ffdshow development #2 [Archive] - Page 4

Shirokuu

7th June 2005, 09:23

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

7th June 2005, 09:31

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

7th June 2005, 09:48

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

7th June 2005, 10:47

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

7th June 2005, 11:25

Leak

7th June 2005, 11:54

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

7th June 2005, 12:04

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

7th June 2005, 12:34

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

7th June 2005, 13:06

Liisachan

7th June 2005, 13:10

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

7th June 2005, 19:09

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

7th June 2005, 19:20

Don't forget that option to create shortcuts in Start Menu is also missing in the last build. ;) :D :)

dinolib2

7th June 2005, 20:16

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

7th June 2005, 20:18

h.264 has it's own deblocking (inloop filter), so enabling postprocessing is not properly a good idea.

DeathTheSheep

7th June 2005, 20:31

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

7th June 2005, 20:44

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

7th June 2005, 21:11

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

7th June 2005, 21:17

It's a sort of choice for enabling or disabling inloop filter when playing back h.264.
However the default is ok.

bob0r

7th June 2005, 21:33

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

7th June 2005, 23:29

Selecting "Info & Debug" option with a movie playing, ffdshow freezes (and the whole movie obv).

Anyone with same problem?

Running Win2003+ffdshow-20050607

Liisachan

8th June 2005, 01:06

no problem on my Win2k
http://subforge.s116.xrea.com/ffdshow.png

iron2000

8th June 2005, 07:15

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

8th June 2005, 08:54

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

8th June 2005, 13:17

patched the cabac typo in ffdshow libavcodec: http://www.webalice.it/f.corriga/misc/libavcodec_ff.7z (it fixes the x264 rev254 playback)

bob0r

8th June 2005, 13:25

@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

8th June 2005, 14:17

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

8th June 2005, 14:35

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

8th June 2005, 16:49

i updated libavcodec dll as you can see... so, relax... :)

Mc Onyx

8th June 2005, 17:24

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

8th June 2005, 17:52

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

8th June 2005, 18:26

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

9th June 2005, 12:13

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

9th June 2005, 12:41

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

9th June 2005, 13:24

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

9th June 2005, 14:52

New FFDShow build is up.

Liisachan

9th June 2005, 15:09

Thanks as always, celtic_druid :)

http://m17n.cool.ne.jp/freeware/mpc/
http://subforge.s116.xrea.com/freeware/mpc/

bob0r

9th June 2005, 16:09

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

9th June 2005, 16:24

20050527, ffdshow.ax = MSVC7.1
20050607, ffdshow.ax = ICL8.1
20050609, ffdshow.ax = MSVC7.1

This is the reason for the size difference.

TheShadowRunner

10th June 2005, 02:00

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

10th June 2005, 04:58

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

10th June 2005, 21:48

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

11th June 2005, 02:14

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

11th June 2005, 04:26

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

11th June 2005, 05:00

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

11th June 2005, 05:18

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

11th June 2005, 05:49

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

11th June 2005, 06:01

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

12th June 2005, 02:34

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

12th June 2005, 04:08

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

12th June 2005, 04:35

Maybe you should try faad instead of realaac.

iron2000

12th June 2005, 13:38

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

12th June 2005, 13:46

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

12th June 2005, 14:09

Yup, i mantain the site and it's constantly overloaded...:(
Yesterday 540 concurrent connections was reached...

iron2000

12th June 2005, 14:10

Heres a mirror for ffdshow:
http://m17n.cool.ne.jp/freeware/mpc/
theres links for other mirrors in there too.

Revgen

12th June 2005, 17:20

I go to bob0r's site (http://x264.nl) if I can't get it from Sharktooth's server.

Rash

12th June 2005, 19:41

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

12th June 2005, 20:43

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

14th June 2005, 01:35

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

14th June 2005, 18:39

@celtic_druid
icl9.0 released.
will you compile ffdshow with icl9.0? what are you (and milan) thinking about it?

celtic_druid

15th June 2005, 01:34

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

15th June 2005, 10:10

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

15th June 2005, 13:35

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

15th June 2005, 13:39

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

15th June 2005, 13:56

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

15th June 2005, 14:54

libavcodec is part of the FFMPEG project. So if you want to bitch about libavcodec, talk to the guys at FFMPEG.

Rash

15th June 2005, 15:30

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

15th June 2005, 19:38

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

15th June 2005, 20:12

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

16th June 2005, 10:55

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

16th June 2005, 12:58

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

16th June 2005, 20:49

Ok, that worked better. All the six boxes were marked by default. Still a little slow though, especially with soft treshold.

bob0r

17th June 2005, 11:17

New ffdshow build needed :eek:
lossless playback! http://forum.doom9.org/showthread.php?t=95962
:thanks: :sly:

celtic_druid

17th June 2005, 11:57

Don't think that libavcodec in ffdshow has been updated yet.

Egh

17th June 2005, 15:18

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

17th June 2005, 15:23

2 celtic_druid:

also, could you please explain what presicely "Fix luminance level" option does?

bob0r

18th June 2005, 20:01

Don't think that libavcodec in ffdshow has been updated yet.

Check again please, sorry i was a bit hasty :o

celtic_druid

19th June 2005, 10:44

http://mirror05.x264.nl/celtic_druid/force.php?file=./ffdshow-20050619.exe

LigH

19th June 2005, 11:03

Yes! You guys are tight at the heartbeat of development!

Liisachan

19th June 2005, 11:30

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

19th June 2005, 11:55

2,480,343 bytes.

Liisachan

19th June 2005, 12:06

k, thanks.

moonman

19th June 2005, 12:12

I have mirrored it here:
http://svartling.hopto.org/index.php?q=node/view/1039

bob0r

19th June 2005, 16:55

Good work.
The lossless encodes i did all work!

Mirrored also, ofcourse :cool:

FredThompson

19th June 2005, 20:46

@celtic_druid,

How about creating a shortcut to the localized help file during installation?

LigH

20th June 2005, 07:38

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

20th June 2005, 13:01

sorry for the delay, i've been busy but finally i mirrored ffdshow too.

Egh

20th June 2005, 23:17

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

20th June 2005, 23:42

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

22nd June 2005, 14:11

Rash

24th June 2005, 03:26

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

26th June 2005, 19:18

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

26th June 2005, 19:27

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

27th June 2005, 15:57

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

27th June 2005, 21:28

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

27th June 2005, 21:37

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

28th June 2005, 04:31

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

28th June 2005, 07:07

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

28th June 2005, 09:26

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

28th June 2005, 20:21

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

29th June 2005, 06:27

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

29th June 2005, 07:05

"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

29th June 2005, 11:33

Musafir_86

29th June 2005, 14:31

-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

29th June 2005, 15:45

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

29th June 2005, 16:24

-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

29th June 2005, 16:26

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

29th June 2005, 20:23

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?

29th June 2005, 21:36

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

29th June 2005, 22:45

celtic_druid

30th June 2005, 01:09

New build: http://mirror05.x264.nl/celtic_druid/force.php?file=./ffdshow-20050630.exe
ICL9

Musafir_86

30th June 2005, 01:21

@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

30th June 2005, 01:44

Changelog?
http://cvs.sourceforge.net/viewcvs.py/ffdshow/ffdshow/?sortby=date#dirlist

Musafir_86

30th June 2005, 01:53

-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

30th June 2005, 02:08

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

30th June 2005, 02:27

@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

30th June 2005, 02:36

Don't know. Maybe the server? ftp shows that the file on their is 267bytes, so it isn't an upload issue.

celtic_druid

30th June 2005, 04:28

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

30th June 2005, 05:21

celtic_druid

30th June 2005, 05:33

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

30th June 2005, 07:06

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

30th June 2005, 08:24

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

30th June 2005, 10:19

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

30th June 2005, 12:10

Speaking of leaks, I noticed earlier that there is a leak in subtitles reader. I'll see about putting up a new build.

Sharktooth

30th June 2005, 12:57

mirrored.

iron2000

30th June 2005, 16:20

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

30th June 2005, 18:56

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

1st July 2005, 12:48

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

1st July 2005, 12:54

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

1st July 2005, 12:56

CD, enjoy your holydays and thanx for your builds :)

yaz

1st July 2005, 14:19

@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

1st July 2005, 15:24

Leak

1st July 2005, 18:15

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

1st July 2005, 19:30

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

2nd July 2005, 02:37

Just tested a clip with RDO and Exhausitive ME and it played back fine here.

Yama4050242

2nd July 2005, 16:56

@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

3rd July 2005, 03:11

Looks like it is probably an ICL9/VC6 issue. The build from the 19th was compiled with MVC7.1.

bob0r

3rd July 2005, 14:18

@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

3rd July 2005, 14:49

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

3rd July 2005, 15:09

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

5th July 2005, 12:45

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

5th July 2005, 17:11

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

6th July 2005, 13:56

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

6th July 2005, 14:12

ffdshow by celtic-druid works well on my AMD Duron (which is not SSE capable).

Defiler

6th July 2005, 14:32

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

6th July 2005, 17:18

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

6th July 2005, 20:04

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.

7th July 2005, 15:53

And what about of ffdshow build for Windows x64?

multiblitz

7th July 2005, 18:33

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

7th July 2005, 19:11

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

7th July 2005, 19:38

Do you have "use overlay mixer" enabled in the "output" register?

Isochroma

7th July 2005, 20:41

Using all video rendering modes in MPC, same problem. Maybe it's MPC?

LigH

7th July 2005, 21:46

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

10th July 2005, 03:09

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

10th July 2005, 12:38

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

10th July 2005, 14:01

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

10th July 2005, 14:09

And what about CoreVorbis, or ffdshow (which I prefer)?

BTW: Brand new CoreVorbis version available, using current libraries.

bond

10th July 2005, 14:14

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

10th July 2005, 20:00

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

15th July 2005, 15:35

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

15th July 2005, 17:50

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

15th July 2005, 18:53

Think someone could start compiling these since Celtic Druid is on vacation?

DeathTheSheep

16th July 2005, 23:16

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

17th July 2005, 12:53

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

17th July 2005, 13:05

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

17th July 2005, 13:15

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

18th July 2005, 10:23

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

18th July 2005, 19:57

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

18th July 2005, 21:01

Palikrovol

18th July 2005, 21:13

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

18th July 2005, 21:30

That did the trick. Thanks!

Oki

19th July 2005, 00:59

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

19th July 2005, 02:23

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

27th July 2005, 12:29

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

27th July 2005, 13:23

WB celtic :)

mirrored:
http://m17n.cool.ne.jp/freeware/mpc/
http://ffdshow.xrea.jp/

esby

27th July 2005, 14:37

mirrored too.

esby

moonman

27th July 2005, 17:41

Mirrored here too:
Link (http://svartling.hopto.org/index.php?q=2005/jul/ffdshow-20050727-released)

multiblitz

30th July 2005, 19:32

celtic_druid

30th July 2005, 19:44

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

30th July 2005, 19:50

Thanks for testing. What is your result with ZP ?

celtic_druid

30th July 2005, 20:19

Didn't try ZP as I don't have it on my system.

Ajaja2005

30th July 2005, 21:45

Where is ff_libfaad2.dll in build by CD? Is realaac better than faad2?

Tuesday

31st July 2005, 12:25

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

31st July 2005, 15:31

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

4th August 2005, 02:12

ffdshow-20050803.exe is mirrored courtesy of Celtic_Druid.
http://m17n.cool.ne.jp/freeware/mpc/
http://ffdshow.xrea.jp/

moonman

4th August 2005, 19:22

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

5th August 2005, 09:16

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

5th August 2005, 13:05

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

5th August 2005, 15:59

it's on the main mirror too

vlada

7th August 2005, 22:40

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

8th August 2005, 12:21

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

8th August 2005, 15:45

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

8th August 2005, 16:12

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

8th August 2005, 17:57

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.