On Tue, Feb 12, 2008 at 06:42:30PM +0000, Måns Rullgård wrote:
Rich Felker <dalias@aerifal.cx> writes:
On Tue, Feb 12, 2008 at 03:24:34PM +0100, Alban Bedel wrote:
On Mon, 11 Feb 2008 22:41:51 -0500 Rich Felker <dalias@aerifal.cx> wrote:
I don't think anyone really cares that much about efficiency when storing shit codecs in NUT. Obviously any good codec will use large frame sizes or compression will not be good.
Ever heard of VoIP? In such applications it is essential to have as little latency as possible. Do the maths and you will see that you need a codec frame size between 10 and 30ms to get a decent experience. At 8KHz that's just a couple 100 samples.
At 8khz you might as well send the contents of /dev/urandom instead of a signal...
Please stop this nonsense. Everybody who has ever used a telephone can certify that 8kHz is perfectly adequate for plain speech.
Perhaps someone could let Rich make a test call on their phone so he can see for himself.
Sorry I should have appended a smiley or something. But seriously, I've seen much better quality at higher samplerates using more general-purpose codecs (vorbis) with similar or marginally higher bitrate. I'm very skeptical that the whole "speech compression" field is just an industry patent game rather than something really useful. Besides, there are plenty of times I'm talking on the phone and want someone to hear a non-vocal sound such as music that's playing or a bad sound coming from a broken machine, etc.. Rich