FantasyPortrait is a high-fidelity, multi-emotion portrait animation generation framework that uses an expressive enhancement learning strategy to capture subtle facial dynamics, suitable for both single-character and multi-character scenarios. The advantage of this technology lies in its unique masking cross-attention mechanism, which effectively prevents feature interference and enhances the quality and expressiveness of the animation. The background of this product stems from a reflection on the shortcomings of existing facial animation methods, particularly the challenges in handling multi-character interactions. In the future, the code and models will be provided in an open-source manner to encourage research and development.