On Thursday, Microsoft's AI division debuted its inaugural suite of self-developed AI models: the MAI-Voice-1 speech model and the MAI-1-preview general model. According to the tech giant, the MAI-Voice-1 speech model boasts the capability to produce a one-minute audio clip within a second, leveraging just a single GPU. Meanwhile, the MAI-1-preview model signals the trajectory for future functional enhancements within Copilot. Presently, MAI-Voice-1 is integrated into diverse applications, including AI anchors delivering news on "Copilot Daily" and generating dialogues for podcasts. Users can engage with this model on the Copilot Labs platform, where they can personalize voice tone and style to their preference. Tailored for users with specific requirements, the MAI-1-preview model is capable of following instructions and has commenced public testing on the AI benchmarking platform, LMArena.