Microsoft launched MAI-Transcribe-2-Streaming and two text-to-speech models, MAI-Voice-2.1 and MAI-Voice-2.1-Flash, on October 1.

Microsoft launched its first streaming transcription model, MAI-Transcribe-2-Streaming, on October 1, alongside text-to-speech models MAI-Voice-2.1 and Voice-2.1-Flash. The transcriber supports 60 languages and returns initial results in about 120 milliseconds; Voice 2.1 supports 23 languages. Flash can generate 45 seconds of audio with roughly 150ms end-to-end latency. The models target faster, more natural voice agents, intensifying competition in real-time AI speech. They are available in public preview; transcription starts at $0.54 per audio hour through year-end.