[MPlayer-dev-eng] [PATCH] Latest CVS Theora update and about Tremor

[Martin Drab]

On Wed, 20 Aug -1, Arpi wrote:

> Hi,
> 
> > > Once we have working vp3, we could add theora header parsing, so
> > > the vp3 decoder could be initialized from tables tored in file,
> > > instead of builtin static data. Afaik no other difference between
> > > theora and vp3.
> > > 
> > > A'rpi / Astral & ESP-team
> > 
> > Really? Form what they're saying on their web page one would think 
> > there's more differences than just header parsing. What would take them so
> 
> not header parsing. vp3 used static (builtin) tables, while theora allows
> the encoder to build the optimal tables, and the store the tables in the
> header. it may be a big improvement in vq-based codecs like vp3.

Ah! Gonna have to filnally find some time to read the spec.
 
> > long to create it then?
> 
> probably it took so long to take the very messy win32/mac targetted vp3
> sourcecode and transfer it to even more messy libogg-dependent shit.
> </flame>

:-)

> also, they probably work a lot on improving the encoder, ie. adding the
> ability to build optimal tables.
> (this is the key thing, see mp3 for example, there are several mp3 encoders,
> from ffmpeg, though xing to l3enc or lame, with huge quality difference,
> they all enoe to teh same format (same decoder can be used) but i bet there
> is big work behind the mp3lame code)

Yea, true. I just wasn't able to find the time to read the Theora spec. So 
I didn't know what was the improvement against VP3. Thanx.

Martin