Announced on September 23, 2026, Nemotron 3 Diarization identifies who spoke and when, including in overlapping speech, and supports up to eight speakers.
On September 23, 2026, NVIDIA announced Nemotron 3 Diarization, a model designed to identify who spoke and at what time. The company says it handles overlapping speech, supports up to eight speakers, and has 100 million parameters.
NVIDIA says the model is publicly available. Check the official announcement and technical documentation to verify availability and details; the supplied material reports no comparative evaluation results. If using AI to study or apply the model, avoid submitting recordings or transcripts containing personal data without appropriate authorization and safeguards.