Developed jointly by Tencent and Shanghai Jiao Tong University, MimicMotion is a high-quality human motion video generation model. This model achieves controllability in the video generation process through confidence-aware pose guidance, improving temporal smoothness and reducing image distortion. Employing an advanced image-to-video diffusion model combined with a spatiotemporal U-Net and PoseNet, it can generate videos of arbitrary length with high quality based on pose sequence conditions. MimicMotion significantly outperforms prior methods in several aspects, including hand generation quality and accurate adherence to reference poses.