Wan-Animate is a unified framework that supports two distinct modes:

1. Animation Mode: Take a static character image and a reference video and animate the character so it mimics the movements/facial expressions from the video.
2. Replace Mode: In a target video, replace an existing character / subject with a different (animated) character, while preserving lighting, environmental consistency, motion, expressions, etc.

It is built on top of the Wan video generation model.

Wan-Animate is considered a step forward over prior open-source or research animated pipelines because of its unified architecture, higher fidelity & expressiveness, and aseamless blending of the environment and subject. According to multiple benchmarks, Wan-Animate shows superior performance relative to prior animation baselines and is competitive with closed-source systems.