ffdshow development #2

[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.