Meta Releases Muse Voice Transcribe: Real-Time Streaming Speech Model with Diarization
bowenc0221 · x · 2026-09-02
Meta has launched Muse Voice Transcribe, its first streaming audio perception model combining ASR, speaker diarization, and endpointing. It supports 70+ languages (25 at launch), hour-long sessions, 20+ speakers, and seamless code-switching. The model uses reinforcement learning to achieve 'adaptive delay,' waiting longer only for difficult words, establishing a new pareto frontier for speed/accuracy trade-offs. It is available via Meta Model API, Meta AI for Mac, and Muse Code.
More from Models
- Mocked benchmark shows DeepSeek's document chart grounding badly misaligned — burkov · 2026-10-18
- Dev complains Claude's habit of saying "not done" is his biggest pain point — jasonkneen · 2026-10-18
- Open-weight TYGER speech decision model makes 80ms turn-taking calls for full-duplex voice agents — solyarisoftware · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18