[Ffmpeg-devel] When is planned to add ogg Theora output in ffmpeg?

Rich Felker dalias
Fri Apr 7 22:49:38 CEST 2006

On Fri, Apr 07, 2006 at 08:46:30PM +0100, M?ns Rullg?rd wrote:
> Rich Felker <dalias at aerifal.cx> writes:
> > On Fri, Apr 07, 2006 at 03:54:56PM +0100, M?ns Rullg?rd wrote:
> >> MPEG4 is generally considered a failed experiment.
> >
> > By whom?
> Codec developers and users (not counting r1pp3rz), mainly.  The
> increased complexity over MPEG2 is often too large compared to the
> quality improvement.  For instance, including a loop filter would have
> added little to the complexity, and could have significantly improved
> the quality.

And ruined performance.

There's a reason the skip* options exist in the lavc h264 decoder:
some of us don't want to waste huge cpu time for marginal quality
gain. The fact that skipping the loop filter (and thus decoding
incorrectly) doesn't produce significant visual error is evidence that
it's pointless, at least for B frames and low key intervals.

Anyway common wisdom seems to be that h264 is only 10-20% better
compression than mpeg4, and thats with all the arith. coding and other
stuff too.

> >> Don't get me wrong, matroska is just as bad, only in different ways.  It
> >> aims to do *everything* with no regard to whether those things make sense
> >> or not.  More often than not, they don't.
> >
> > At least it works.. aside from the idiotic lacing which makes A/V sync
> > difficult.
> I've never noticed that problem with matroska.  My main complaint
> about matroska is that it is overly complex.  Implementing a fully
> featured demuxer is a major piece of work.

Lacing means that timestamps for the majority of frames are
suppressed. This is extremely problematic for vfr (which includes
vorbis) audio. See the recent mplayer bug about this.


More information about the ffmpeg-devel mailing list