[Community Work] Request to Add One-to-All Animation in Community Work

[ssj9596]

We would like to request adding One-to-All Animation to the Community Work section.
Our method is based on Wan 2.1, and we provide both the 1.3B and 14B model variants.

One-to-All Animation (One-to-All) supports both aligned and misaligned pose-driven animation.
Beyond standard aligned pose animation, it can handle non-aligned settings such as half-body to full-body transfer and other cross-proportion mappings. This allows a single reference image to be driven by any dance video across varying scales, while also supporting its misaligned subtask of image pose transfer.

Additionally, One-to-All enables long video generation, achieving seamless continuity across multiple chunks.

We have open-sourced the entire pipeline, including:

• Training code
• Multi-resolution, multi–aspect-ratio, and multi-frame training framework
• A cross-domain cartoon dance dataset
• Reproduction scripts to fully replicate the metrics reported in the paper

Project Page: https://ssj9596.github.io/one-to-all-animation-project/
ArXiv: https://arxiv.org/abs/2511.22940
Code: https://github.com/ssj9596/One-to-All-Animation

Thank you very much for your amazing work!