[vorbis] RC3 ?

safemode safemode at speakeasy.net
Mon Oct 1 18:05:39 PDT 2001



On Monday 01 October 2001 19:36, Gian-Carlo Pascutto wrote:
> ----- Original Message -----
> From: "safemode" <safemode at speakeasy.net>
> To: <vorbis at xiph.org>; "Jonathan Walther" <krooger at debian.org>
> Sent: Tuesday, October 02, 2001 12:54 AM
> Subject: Re: [vorbis] RC3 ?
>
> > I'm more concerned with a psycho-acoustic engine that's vorbis biased. 
> > We have lame but lame's psycho-acoustic engine is hand tuned for mp3
>
> encoding.
>
> > I somehow doubt oggenc has an acoustic engine that rivals lame's in
> > performance
>
> Hmm, you could be surprised. But actually, I think they have little
> in common because of the fundamental differences between the two formats.
>
> > and i dont think lame uses any different routines for when it's
> > encoding to vorbis (which i haven't been able to do in rc1 or rc2 with
>
> lame -
>
> > cvs).
>
> LAME uses the Vorbis encoder to encode to Ogg. It does _not_ have it's
> own routines for that.
>
> > Does anyone know if there are any people working on this ?
>
> Hmm, LAME hasn't supported encoding to Vorbis for a long time.
>
> > Or does vorbis suffer the same frequency problems that mp3 does?
>
> Not at all...

Then this has all confirmed my assumptions.  This would be much more 
important to those who wish to go through the task of re-encoding their 
collections (that is not converting mp3 to ogg, which would be a waste of 
time unless it's some moral obligation) since having a good psycho-acoustic 
engine can mean the difference between a file completely sounding like crap 
and one that sounds almost exactly like the original, regardless of the codec 
being used.  
And i wasn't really saying that lame could encode the ogg files with it's own 
routines, I know it uses libogg and libvorbisenc etc for that.  I was more 
concerned with lame's psycho acoustic engine which runs through the audio 
before getting to the ogg encoder.   Thus it makes much more sense to use 
lame to encode oggs because it's acoustic engine is fairly mature.  The 
problem is that it's tuned for mp3s and not for ogg files.  So any real audio 
advantages that the vorbis format allows wouldn't be realized unless there 
are some vorbis tweaks that are enabled when the --vorbis option is set or 
whatever when using lame.  I'm just trying to find a way to not start from 
scratch in building a high quality vorbis biased psycho acoustic engine.   
Perhaps the answer is ripping Lame's psycho acousic engine out and making a 
completely ogg vorbis encoder.  

Regardless of any speed / compression improvements rc3 may have, all we have 
is a very skimmed down encoder that is lacking in an area that took the lame 
crew years to get to where it's at now.  I dont think it will take that long 
for an ogg encoder that builds off of lame though, and maybe then we'll start 
to hear some noticable differences enough to justify the longer encode times 
and more cpu intensive decodes.  oggenc was not intended from as far as i can 
tell as anything more than an example program.  Maybe i'm wrong but i think 
ogg needs a real encoder before people start encoding whole collections.    

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.



More information about the Vorbis mailing list