Re: Tcrequant breakthrough!

Otto J. Makela wrote:
> I've now figured out one situation where tcrequant (and also M2VRequantiser)
> makes a mess of it: when given a m2v extracted from a full mpeg. This happens
> to be the format I usually store videos before mastering them to dvd's.
> 
> When you take an audio and video file, and combine them into an ersatz "DVD"
> mpeg stream with mplex, tcdemux no longer extracts it to the same m2v file
> from the mpeg. Neither is tcrequant able to do its magic on it.
> 
> mplex -f 9 -o am-dvd.mpg am.m2v am.mp2
> tcdemux -i am-dvd.mpg -x mpeg2 | tcrequant -f 1.3 -o am-extracted-requant.m2v
> 
> After this, am-extracted-requant.m2v is garbled garbage.
> 
> Maybe this is actually some kind of a fault in tcdemux, as even mplex no
> longer recognizes the extracted stream as a m2v video stream? By the way, why
> doesn't tcdemux have an "-o" option like most of the other tc* utilities?
> 
> As before, I've prepared a sample (8MiB) for you at:
> 	http://www.otto.net/~otto/requant-sample2.tar.gz
> 
> I would appreciate someone (who understands the intricacies of video files
> better than me) checking this out and figuring out if my conclusions are
> correct. Thank you!

I haven't investigated this idea, but might the problem here be a result 
of using tcrequant on an input stream rather than a file?  The new 
Metakine code needs to know the stream length and perhaps the old 
version actually needed it too. Does the pipe provide that?  Can you run 
tcdemux with > outfile.m2v?

What still worries me is the difference that I saw between the CentOS5 
and FC10 results.

Cheers

John P