Wan-Animate-2 framework enhances character animation with direct video input

Researchers have introduced Wan-Animate-2, a novel framework for character animation that directly processes driving videos using a redesigned Diffusion Transformer. This approach enhances motion generation fidelity and preserves character identity by removing intermediate motion extractors. The framework also incorporates text-driven viewpoint control, allowing for independent adjustment of the output camera perspective from the input video. Additionally, an efficient variant, Wan-Animate-2-Lite, has been developed to achieve real-time inference speeds suitable for streaming animation. AI

IMPACT This framework could enable more efficient and high-fidelity character animation for various applications, potentially impacting content creation and virtual environments.

RANK_REASON The item describes a new research framework and its associated model weights and inference scripts, fitting the definition of a research release. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Hugging Face Trending Models →

• Bang Zhang
• Dechao Meng
• Diffusion Transformer
• Gang Cheng
• Guangyuan Wang
• Hai Xu
• Hugging Face
• Ke Sun
• Mingyang Huang
• Peng Zhang
• Qwen3.7-Plus
• Ruoshi Zhang
• Wan-Animate-2
• Wan-Animate-2-Lite
• Xingjun Wang
• Zhe Zhang
• Zhongyi Zhang

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →