Nemotron 3 Diarization in action: streaming 8-speaker tracking powers a memory-equipped robot
futterneid · reddit · 2026-09-23
A hands-on look at NVIDIA's open-source Nemotron 3 Diarization, which fills a key gap for local voice agents — knowing who is speaking:
- Outputs speaker labels (not transcripts), streams in real time, and tracks up to 8 speakers.
- The author tested with 1-second streaming chunks and reports excellent quality.
- Demo: plugged into a speech-to-speech setup on a DGX Spark driving a Reachy Mini robot — when someone new speaks, the robot asks their name and remembers it for the conversation.
- Ships with day-zero Transformers integration and a commercial-friendly license is coming.
Related event: NVIDIA Open-Sources Nemotron 3 Diarization, Tops Benchmark(16 posts)→
More from Embodied
- Cube-Solving Robot Demo Is Its Best Case: ~90 Moves per Solve on Average — haydendevs · 2026-11-11
- ETH researcher vibe-codes SuperMimic humanoid reproduction in 1h, LAFAN passes in 1.5 GPU-hours — ChongZzZhang · 2026-11-11
- Open-source companion robot Lamp ships at $699 with custom personality, skills and an app economy — dee_hw · 2026-11-11
- Must-read papers of the week: HALO humanoid motion, GR00T 2, VLA policies, agent tooling — TheTuringPost · 2026-11-11
- Microsoft's DLSS rival Auto SR launches on Xbox Ally X with handheld 1080p and 120fps support — tomwarren · 2026-11-11
- NVIDIA's PhysicalAI-Human-1M-100K robot dataset trends on Hugging Face — nvidia · 2026-11-11