ffmpeg2theora with derf's theora codec support

[bond]

j^ has ported his ffmpeg2theora tool to be able to be used with derf's theora encoder which makes use of the full features theora offers and not only the ones which were also there in vp3 (as supported in the reference theora encoder)

the sources can be found here: ffmpeg2theora-exp (http://trac.xiph.org/cgi-bin/trac.cgi/browser/experimental/j/ffmpeg2theora-exp/) and derf's encoder (http://trac.xiph.org/cgi-bin/trac.cgi/browser/experimental/derf/theora-exp/)

anyone wanting to make a compile for testing this? :)

[ak]

You can get one here: http://4nykey.nm.ru/bin/ffmpeg2theora-svn.exe

[bond]

thx, but are you sure that this is with derf's encoder? did you do the compiles?

[ak]

Well, yes, it says 'ffmpeg2theora-exp 0.13-svn' and oggs it produces decodable only with libtheora-exp. If I upped the right one that is :)

[bond]

thanks a lot! :)

[hellfred]

I just cant get ffmpeg2theora to crop my clip
:stupid:
ffmpeg2theora -h states:
ffmpeg2theora-exp 0.13-svn

usage: ffmpeg2theora-exp [options] input

Output options:
--output,-o alternative output
--width, -x scale to given size
--height,-y scale to given size
--aspect define frame aspect ratio: i.e. 4:3 or 16:9
--crop[top|bottom|left|right] crop input before resizing
...
Here a list of commands that I have tried, all in vain.

::ffmpeg2theora --crop=8:8:8:8 -o out.ogg prodigy90_med.mpg
::ffmpeg2theora -croptop 32 -cropbottom 32 -cropleft 32 -cropright 32 -o out.ogg Suzuka.avi
::ffmpeg2theora -o out.ogg Suzuka.avi
::ffmpeg2theora -i Suzuka.avi -croptop 32 -cropbottom 32 -cropleft 32 -cropright 32 -o out.ogg
::ffmpeg2theora --crop 8|8|8|8 -o out.ogg Suzuka.avi > out.txt
::ffmpeg2theora --crop 8 | 8 | 8 | 8 -o out.ogg Suzuka.avi > out.txt
::ffmpeg2theora --crop [8 | 8 | 8 | 8] -o out.ogg Suzuka.avi > out.txt
::ffmpeg2theora --crop [8|8|8|8] -o out.ogg Suzuka.avi > out.txt
::ffmpeg2theora --crop=[8|8|8|8] -o out.ogg Suzuka.avi > out.txt
::ffmpeg2theora --crop=[8 | 8 | 8 | 8] -o out.ogg Suzuka.avi > out.txt
::ffmpeg2theora --crop 8:8:8:8 -o out.ogg Suzuka.avi > out.txt
ffmpeg2theora --crop=8:8:8:8 -o out.ogg Suzuka.avi > out.txt
PAUSE


E:\Rechner\Neu\Import\ffmpeg2theora>ffmpeg2theora --crop=8:8:8:8 -o out.ogg Suzu
ka.avi 1>out.txt
ffmpeg2theora: ambiguous option -- crop
Input #0, avi, from 'Suzuka.avi':
Duration: 00:24:03.8, start: 0.000000, bitrate: 1018 kb/s
Stream #0.0: Video: mpeg4, yuv420p, 640x480, 23.98 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Resize: 640x480
Resample: 48000Hz => 44100Hz
0:00:00.04 audio: 0kbps video: 7480kbps
The output shows ffmpegg2theora`s typical error message, if ffmepg2theora starts up at all. As one can see, the source is modolo 16, and encoding without cropping works just fine. But I want to encode some TV-captures with bad edges that need to be cutted badly.
Can anybody please enlighten me, how to set cropping values.

Hellfred

EDIT:
Googling for "ffmpeg2thora crop" brought up the solution:

ffmpeg2theora --croptop 16 --cropbottom 16 --cropright 32 --cropleft 16 -o out.ogg Suzuka.avi > out.txt

[bill_baroud]

how's the quality compared to "reference" theora ?

[hellfred]

how's the quality compared to "reference" theora ?
Sorry, that i do not know, as i never used that one. But just try yourself. Win32-binaries are available for both and the usage of ffmpeg2theora is quite simple. I have to hunt bugs right now. :)

Hellfred

[bond]

hellfred, i also struggled with its crop option and this is how it works:
--croptop 82 --cropbottom 78 aso :)

how's the quality compared to "reference" theora ?i would say its worse than the reference one

but derfs decoder is both faster and complete

[hellfred]

Yes, last night I encoded the same clip already used for testing derfs encoder with standardt ffmpeg2theora, featuring the mmx-enhanced reference encoder. Derfs encode was a little bit sharper (only set quality with -v 10 to max) but had much more block artefacts e.g. in the sky and more ringing. Source was a vob that needed to be deinterladed (-d).

[PatchWorKs]

Why not porting these 'enhanced' features to ffvfw ?

...FairUse (http://www.dvd-guides.com/component/option,com_simpleboard/Itemid,61/func,view/id,11493/catid,26/) is going to become the 1st DVD Backup tool that encodes by it. :eek:

[dimzon]

how about AviSynth support???

[hellfred]

Hi
I have made yet anoter test comparing the two ffmpeg2theora. This time the source was a video clip, captured with my analog tv-card. The Source has only low quality, was captured als lossless MJPEG and was being properly deinterlaced croped and resized (modolo16) before encoding. Deinterlacing was done using kerneldeint.
Again, derf showed much more blocks as the mmx-enhanced reference encoder.

Hellfred

[bond]

how about AviSynth support???you can feed ffmpeg2theora with .avs via akupenguin's avs2yuv tool

Why not porting these 'enhanced' features to ffvfw ffvfw is not existing anymore

and ffdshow already includes the theora reference encoder and can even output to .ogg! (yeah .ogg, not .ogm)

[PatchWorKs]

ffvfw is not existing anymore

What you mean ?
It's present in the last ffdshow build. :confused:

[celtic_druid]

Well not exactly. There is a VfW wrapper called ff_vfw, but ffvfw as a project as said doesn't exist any more... There is just ffdshow.

[PatchWorKs]

oh, well... you mean they merged. I know.

[Sirber]

you can feed ffmpeg2theora with .avs via akupenguin's avs2yuv toolAny way to feed ffmpeg2theora with .avs and have sound? :)

[celtic_druid]

Well I guess you could pipe from avs2wav to ffmpeg.

[Sirber]

Is there a way to pipe video and audio from .avs at the same time? ;)

[celtic_druid]

Just having a look at Mobilehackerz changes and it looks like it supports audio. Just need to apply them to ffmpeg2theora and you are set.

[Sirber]

to avs2yuv? I don't understand

[celtic_druid]

No I am talking about ffmpeg2theora supporting avs input directly.
Just build a current cvs ffmpeg with AVISynth input. I'll see about an ffmpeg2theora build.

[Sirber]

I'm not familiar with compiling using mingw. I'll be pretty happy if you can make a ffmpeg2theora (v0.16) build with avs input :D

[celtic_druid]

get_ffmpeg_cvs.sh
autogen.sh
make

http://celticdruid.no-ip.com/xvid/
Those who don't have access will have to wait for it to be mirrored.

[Sirber]

I can't find "ffmpeg2theora-avs.7z" in the sections nor mirrors. Where did you place it?

Thanks for your efforts! :D

[celtic_druid]

It is under Misc or rather will be when it gets mirrored. At the moment it is on my server, but not publically available.

[Sirber]

Could you please PM me a link? I'd like to try ASAP.
I wanna use it with FreeAnime :)

[Sirber]

C:\Projets\FreeAnime\output\ffmpeg2theora.exe -v 3 -S 2 --optimize -a -2 -H 44100 "D:\Tests\1.avs"

Doesn't work. Says:
The Vorbis encoder could not set up a mode according to
the requested quality or bitrate.

-a -1 work though.
-a -2 was working in 0.16.

About AVS, it works :D

Input #0, avs, from 'D:\Tests\1.avs':
Duration: 00:00:23.8, start: 0.000000, bitrate: 0 kb/s
Stream #0.0, 23.98 fps(r): Video: rawvideo, yuv420p, 640x480
Stream #0.1: Audio: pcm_s16le, 48000 Hz, stereo, 1536 kb/s
Resize: 640x480
Resample: 48000Hz => 44100Hz

[celtic_druid]

Different libvorbisenc I guess. Not all versions do -2. Can't remember what lib I have installed. Guess I should update to the latest aotuv. Don't think it would work with lancer without some mods.

edit:
pkg-config --modversion vorbisenc returns
1.1.3svn

[Sirber]

That's the latest? 1.1.3

[celtic_druid]

Latest is 1.1.2 I believe. So yeah I had a current svn build. Think it is as simple as Xiph source not doing -2. Although didn't they merge an earlier aotuv? Think -2 has been supported since aotuv beta3 or so.

[Sirber]

I've seen -2 in the latest oggdropxp and -2 is avalible in belight.

Sad :(

[celtic_druid]

That would be the aotuv_b4.51 version of oggdrop right?

Anyway, I linked against aotuv b4.51 so -2 is now available.

[Sirber]

Many thanks!!! :D

same place? ;)

[Sirber]

Works! :D

Good work!!! :D

[PatchWorKs]

That would be the aotuv_b4.51 version of oggdrop right?

Check out Ogg Vorbis acceleration project (http://homepage3.nifty.com/blacksword/index_e.htm) too... :cool:

[celtic_druid]

Don't think it would work with lancer without some mods.

Still I might investigate.

[Sirber]

The website don't explain much. What is Lancer about?

[pest]

What is Lancer about?

sse & sse2 optimized. about 60%-90% faster.

[Sirber]

sse & sse2 optimations. about 60%-90% faster.That is something nice :D

[Selur]

does anyone know if there are plans for 2pass encoding with theora?

[bond]

does anyone know if there are plans for 2pass encoding with theora?ffdshow supports it already

[Sirber]

CLI is better than VFW. Could be cool to have it in ffmpeg2theora .

[bond]

there are so many things that would be cool, its that someone has to code it ;)

[Sirber]

not sure I could do it with php and delphi ;)

[Sirber]

any way to put avisynth support in ffmpeg2theora?

[celtic_druid]

Read above. Already been done. I used Mobilehackerz patch, so video and audio is supported.

[Sirber]

seems I have ffmpeg2theora 0.16 and it dislike avs
can I have a link for your latest version please? thanks!

[edit]

found it, thanks!