Microsoft has introduced a new AI model that can generate lifelike talking faces of virtual characters, using a single photo and an audio clip.
According to the company’s statement, the new model, VASA-1, converts a photo and an audio file into a realistic talking face video, with lip movements synchronized with the audio, and realistic facial expressions and natural head movements.
It said that this new AI model paves the way for real-time engagements with lifelike avatars that mirror human conversational behavior. But so far, the videos generated by this new AI model still lack the authenticity of real videos.
Microsoft added that its research focuses on generating visual affective skills for virtual AI avatars, aiming for “positive application.” It stressed that its new tool is not intended to create content that is used for misleading or deceiving.
The company warned that the new AI tool could still be misused for impersonating humans, expressing its firm opposition to creating any misleading or harmful contents of real persons.
Still, the new tool has considerable positive potential, including enhancing educational equity, improving accessibility for individuals with communication challenges, offering therapeutic support for those in need, among others.
These benefits emphasize the significance of Microsoft’s research and other related experiments, dedicated to developing AI responsibly with the aim of promoting human well-being.
The tech giant said it doesn’t have plans to release an online demo for the new tool or any further details related to its implementation until it makes sure the technology will be used responsibly and in conformity with proper regulations.
