Meta Unveils Muse Voice Transcribe with Real-Time Multilingual Speech Recognition

Meta has unveiled Muse Voice Transcribe, a new real-time speech-to-text AI model designed to make voice transcription faster and more useful, particularly for multilingual conversations. The model can transcribe speech as it happens and supports more than 70 languages, including five major Indian languages—Hindi, Tamil, Telugu, Malayalam and Kannada. Meta says Muse Voice Transcribe is designed to handle code-switching, allowing users to move between languages during the same conversation without requiring separate models or additional processing. The system can also distinguish speakers in recordings involving more than 20 voices and process audio lasting over an hour.

One of the model’s key features is its approach to balancing speed and accuracy. Rather than using a fixed listening duration before producing text, it determines the appropriate timing on a word-by-word basis, allowing simpler words to be transcribed quickly while giving more time to difficult-to-recognise speech. Meta said 25 of the more than 70 supported languages had been validated at launch, and claimed the model ranked first on the Artificial Analysis streaming speech-to-text leaderboard as of September 1, 2026. Muse Voice Transcribe is available through Meta’s Model API, priced at $3 per 1,000 audio minutes, or about $0.18 per hour according to Meta. The technology is already being used for dictation in Meta AI for Mac and Muse Code.