OmniAvatar: The AI digital human technology breakthrough that brings still photos to life

OmniAvatar is an audio-driven digital human system jointly developed by Zhejiang University and Alibaba Group, capable of generating natural and smooth full-body motion video based on static photos, audio and text prompts. Compared with the traditional "talking avatar" technology, the system achieves breakthroughs in body movement coordination, high-precision audio and video synchronization, and text control. After testing, it is the only model that can synchronize facial and full-body animation, and is ahead in image quality, video smoothness and mouth synchronization. The project has been open-sourced and the paper is published in arXiv.