Microsoft has introduced three new AI models that can handle text, voice, and images. The move shows the company wants to build more of its own AI tools, even while it continues working closely with OpenAI.
These models come from Microsoft’s AI research team led by Mustafa Suleyman. The group formed last year and has been focused on building systems that feel more practical in real use.
MAI-Transcribe-1
This model focuses on turning speech into text. It supports 25 languages and is built for speed. Microsoft says it runs much faster than its earlier Azure tool. That makes it useful for tasks like meetings, interviews, and customer calls.
MAI-Voice-1
This one handles voice generation. It can create audio almost instantly, up to a full minute in a second. Users can also shape custom voices. That opens up options for content creators, apps, and voice assistants.
MAI-Image-2
This model works on visuals. It can generate images and even video from text prompts. It first showed up in Microsoft’s testing platform, MAI Playground, and is now being rolled out more widely.
All three models are now available through Microsoft Foundry. Some features are also accessible in the testing platform. Microsoft is also trying to stay competitive on pricing, offering lower rates than some rivals.
Even with these launches, Microsoft is not stepping away from OpenAI. The partnership is still a big part of its strategy. At the same time, the company is building its own models to avoid relying too much on one source.
The AI space is getting crowded. With these releases, Microsoft is making sure it stays in the race.
