[vorbis-dev] BCB build successful?

[Gregory Maxwell]

On Wed, Jan 31, 2001 at 07:21:59PM -0800, David K. Gasaway wrote:
[snip]
> > So if the binaries you had have different codebooks, it would output
> > different files which both should reduce (rehuff) down to a 3rd
> > identical file ... though I suspect its different sources at this
> > point?
> 
> Well, I don't follow that conclusion, but I don't know too much about 
> audio coding (specifically).

Nope. The codebooks in vorbis do more then optimal huffman encoding. The
codebook space is smaller then the input space, i.e. they are lossy.

Vorbis codebooks are of two types: fully trained or lattice books.

Lattice books be represented as a simple regular lattice of values, part of
creating codebooks is picking lattice with 'good' spacings.

Fully trained books actually transmit a complete set of possible values. A
training utility needs to optimally cluster these values.

If the only difference between differnt books in an encoder is the huffman
part, then yes, rehuff will 'fix them'. However, it's unlikely that you
would find two differnt encoders with the same codebooks params but differnt
huffs, becuase it's pretty much just Monty making the codebooks. The only
way it would occure is if you use the same lattice spacings but huffed the
book with a differnt training set, for fully trained books this is unlikely
because you the book values depend on the training set.

--- >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.