Wan2.2-Animate - An open-source action generation model from Alibaba Tongyi

What is Wan2.2-Animate?

Wan2.2-Animate is a motion generation model developed by the Alitongyi team. The model supports both motion mimicry and role-playing modes, accurately replicating facial expressions and movements from performer videos to generate highly realistic character animation videos. The model can seamlessly replace animated characters into the original video, perfectly matching the scene's lighting and color tone. Built on the Wan model, it controls limb movements through spatially aligned skeletal signals and recreates expressions using implicit facial features extracted from the source image, achieving highly controllable and expressive character video generation. The Wan2.2-Animate model can now be experienced online directly on the Alitongyi Wanxiang official website.

Main functions of Wan2.2-Animate

• Action imitation Input a character image and a reference video. The model will transfer the character's movements and expressions from the reference video to the input image character, giving the static image character dynamic expressiveness.
• Role replacement While preserving the original video's actions, expressions, and environment, the characters in the video are replaced with characters from the input images, achieving seamless integration.

The technical principle of Wan2.2-Animate

• Input paradigmThe input paradigm of the Wan model was modified to unify the reference image input, time frame guidance, and environmental information into a common symbolic representation, thus adapting to the needs of character animation tasks.
• Body movement control: Replicating body movements using spatially aligned skeletal signals. Skeletal signals can accurately describe a character's limb movements, enabling precise control of character limb movements by combining the signals with the character image.
• Facial expression controlThe method uses implicit facial features extracted from the source image as driving signals to reproduce facial expressions. Implicit features can capture subtle changes in a character's facial expressions, achieving highly realistic expression reproduction.
• Environmental integrationTo enhance the integration of characters with new environments during character swapping, an auxiliary Relighting LoRA module was developed. This module maintains character appearance consistency while matching the lighting and color scheme of the new environment, achieving seamless environmental blending.

How to use Wan2.2-Animate

• Visit the Tongyi Wanxiang official websiteVisit the official website of Tongyi Wanxiang and log in to the platform.
• Upload pictures and videosUpload a character image and a reference video.
• Select modeChoose between action imitation or character replacement mode.
• Generate animationClick the "Generate" button, and the model will automatically process the output results.

The project address for Wan2.2-Animate

• Project official websitehttps://humanaigc.github.io/wan-animate/
• HuggingFace model libraryhttps://huggingface.co/Wan-AI/Wan2.2-Animate-14B
• arXiv technical paperhttps://arxiv.org/pdf/2509.14055

Application scenarios of Wan2.2-Animate

• Video editingIn video editing, you can replace the characters in the video with animated characters, and they can be perfectly integrated into the original video environment to achieve a seamless visual effect.
• Game developmentIn game development, character animations are generated in real time based on player motion capture data, making the movements of game characters more natural and fluid, and enhancing the game's immersion and interactivity.
• Virtual Reality and Augmented RealityIn virtual reality (VR) and augmented reality (AR) applications, create realistic virtual characters that interact naturally with users, providing them with a more authentic and immersive experience.
• Education and TrainingIn the field of education, animated characters can be created as teaching assistants to attract students' attention through their expressions and movements, thereby increasing the fun and interactivity of teaching.