Outputs words when silence is input to it. How to stop this?
#54
by
polonuim210
- opened
Trying to integrate into a real time application but I feel that I can't because it incessantly outputs "Yeah" or "Okay" when there is the slightest audio signal sent to it.
Do I just need a better VAD? Why can't it output "..." or something for silence?
Thanks.
Easiest fix is just to add a VAD. Start with WebRTC VAD and try the higher “aggression” levels. If that still doesn’t feel good enough, switch to something model-based like Silero VAD — usually performs much better.