NVIDIA's Nemotron 3 Diarization Model Now on Hugging Face

Key Info

NVIDIA's Nemotron 3 Diarization model, which tracks who spoke when even with overlapping voices, is now available on Hugging Face. It supports up to eight speakers and has 100M parameters.

Highlights

• Handles overlapping speech to produce accurate transcripts.
• Supports up to eight speakers in a conversation.
• 100M parameter model available on Hugging Face with day-0 transformers support.