Meta Superintelligence Labs on Tuesday introduced Muse Voice Transcribe as its first real-time audio perception model. Unlike a basic transcription system that processes a recording after the fact, Muse produces text continuously while identifying speakers and detecting when speech begins and ends.
The model supports audio with more than 20 speakers and recognizes speakers switching languages during a conversation. Meta says Muse was trained across more than 70 languages, with 25 extensively validated for the initial release.
Those validated languages include Hindi, Tamil, Telugu, Kannada and Malayalam, making the model potentially useful in…
Read the full article at TECHREPUBLIC.COM
