MICROSOFT TESTS NEW MAI REALTIME VOICE MODEL

Summary from the TLDR IT newsletter

Microsoft appears to be testing MAI Realtime, its first native real-time voice model, through a hidden early-access listing in MAI Playground. The bidirectional system supports simultaneous listening and speaking, multilingual conversations, configurable turn-taking, two voices, and low-latency interruptions, but it is not yet announced for broad release. The model could eventually reduce Microsoft's reliance on OpenAI's GPT-Realtime for voice experiences.