[CELT-dev] Just a question on how things are going

[Jean-Marc Valin]

John Ridges <jridges at masque.com> a écrit :
> In my case, I'm afraid the IETF standard is much too far away for me  
> to wait. So I guess what you're saying is that waiting for the  
> frozen bitstream for someone (like me) who's only using CELT isn't  
> that important because CELT probably won't be improved upon as a  
> standalone codec after the bitstream is frozen anyway. It seems kind  
> of a shame but I can understand your reasons for it.

I don't see what you don't like in this. I'm currently working getting  
a CELT 1.0 codec (with frozen bit-stream) as soon as possible. The  
only difference is that this 1.0 version will actually be part of  
something bigger/better. Once the bit-stream is frozen, then the only  
thing you can improve is the encoder (and that will likely happen),  
whether or not it's part of a bigger codec. In the end, I don't see  
what disadvantage you could find in this.

> I've got one quick question if you don't mind. Do you know offhand  
> at what bitrate the pitch predictor ceased to be useful? In my  
> application my minimum bitrate is 35 Kbits/sec. If the pitch  
> predictor helped at that bitrate, I might want to settle on a  
> version of CELT that still had it. Thanks,

The pitch predictor was moderately useful up to around 50-55 kb/s. So  
it definitely helped at 35 kb/s. That being said, I'm not sure CELT  
should really be used below about 40 kb/s in the first place. I find  
that it tends to have too much distortion below that point. OTOH, the  
hybrid codec which we've been working on can produce nearly perfect 48  
kHz speech at 32 kb/s.

Cheers,

     Jean-Marc