Meta launches Muse Voice Transcribe for multilingual live speech transcription
Meta just launched Muse Voice Transcribe, a new tool that can turn live speech into text instantly, even if you're switching between more than 70 languages like Hindi or Tamil.
Built by Meta Superintelligence Labs, it doesn't just transcribe; it can separate out different speakers and handle long recordings (more than an hour) without extra work.
Plus, it recognizes more than 20 voices in real time.
Muse Voice Transcribe uses 80ms chunks
Muse Voice Transcribe slices audio into tiny pieces (just 80 milliseconds each) and uses reinforcement learning to boost accuracy.
It's smart enough to keep up if you switch languages mid-conversation, and uses context clues to get words right.
Already live in Meta AI for Mac and Muse Code for dictation, you can access the tech via the Meta Model API at $3 per 1,000 minutes, making real-time transcription way easier across apps.