Description
Wan 2.2 is an open-source video generation model described as the first to bring a Mixture-of-Experts (MoE) architecture into video diffusion. Built by Alibaba's Tongyi Lab, it lets users create cinematic videos from text or images at 480P and 720P resolution, with complete model weights available for download on GitHub.
Read More
The model emphasizes cinematic vision control through a deep command of shot language, offering fine-grained control over lighting, color, and composition for versatile styles. Its I2V-A14B image-to-video model delivers stable video synthesis with realistic physics, natural movement, and reduced unrealistic camera movements, while advanced motion understanding recreates complex motion such as dancing, parkour, and figure skating with enhanced fluidity. The MoE architecture enlarges model capacity while maintaining computational efficiency.
Wan 2.2 also includes an enhanced visual creation pipeline that generates images optimized for downstream video integration, tuned with curated aesthetic data for lighting, composition, and color. Trained on 65.6% more images than previous versions, it improves generalization across motion, semantics, and aesthetics. Being fully open-source, developers and creators can download it and run it themselves, and a global community shares their work built with the model.
- CategoryAI Video Generators
- Pricing ModelOpen source
- DeploymentSelf-hosted
- PlatformsWeb
Wan 2.2's Core Features
Open-source model with complete downloadable weights on GitHub
Mixture-of-Experts (MoE) architecture for larger capacity with efficiency
Text-to-video and image-to-video generation at 480P and 720P
I2V-A14B model for stable synthesis with realistic physics
Advanced motion understanding for complex, fluid movement
Cinematic control over lighting, color, and composition
Video-optimized image generation pipeline with aesthetic tuning
720p output at 24fps with optimized performance
How to use Wan 2.2?
Download the model: get Wan 2.2 with its complete weights from GitHub.
Choose a mode: use text-to-video or image-to-video generation.
Provide input: write a text prompt or supply a starting image, optionally prepared with the visual pipeline.
Set cinematic controls: specify lighting, color, composition, and motion for your desired style.
Generate: produce a 480P or 720P cinematic clip.
Wan 2.2's Use Cases
- Cinematic text-to-video
- Image-to-video creation
- Self-hosted generation
- Motion-heavy content
