X-Dance is a test dataset specifically for image-to-video animation generation, released in 2025 by Nanjing University in collaboration with Tencent and the Shanghai Artificial Intelligence Laboratory. The related research paper is titled "SteadyDancer: Harmonized and Coherent Human Image Animation with First-Frame PreservationThe study aims to evaluate the robustness and generalization ability of models in real-world scenarios when dealing with challenges such as identity preservation, temporal coherence, and spatiotemporal misalignment. This dataset contains 12 driving videos, including 8 high-dynamic dance movements and 4 low-amplitude everyday behaviors, covering various non-ideal real-world scenarios such as motion blur, occlusion, and dramatic pose changes. For these action sequences, the dataset is accompanied by multi-source reference images, including anime characters, half-body portraits, transgender/cross-style figures, and pose images significantly different from the actions, to simulate common problems in real-world applications such as spatial inconsistencies and temporal discontinuities.

Citation

@misc{zhang2025steadydancer, title={SteadyDancer: Harmonized and Coherent Human Image Animation with First-Frame Preservation}, author={Jiaming Zhang and Shengming Cao and Rui Li and Xiaotong Zhao and Yutao Cui and Xinglin Hou and Gangshan Wu and Haolan Chen and Yu Xu and Limin Wang and Kai Ma}, year={2025}, eprint={2511.19320}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2511.19320}, }