ogg dev - Dec 2005 - Xiph QT Theora Codec

Taking a look ahead at your QT (OSX version) work, you have Theora  
Encoding and Decoding on the list.  This could be very beneficial as  
many times the only way to get a constant frame-rate out of QT is to  
export (and for DV video recorded in QT, this can be a large issue).   
Another major issue is that the video coming from DV (at least the  
sources I have used) is almost always interlaced, and since there is  
no deinterlace in the filter options (not to mention lack of a filter  
chain), but the codecs seem to be the best way to put a deinterlace  
into it prior to actual encoding of the frames.
Whether it is better to do this in an intermediary file or not is up  
to implementation.

I see by the current 0.1.3 sources that FLAC is probably next on the  
list, which makes Theora in the distance, but I thought I'd ask now.   
I have been generating Theora files using one of the CLI tools, a  
script file that calls mplayer for audio and dumps to a FIFO and  
calls it for YUV and dumps to FIFO then runs the Alpha 5 encoder on  
the FIFO files to generate a final OGG.  Due to the DV not going in  
at a constant 29.97, the audio desyncs, so I have to export the  
initial DV to another DV with a normalized FPS of 29.97.  When  
dealing with 10GB/hr of video, this process becomes large and  
unwieldy quickly.  I take that file and use the 2ogg.sh to encode it  
perfectly, but QT Theora would take it straight from the original  
source with less loss from intermediary transcodes.

On Mon, Dec 19, 2005 at 02:09:56PM -0500, Anders Kirchenbauer wrote:
> Taking a look ahead at your QT (OSX version) work, you have Theora  
> Encoding and Decoding on the list.  This could be very beneficial as  
> many times the only way to get a constant frame-rate out of QT is to  
> export (and for DV video recorded in QT, this can be a large issue).
Yes, the ability to export Ogg Theora (or even vorbis+vp3 in .mov) from 
quicktime applications has for a long time been a major limitation for 
the format.
> Another major issue is that the video coming from DV (at least the  
> sources I have used) is almost always interlaced, and since there is  
> no deinterlace in the filter options (not to mention lack of a filter  
> chain), but the codecs seem to be the best way to put a deinterlace  
> into it prior to actual encoding of the frames.
Hmm. So we need to include a deinterlacing filter in the theora 
component? Does CoreVideo help at all with the filter chain issue?
>                                          Due to the DV not going in  
> at a constant 29.97, the audio desyncs, so I have to export the  
> initial DV to another DV with a normalized FPS of 29.97.
How does that work? I thought DV was always fixed framerate. I do 
understand that quicktime is natively fixed framerate (in fact we 
occasionally get complaints in the other direction, because theora
can't perfectly reproduce the variable framerate video from cheap
webcams and capture cards with sloppy sync.)

Otherwise, yes, not being able to do direct encoding from your source
is always painful. Have you tried ffmpeg2theora. I don't think it
talks to quicktime, but it should be able to handle dv input.

 -r