Skype publishes SILK audio codec source code | Hacker News Reader

Skype publishes SILK audio codec source code

h-online.com

18 pointspmjordan10 comments

Original post from the Skype blog: http://share.skype.com/sites/en/2010/03/advances_in_audio.ht... Link to SILK on Skype Dev Zone: https://developer.skype.com/silk

It's worth noting in relation to the HTML5/Theora/VP8/H.264 shenanigans that Skype currently use VP7 as their video codec.

I like this summary of why royalty free audio codecs makes business sense for them:

"1. Technological progress saturates.

2. Patents expire.

Therefore, the performance advantage of royalty-bearing standards diminishes with time, and high-quality, royalty-free standards are unavoidable. I'm convinced that today we have reached this point of commoditization for audio and speech coding technology."

from http://www.ietf.org/mail-archive/web/codec/current/msg00897....

What I don't get is them on one side pushing this at the IETF to "convert" it into an RFC and at the other side requiring that license agreement to be signed.

Doesn't the IETF disallow something like that? I always thought that specifications defined in RFCs were free to implement by anybody.

I wonder what happened to the hinted-at open sourcing of the linux GUI a couple of months back. I really hope this is not it and we still get that.

When I make a Skype call, I always seem to get G729 and not SILK.

Is SILK only use for calls between Skypes (and not Skypeout)?

I'm guessing here, but as a Skypeout call at one time has to go into the classical telephone network, probably using some existing VoIP infrastructure in between, the higher sampling rate of SILK would not be of any advantage and the audio would have to be recompressed (probably to G729 anyways if the call goes over some "classical" VoIP network.

Considering that stacking up lossy compression algorithms is never a good thing for the quality, they might opt for encoding to G729 right at your end of the call.

That makes sense. I just called the Skype Test Call and it used SILK_V3, so that would be further confirmation.

Do you think that Skype might be losing out by providing this under a non-commercial use license? It'll now surely get reimplemented by a free software group (ffmpeg comes to mind...) and that'll probably become the defacto implementation by reason of freeness of license. This means that Skype may lose on some commercial support opportunities since the implementation most use will come from another vendor. Any thoughts on this?

I get the impression they are in a transitional period and they will make the code and the patents freely available. At the moment they seem to be sending mixed messages and indeed even non-commercial use isn't currently allowed beyond testing without an IP licence agreement according to the download page.

On the other hand perhaps they intend for this codec to be entirely replaced by the output of the IETF codec work they are participating in (which may be roughly the same codec, or a descendant/hybrid of it and other codecs) in which case they don't actually want anyone to use this codec as it will then be a competitor to their first choice.

I thought it was a simple algorithm:

voice + crackles * importance of call

Reply on news.ycombinator.com