Meta on Tuesday introduced Muse Voice Transcribe, its first real-time audio model, which can transcribe dictation and conversations involving more than 20 speakers across multiple languages.
The model is available now through Meta’s Model API and within Muse Code, and Meta said people can also use it in the recently released Meta AI Mac app.
Chief Executive Mark Zuckerberg demonstrated the system in a post on X, showing it distinguishing between speakers in real time and switching between languages within the same conversation.
Zuckerberg said the model can also capture mid-sentence code-switching, transcribing phrases that combine words from multiple languages.
“The model decides when to listen. It waits a little longer on hard words and commits faster on easy ones, using adaptive delay to predict each token and increase accuracy,” Zuckerberg wrote on X.
Muse Voice Transcribe is MSL’s first real-time audio perception model — rolling out today. SOTA in streaming speech-to-text, it handles speaker diarization, and endpointing natively in a single model. pic.twitter.com/LViMDSkbim
— Mark Zuckerberg (@finkd) September 1, 2026
He said the model was trained across more than 70 languages, with 25 validated at launch, and can handle “messy, real audio” as well as hour-long sessions with more than 20 speakers.
Meta described Muse Voice Transcribe as Meta Superintelligence Lab’s first real-time audio perception model, with speaker diarization and endpointing built into a single model.
The launch came less than a week after Google introduced Gemini 3.5 Transcribe, an audio model with similar capabilities.
Meta has not outlined plans to integrate Muse Voice Transcribe into its main consumer services, but the company said the Meta AI Mac app can extend the model’s dictation features into other apps.
Alexandr Wang said in a post on X that the model was already powering dictation in the Meta desktop app and Muse Code.
3/ already powering dictation in the meta desktop app and muse code. live now via meta model apihttps://t.co/MFosERCV0E pic.twitter.com/UESTQhKJrH
— Alexandr Wang (@alexandr_wang) September 1, 2026
Meta priced Muse Voice Transcribe at $3 per 1,000 audio minutes and made a demo available on its research blog.
Muse Voice Transcribe is the latest product release from Meta Superintelligence Lab, which in recent weeks has also introduced a dedicated coding agent, an open-weight model and the Meta AI Mac app.
Featured image credit
