Microsoft launches MAI-Transcribe-2-Streaming, calls it most accurate real-time transcription
Microsoft just launched MAI-Transcribe-2-Streaming, which Mustafa Suleyman, Microsoft AI CEO, calls "the most accurate real time transcription model in the world."
It handles 60 languages, figures out which one you're speaking automatically, and is built for developers working on autonomous conversational voice agents.
Mustafa Suleyman says it's not only faster (by 55%) but also way cheaper (60% less) than rivals like ElevenLabs.
Microsoft MAI-Voice-2.1 supports 23 languages
Alongside that, Microsoft introduced MAI-Voice-2.1, its top-tier engine that can read out text in 23 languages while preserving its unique vocal identity while adapting pronunciation if you switch languages.
The Flash version is built for speed, cranking out 45 seconds of audio with barely any lag (just 150 milliseconds), making it super handy for businesses needing instant AI-powered responses.