Hi Guilles, You may have seen my blog article go by about my switching to Pulseaudio -- there are a few Voxin related bits I wanted to bring to your attention. I made a small tweak to the buffer-allocation logic in atcleci.cpp to get lower latency from Pulse, the change should not affect pure alsa systems. With the above, latency with pulse did come down from the default which was painfully high; char-echo when typing is now tolerable. However, observed latency is still higher than with pure alsa when speaking a buffer full of text, e.g. try the "cnn market data" URL template and observe the delay between when the auditory icon is played and when speech commences; it's significantly longer than on a pure alsa system. 2. Something in the voxin layer appears to die more frequently -- symptom -- speech goes silent -- but the tclsh process running the server is still live -- and you need to press C-e C-s to get speech back. Mostly FYI, but something you could perhaps keep an eye out for while testing the packages you build --=20 Thanks, --Raman(I Search, I Find, I Misplace, I Research) =E2=99=89 Id: kg:/m/0285kf1 =F0=9F=A6=AE -- Thanks, --Raman(I Search, I Find, I Misplace, I Research) ♉ Id: kg:/m/0285kf1 🦮
|May 1995 - Last Year|Current Year|
If you have questions about this archive or had problems using it, please contact us.