Hallo3 is a technology for portrait image animation that utilizes a pre-trained transformer-based video generation model. It is capable of generating highly dynamic and realistic videos, effectively addressing challenges such as non-frontal perspectives, dynamic object rendering, and immersive background generation. This technology has been jointly developed by researchers from Fudan University and Baidu, showcasing strong generalization capabilities and bringing new breakthroughs to the field of portrait animation.