SteadyDancer: Harmonized Human Image Animation
SteadyDancer is a first-frame-preserving I2V animation framework built on Wan2.1. Delivers harmonized and coherent human dance animation with identity consistency.
SteadyDancer
Wan2.1Human AnimationDanceI2VFirst-frame-preserving image-to-video animation framework built on Wan2.1 I2V. Delivers harmonized and coherent human dance animation with robust visual identity preservation and temporally consistent motion.
| Developer | MCG-NJU |
| Release Date | 2025-11 |
| Architecture | Wan2.1 I2V with pose-driven motion modeling |
| Capabilities | Image-to-video human animation, dance motion generation, first-frame preservation |
| Benchmarks | X-Dance, RealisDance-Val |
Overview
SteadyDancer is an Image-to-Video (I2V) paradigm-based framework that achieves harmonized and coherent human animation with robust first-frame preservation. Built on the Wan2.1 I2V model, it enhances it with pose-driven motion modeling to generate stable, identity-consistent character videos.
The framework addresses common challenges in human animation such as identity drift from spatio-temporal misalignments and temporal jitter during complex motion. SteadyDancer generates high-fidelity dance videos that maintain visual identity across frames while accurately following the driving pose sequence.
ComfyUI Integration
SteadyDancer is available in ComfyUI via the Kijai WanVideoWrapper custom nodes. Both fp16 and fp8 scaled model variants are hosted on Hugging Face under Kijai/WanVideo_comfy. The model works with the WanVideoWrapper nodes and supports pose-guided generation workflows.
Resources
| Resource | Link |
|---|---|
| GitHub | github.com/MCG-NJU/SteadyDancer |
| HuggingFace (Kijai) | Kijai/WanVideo_comfy |
| Project Page | mcg-nju.github.io/steadydancer-web/ |
Guides and workflows related to this model series.