[CELT-dev] Description of how CELT packs into Ogg?

[Andrew Lentvorski]

On 4/17/11 11:27 AM, Timothy B. Terriberry wrote:
>> Is there are description somewhere of how CELT packs into the .oga file?
>>
>> Trying to reverse engineer it from the intertwined celtenc code is
>> suboptimal.
>
> No, and this is likely to change/be superseded by an Opus mapping in the
> (hopefully near) future. At a minimum an extension for multichannel
> (through multiple mono/stereo streams) is required. We simply haven't
> gotten this far down our TODO list yet.

Well, drat.  I just needed it in order to pull all the Ogg handling code 
into a different file so I can dump it all at once when I shift to the 
embedded system without risk of introducing bugs.

I'll just have to untangle the Ogg and CELT code.  Ergh.



Also, what's the deal with having to support SILK *and* CELT in Opus. 
That seems like it's going to block usage of CELT in resource 
constrained devices.

While it wasn't the only reason (and probably not even the main reason), 
one of the mistakes that Ogg made that was a barrier to adoption was 
that it requires quite a bit more RAM than an MP3 decoder.  Requiring 
the extra resource to support two codecs instead of just one in Opus 
seems like a mistake.

Is there something I'm missing?  If this is a rehash, just give me a 
pointer to the relevant discussion on the mailing list archives.

-a