Meta Launches Muse Voice Transcribe for Real-Time Multilingual Transcription

Meta launched Muse Voice Transcribe, a real-time audio perception model designed for multilingual streaming transcription. Developed by Meta Superintelligence Labs, the model uses adaptive delay to balance speed and accuracy, supporting native code-switching and speaker diarization for more than 20 distinct voices.

The technology is integrated into Muse Code and Meta AI for Mac. Mac users can access system-wide dictation by holding the Fn key to input text into any application. While the model was trained on over 70 languages, 25 are validated at launch.

Developers can access the model via the Meta Model API at a cost of $3 per 1,000 audio-minutes. Meta reports that the model currently holds the top position on the Artificial Analysis streaming speech-to-text leaderboard.