According to Microsoft, not only does VASA-1 create "subtly synchronized" lip movements with sound, but VASA-1 also shows the character's charisma through face and head movements, thereby increasing realism.

The American software company says this is a "face kinematics system" trained through a motion model that is "comprehensive better than previous methods". this AI supports portraits with the head and upper body, with happy, angry, surprised or serious expression.

In the video introduced by Microsoft on YouTube, the company's AI can make photo created by Dall-E 3 and StyleGan2, talk naturally. Users can also adjust facial parameters, expressions, voice and many other details. image specialist site PetaPixel rates VASA-1 as having better performance than similar AIs, making it difficult to recognize if viewed for the first time.

Source: vnexxpress

Translated by Nguyen Thi Luc.