Re: [rtcweb] Confirmation of consensus on audio codecs

[Jean-Marc Valin <[e-mail]>]

> (speaking for Nokia)
> 
> Nokia thinks that there are complexity issues with the Opus codec.
> There are also possible IPR risks associated with the "royalty-free"
> nature of Opus. We do not recommend Opus to be taken as a mandatory
> codec for RTCWeb at this point. From high quality low bit rate and
> mobile applications point of view the 3GPP AMR-WB codec (also known
> as ITU-T G.722.2) is the most preferable for us. For interoperability
> with implementations restricted to unencumbered codecs, we prefer
> G.711 and G.722.

Sorry, but I think G.722 (or AMR-WB) is just not enough. Speech-only, AM
radio quality is not what we need here. Skype has been shipping a
super-wideband codec for about 3 years now, so going back to wideband as
the highest guaranteed interoperability would make webrtc irrelevant.
Let's look at the options here. On one side we have Opus, that can
easily handle all possible rtcweb applications because it does
narrowband, wideband, super-wideband and fullband, can handle both
speech and music, and scales to high-quality stereo (for music
applications). On the other side, if you want the same functionality,
you would need at least the following codecs: AMR-NB, AMR-WB, G.722.1C,
AAC-LD, HE-AAC, and AAC-LC. And on top of the insane IPR cost of
shipping these 6 codecs, you end up with an inflexible solution (hard to
switch between 6 codecs in real-time based on the network bandwidth),
with slightly lower quality, and that in practice nobody would fully
implement.

	Jean-Marc