VAD (Voice Activity Detection) on nemotron-asr-streaming

AI & Data Science NVIDIA Nemotron
,
1

I’m deploying the nemotron-asr-streaming NIM container, and I can’t get VAD (Voice Activity Detection) to work.

If I commit at a regular interval, I get a text at a regular interval. But if I enable VAD (using the NeMo framework’s configuration), it never sends data except when stopping recording.

Is it supposed to work?

Thanks,

Luc