Meta launches real-time AI transcription model Muse Voice Transcribe

In a post on X, Meta CEO Mark Zuckerberg demonstrated the model's ability to automatically differentiate speakers and switch between languages. He said the transcription model picks up code-switching, transcribing sentences that use words from multiple languages. Zuckerberg returned to X after three years of not posting on the platform.

According to Zuckerberg, the model uses adaptive delay to decide when to listen, waiting longer on difficult words and committing faster on easy ones to increase accuracy. It was trained on more than 70 languages, with 25 validated at launch, and can handle hour-long sessions with more than 20 speakers.

The release comes less than a week after Google unveiled Gemini 3.5 Transcribe, an audio model with similar capabilities. Engadget reported that while Google plans to integrate its model into Android and eventually Chrome, it is unclear whether Meta will integrate Muse Voice Transcribe into its flagship services.

For now, users can experience the model through Meta's recently released Meta AI Mac app, where it powers dictation features across other applications. Muse Voice Transcribe is also available to developers via Muse Code and Meta's Model API, priced at $3 for 1,000 audio minutes. A demo version is available on Meta's research blog.

Muse Voice Transcribe is the latest release from Meta Superintelligence Lab, which has recently introduced a dedicated coding agent, an open-weight model, and the Meta AI Mac app.