For those interested, the paper we wrote to describe how the CELT codec works has just been accepted for publication. Here it is:

J.-M. Valin, T. B. Terriberry, C. Montgomery, G. Maxwell, A High-Quality Speech and Audio Codec With Less Than 10 ms delay, To appear in IEEE Transactions on Audio, Speech and Language Processing, 2009.

The paper is actually based on version 0.3.2, but the main principles are still the same.

I didn't understund a word of all these equations but it was nice to see your faces I'll sure try CELT once it gets a win32 binary & some playback support. I already hope it will have some tool similar to vcut personnaly.

Glad to see this, Jean-Marc. Congratulations. I'll be sure to have a deeper look later.

If the equations are a bit thick for you, you may appreciate the higher level presentation material on the presentations page.

There is a win32 binary on the site on the download page.

CELT is really intended for streaming rather than stored-file use… The file reader/writers exist primarily as testing and evaluation infrastructure. As a result file manipulation tools are not a high priority, especially while the file format is not finalized.

For those interested, here's one on the low-complexity mode of CELT:

J.-M. Valin, T. B. Terriberry, G. Maxwell, A Full-Bandwidth Audio Codec with Low Complexity and Very Low Delay, Accepted for EUSIPCO 2009.