> Back to topic: > It would be great when some pulseaudio developers can look into the > pulseaudio driver of speech-dispatcher. > I think the main problem can be found there. Done! The new pulse.c works much better. Pulseaudio still introduces a barely perceptible delay, but maybe we can get rid of that with some tuning. Also, I'd posted a question about why short sounds were not being played immediately when queued through the pulse-simple API. I was able to get around this problem by setting the the tlength buffer attribute to 100, though I haven't tuned that value yet. Bill
