Re: [AVT] Requesting feedback on new CELT codec

[Jean-Marc Valin <jean-marc.valin at usherbrooke.ca>]

Even, Roni a écrit :
> Jean-Marc,
> I looked at your web site. Note that the ITU-T approved G.719 which is a
> low complexity full band codec (40kHz) at rates from 32 - 128 kbit/s

It seems to me like G.719 is basically a further extension on the
G.722.1 and G.722.1C codecs. From what I could read, it has an
algorithmic delay of 40ms, whereas CELT's delay is tunable between 2ms
and 15ms. In terms of quality, I also did some listening tests and CELT
with 8.7ms delay came up ahead of G.722.1C despite having less than 1/4
of the delay. The comparison with AAC-LD closer, but CELT still had an
edge at 48 kbit/s, again with 1/4 of the delay. The rest results are at:
http://www.celt-codec.org/comparison/

Christian Hoene a écrit :
> I would suggest you to support multiple coding rates (from 16 to 200) and
> various coding delays ranging from 2ms (for interactive ensemble
> performances) to higher delays (e.g., 40ms if the is beneficial for the
> audio quality).

Actually, this is pretty much already supported. The bit-rate starts at
around 24 kbit/s (to have vaguely decent quality) and can currently go
up to ~128 kbit/s/channel. The delay can be tuned from 2ms to 15ms. It's
even possible to go below 2ms, but performance suffers a lot. I don't
think going above 15ms is really worth it for CELT.

> Then, various application scenarios and the use of CELT over congestion
> aware transport protocols like DCCP or TCP would make sense.

Note that the bit-rate can vary on-the-fly, but the delay (linked to the
frame size) has to be constant throughout the session. Is there any
other important feature to have?

Cheers,

	Jean-Marc

> Roni Even
> 
>> -----Original Message-----
>> From: avt-bounces at ietf.org [mailto:avt-bounces at ietf.org] On Behalf Of
>> Christian Hoene
>> Sent: Saturday, July 26, 2008 8:51 PM
>> To: 'Jean-Marc Valin'
>> Cc: avt at ietf.org
>> Subject: Re: [AVT] Requesting feedback on new CELT codec
>>
>> Hello Jean-Marc,
>>
>> Great work!
>>
>> I would suggest you to support multiple coding rates (from 16 to 200)
>> and
>> various coding delays ranging from 2ms (for interactive ensemble
>> performances) to higher delays (e.g., 40ms if the is beneficial for
> the
>> audio quality).
>>
>> Then, various application scenarios and the use of CELT over
> congestion
>> aware transport protocols like DCCP or TCP would make sense.
>>
>> With best regards,
>>
>>  Christian Hoene
>>
>>
>>
>>
>>> -----Original Message-----
>>> From: avt-bounces at ietf.org [mailto:avt-bounces at ietf.org] On Behalf
> Of
>>> Jean-Marc Valin
>>> Sent: Tuesday, July 22, 2008 5:55 AM
>>> To: IETF AVT WG
>>> Subject: [AVT] Requesting feedback on new CELT codec
>>>
>>> Hi everyone (bit of background, I'm the author of the Speex speech
>>> codec),
>>>
>>> I have been working on a new codec called CELT. It is not a
> successor
>>> to
>>> Speex, but a much higher quality codec (44.1/48 kHz) with very low
>>> delay
>>> (<10 ms) and rates in the 32-200 kbps range. The bit-stream isn't
>>> frozen
>>> yet and before I freeze it, I would be very interested in feedback.
>> Is
>>> there anything I could add (or remove) to make everyone's life
>> easier.
>>> Any application I'm not considering?
>>>
>>> The website is http://www.celt-codec.org/ and it includes some
>>> comparisons and audio samples. Like Speex, CELT is completely open.
>>>
>>> Cheers,
>>>
>>> 	Jean-Marc
>>> _______________________________________________
>>> Audio/Video Transport Working Group
>>> avt at ietf.org
>>> https://www.ietf.org/mailman/listinfo/avt
>> _______________________________________________
>> Audio/Video Transport Working Group
>> avt at ietf.org
>> https://www.ietf.org/mailman/listinfo/avt
> 
> 
_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www.ietf.org/mailman/listinfo/avt