cpu than the last version or works worse on my computer.

[Liisachan]

ffdshow-20050803.exe is mirrored courtesy of Celtic_Druid.
http://m17n.cool.ne.jp/freeware/mpc/
http://ffdshow.xrea.jp/

[moonman]

Mirrored here too (http://svartling.hopto.org/index.php?q=2005/aug/ffdshow-20050803-released) (My Streamload account can be blocked by too much traffic sometimes).

[PatchWorKs]

Have you heard about the next version of FairUse ?

It can use ffVFW for encoding ! I succesfully backuped a movie in 2-pass theora !

Check it out !!!

FairUse Wizard 2 - Release Candidate 1 (http://www.dvd-guides.com/component/option,com_simpleboard/Itemid,61/func,view/id,11718/catid,26/) 3ad @ DVD-guides.com's forum.

[buzzqw]

mirror here too ffdshow 03/08/2005 Direct Download Link (http://www.64k.it/andres/modules.php?name=Downloads&d_op=getit&lid=188)

no download bandwith problem.

BHH

[Sharktooth]

it's on the main mirror too

[vlada]

One question to the ffdshow-fan-group: I played recently a HDTV-copy of Star Wars II in avi-fomat using ffdshow. Runs beatifully, but when I try to switch in ZP the audio stream from english to german, the audio stream starts from the very beginning of the movie, while the picture goes on. As ZP does not remember which stream to use after I press stop, the movie always starts in english and the switching results always in an out-of-sync of sound and picture. Any ideas how to resolve this ? Is this a known ffdshow-bug ?
I don't think it is a bug in FFDShow. It is most likely caused by an AVI splitter or a stream switcher. 2 sound tracks in AVI is a hack, you should try converting the file to Matroska format using MKVToolnix. Or you can try Haali's Media Splitter (AKA MatroskaSplitter) with AVI support. FFDShow is just a decoder, it shouldn't be responsible for a synchronisation problem.

And now according to FAAD2 - I heard there are some problems with this decoder, because Nero bought it and the current license is not compatible with GPL. You can find more about it in Hydrogenaudio forums.

[LigH]

2 sound tracks in AVI is a hack
Absolutely not! AVI can handle several audio streams. It just gets several "auds" chunks. Analyse such an AVI with RIFFWALK or abcAVI, you won't notice anything strange.

[Shirokuu]

Absolutely not! AVI can handle several audio streams. It just gets several "auds" chunks. Analyse such an AVI with RIFFWALK or abcAVI, you won't notice anything strange.

I was in the same understanding of AVI containers as being not able to handle more than one audio track as Vlada was. Of course, I believe LigH but a question comes to mind then: What's the disadvantage of using multiple audio streams in an AVI container then? There must be a reason for so few people using the AVI container to mux several different streams. I, for instance, have never seen such an AVI. (Not that it proves not to exist ;) )

[esby]

Mostly because you won't need it.
If you follow the avi spec, you can have an avi with several video, several audio and several other streams.
Now your problem starts when you try to play such avi file.
Playing will be a problem, since most players will only allow one video stream and maybe several video stream.
For the non audio/video, sub for example, you can have filter that do the job.
Now let's suppose you can code and exploit the format specification, this won't be an hack, since it's in the format, but it won't be supported by most of the player, since they don't consider all of the specifications as useful, and they are right in a way,
who would say that three video streams in an avi is this useful?

Now for several audio, I think this may work, with a stream switcher, after all, you only listen to one of the stream usually. Now, including several audio, means usually, to be include the subbed version also, it could work, but it is easier to use another format for that.

esby

PS: And of course, you'll find the usuall trolls about the 'corrupted' avi with vbr mp3, better to use another format here, since the compatibility is problematic depending the player used.

[LigH]

More details:

- AVI was not made to contain VBR audio; it is possible (see Alexander Noe's documentation about AVI Myths), but several splitters or decoders fail synchronising, or even the whole playback, after a time jump.

- The stupid Microsoft Windows Media Player is unable to play only one stream and switch between them, instead it plays every stream at the same time. The "Morgan Stream Switcher" was made to solve this problem, but now that much better filters and splitters exist, the MMSwitch is not useful anymore and shall be avoided if your media player can handle it.

[Shirokuu]

Thanks for the thorough explanation. I've just googled on Alexander Noe's document on AVI, MKV and OGM containers and it's an excellent read. If other people reading this dev thread are interested, you can find it here (http://www-user.tu-chemnitz.de/~noe/Video-Zeug/containers.pdf).

[Doom9]

this thread is too vast and contains too many subject. In the future, please open new threads about specific issues in the appropriate forum. E.g. if you have a problem with ffdshow's decoding of x264, use the MPEG-4 AVC forum, if you have a problem with Snow, use this very forum, if you have a problem with a specific container, use the container forum, for audio en/decoding issues, use the audio forum, for subtitles, use the subtitle forum. Thank you for your cooperation.