Back to AIBriefs
LaunchAI Models

Microsoft introduces MAI-Transcribe-1.5 speech-to-text model

The model achieves best-in-class Word Error Rate across 43 languages and can transcribe an hour of audio in under 15 seconds. It includes keyword biasing that improves WER by up to 30% on FLEURS and is integrated into Copilot, Teams, GitHub, and Dynamics 365 Contact Centre.

·
Jun 2, 9:54 AM