MicrosoftMAI Transcribe

MAI-Transcribe-2

MAI-Transcribe-2 is a public-preview transcription API with 60-language support.

Preview

What the release provides

It provides speaker labels, word timestamps, keyword biasing, language switching and clean or verbatim output. The current guide warns that diarization can fail on recordings around 15 minutes or longer, even when transcription without diarization succeeds.MAI-Transcribe Azure Speech documentation

How to use it well

Practical guidance. Choose clean or verbatim output according to the document's purpose. Review specialist names against the recording and keep corrections aligned with timestamps. Treat speaker labels as inferred until checked; recognition of words and identification of speakers are separate quality questions.

Specifications & access

MAI-Transcribe-2 specifications
DetailValue and conditions
StatusPreview
AccessAPI
Model identifiersMAI-Transcribe-2