Microsoft introduces MAI-Transcribe-2 with $0.10-per-hour launch price

Microsoft introduced MAI-Transcribe-2, a speech-recognition model with speaker diarization, word-level timestamps, configurable transcription styles, keyword biasing and automatic language identification. Microsoft said the model supports 60 languages and will launch at $0.10 per hour through the end of the year. The company also claimed faster processing and stronger accuracy than several named competing transcription models, citing its own benchmark and third-party evaluation references.