Microsoft releases new models for real-time voice agents
Microsoft AI has introduced MAI-Transcribe-2-Streaming, a model designed to transcribe speech in real time. Microsoft says it ranks first for accuracy on Artificial Analysis, supports 60 languages and produces its first partial results in just over 100 milliseconds. This latency is intended to let voice agents begin responding while a speaker is still talking. The introductory price through the end of the year is $0.54 for an hour of audio.
The company also launched MAI-Voice-2.1 and MAI-Voice-2.1-Flash for text-to-speech. MAI-Voice-2.1 can use the same voice across 23 languages while providing a native accent in each language. Microsoft reports 150 milliseconds of latency for the Flash version, priced at $15 per million characters instead of $22. Both models can clone a voice from a few seconds of reference audio and include safeguards intended to limit misuse.
The models are available through Microsoft Foundry and MAI Playground, among other platforms, while both voice models are also offered on OpenRouter. In a test involving 4,000 participants, about half judged the generated voices to belong to a real person.
Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.