Meta Superintelligence Labs released Muse Voice Transcribe, our first real-time audio perception model.
Muse Voice Transcribe delivers streaming transcription, speaker separation across 20+ voices in hour-plus recordings, and native code-switching, from a single model with no post-processing step. Trained across 70+ languages with 25 validated at launch. It ranks first on the Artificial Analysis streaming speech-to-text leaderboard as of September 1, 2026. Rather than locking in one speed versus accuracy setting, it decides per word how long to listen before committing, which is what lets it stay fast without giving up accuracy on hard words.
Available today via the Meta Model API, and already running dictation in Meta AI for Mac and Muse Code. API Pricing: $3.00 per 1,000 audio-minutes, equivalent to $0.18 per hour.
More Information in the research blog: Introducing Muse Voice Transcribe and social post on X.