[vorbis-dev] thoughts about MP3 -> OGG convertor

Daniel B. Miller dan at on2.com
Tue Dec 18 12:18:42 PST 2001



for the record, what Linus is proposing is called 'transcoding' and is
done all the time in video (well some of the time anyway).  Usually it
applies to changing the bitrate on a given stream without changing the
format (say MPEG2 VBR @ ~8mbit production stream to MPEG2 CBR @3mbit
broadcast).  We are looking into ways to transcode from MPEG to VP4, a
similar issue to transcoding from MP3 to Vorbis.  THe trick is in seeing
how the two algorithms match up.  It's possible that by knowing alot about
the MP3 bitstream, you can reuse certain codec-internal data, and make
better choices on encoding the vorbis stream.

I'm not presently familiar with either audio codec, so I can't say if this
is feasible or not, only that it's not inconcievable.

 ___  Dan Miller
(++,) Founder, CTO, On2.com

On Tue, 18 Dec 2001, Dirk Knop wrote:

> Ahoy List,
>
> finally, after a long time, a new post from me ;)
>
> Linus Walleij wrote:
>
> >On Tue, 18 Dec 2001, John Zitterkopf wrote:
> >
> >>Why is this a bad IDEA? I take it it's because of the "lossy" nature of
> >>both types.
> >>
> >Its a bad idea because mp3-compression and vorbis compression affect the
> >audio in different ways: artifacts from the mp3compression will be mixed
> >with artifacts from the vorbis compression.
> >
> [...]
>
> This is absolutely correct. No additions to make :)
>
> >
> >Cliffhanger: I PRESUME it could be possible to convert the kind of
> >mathematical information stored in MP3 files to the kind of
> >mathematical information stored inside vorbis files, without
> >decompressing the audio and recompressing it inbetween, thus creating a
> >minimal loss transfer method. If you love maths, transforms and the
> >like, by all means have a look at it.
> >
> I have to disagree.
> The decoder is "fixed" in the way it works. All the psychoacoustic
> castrations are done by the encoder. The decoding process is just a iDCT
> of the data created by the encoder, so there is no difference between
> the mathematical data and the decompressed output - it's absolutely the
> same. The "mathematical data" is already lossy - that's the point in
> doing ogg / mp3 /ac3 / anything-compression.
> So the data gets "flawed" during encoding, no need to step in at a
> certain point of decompression.
> Maybe i got you wrong here, but i think that's what you wanted to say.
>
> (Erm, don't get me wrong, this is no offense or flame!)
>
> Regards,
> Dirk
>
>
>
> --- >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-dev-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.
>

--- >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-dev-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-dev mailing list