Microsoft's new VASA-1 AI framework generates super-realistic talking heads that can even sing songs - SiliconANGLE
SiliconANGLE
|
Summary
Microsoft has disclosed a new AI framework named VASA-1, designed to animate highly realistic talking heads from a single photograph and an audio file. This technology is capable of synthesizing facial expressions, head movements, and synchronizing lip movements with the provided audio, allowing even for singing. While similar technologies have been released by Nvidia Corp. and Runway AI Inc., Microsoft's VASA-1 showcases a higher degree of realism with fewer artefacts around the mouth region. Currently a research initiative, VASA-1 has not been made publicly available, citing potential risks of misuse such as creating convincing deepfakes. Microsoft aims to explore positive applications, emphasizing virtual character animation for video games and other media, while ensuring responsible use of the technology.
