Streaming Nemotron 3 Diarization

Excerpt

I’ve been playing with Nemotron 3 Diarization , and it fills a gap I’ve had with local voice agents: keeping track of who is speaking. It’s a diarization model, so it gives you speaker labels rather than transcriptions or people’s names. It can stream its output and track up to eight speakers. I’ve been trying it with one-second streaming chunks, and the quality has been really good in my tests. I plugged it into my speech-to-speech setup on a DGX Spark and connected it to a Reachy Mini. The fun