AI & agents
Microsoft introduces models for real-time voice agents
- Published
- Source
- Microsoft AI Blog
Summary
Microsoft announced MAI-Transcribe-2-Streaming for live transcription in 60 languages, alongside MAI-Voice-2.1 and Flash speech-generation models. It produces partial transcripts before a speaker finishes and revises them as context arrives. Speed and accuracy claims come from evaluations cited in the announcement, not verified workplace productivity gains.
Why it matters for our work
Less waiting for speech input could give AI more time to retrieve material and prepare responses during customer support or research interviews. Because partial transcripts can change, consequential actions should depend on confirmed text and human intent.
Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.