The implications of this technology are concerning, as it could enable malicious actors to create convincing deepfake videos from a single image. Combined with facial animation and voice capture technology, it could be used to make a person appear to say or do anything. The developers are preparing the demo and code for public release, raising questions about the potential misuse of this technology.
Key takeaways:
- Researchers at Alibaba Group’s Institute for Intelligent Computing have developed a new generative video technique called Animate Anyone, which can create realistic videos from still images.
- The technique extracts details like facial features, patterns, and pose from a reference image, then maps those details onto slightly different poses to create a series of images.
- While the technology is not perfect and struggles with certain aspects like the eyes and hands, it is a significant improvement over previous models, producing fewer artifacts and retaining more important details.
- The developers are preparing the demo and code for public release, raising concerns about the potential misuse of the technology to create deepfakes.