Meta Muse Voice Transcribe is now available, marking the company’s first real-time audio perception model. Developed by Meta Superintelligence Labs, it delivers multilingual, streaming transcription across Meta AI for Mac, Muse Code, and third-party projects through the Meta Model API.
The model combines streaming automatic speech recognition with speaker diarization and endpointing. That means it can transcribe speech as it happens, distinguish between more than 20 different voices across recordings, and detect when a speaker has finished talking, all without a separate post-processing step.
How Muse Voice Transcribe works
Muse Voice Transcribe was trained across more than 70 languages, with 25 validated at launch. It handles audio longer than an hour and supports native code-switching within or between sentences. Language, keyword, and context biasing can further sharpen recognition accuracy.
Instead of relying on a single fixed balance between speed and accuracy, the model determines how long to listen before committing to each word. Meta calls this feature adaptive delay. The system moves quickly through simpler speech while drawing on more audio context for harder-to-parse words.
According to Meta, the model ranks first on the Artificial Analysis streaming speech-to-text leaderboard as of September 1.
System-wide dictation on Mac
On the Mac, Muse Voice Transcribe powers system-wide dictation. Users can hold the Fn key to dictate directly into any application. The technology is already running inside Meta AI for Mac and Muse Code.
Developers can access the model today through the Meta Model API, priced at EUR 3 per 1,000 audio-minutes. That works out to roughly EUR 0 per hour of transcription.
The launch expands Meta’s growing lineup of AI tools for Apple platforms, following the recent debut of Meta AI on Mac and the release of Muse Code. The company positions Muse Voice Transcribe as a foundation for faster, more flexible voice-driven workflows across desktop and developer applications.
Meta AI for Mac is available now with Muse Voice Transcribe built in, offering hold-to-dictate functionality via the Fn key.
Source
Image: 9to5mac.com
