Microsoft's VASA-1 AI model can create realistic talking faces from a single image and accompanying audio.

The AI captures intricate facial expressions and natural head movements, and works with various types of input, including artistic images and non-English languages.

VASA-1 can produce high-resolution video at up to 45 fps offline and 40 fps during online streaming with minimal latency.

Concerns have arisen over the technology's potential for misuse in scams, prompted by a video of the Mona Lisa singing and rapping using the AI.

Microsoft is focusing on responsible development, delaying the release until regulatory frameworks and guidelines for safe use are established.