Wan2.2
Open-source MoE video generation model by AlibabaTurn text or images into 720P video, free and open source
Highlights
What you can do with it
- An independent filmmaker used its film controls for professional shot language
- A researcher studied and modified the MoE code to speed up diffusion research
- A content creator made 720P videos without paying subscription fees
How to use it
- 1 Download the model and code from the Wan2.2 GitHub page
- 2 Pick a function such as text-to-video, image-to-video, or visual enhancement
- 3 Prepare your input, such as a text prompt or a still image
- 4 Set parameters like resolution and frame rate per the documentation
- 5 Run the code and wait for the video to generate
- 6 Review the result and repeat the input and settings to refine it
Who it is for
Pros
- Open source with full weights, free to run and modify
- Strong motion and film-style control at 720P
- Covers both text-to-video and image-to-video
Limitations
- Running it locally needs technical setup and suitable hardware
- Open-source license and model terms can change with updates
- Output still depends on prompts and compute, and 720P is the top resolution
What Wan2.2 is
Wan2.2 is an open-source video generation model created by Alibaba Tongyi Laboratory, described as the first open-source model built on a Mixture of Experts (MoE) architecture. It converts text prompts or still images into video at 480P or 720P with a 24fps frame rate. The model is released with complete weights, so developers, filmmakers, and creators can download, run, and modify it locally instead of paying for a subscription. Its strengths include advanced motion understanding, stable video synthesis with realistic physics, and film-level control over lighting, color, and composition, all aimed at making professional-looking video creation more accessible. Because it is open source, the community can study and build on the same code.
What you can do with it
- Turn text prompts or images into 480P or 720P video at 24fps
- Rely on the MoE architecture for stable, realistic motion and camera behavior
- Use film controls to shape lighting, color, composition, and style
- Download the full model and code to run or customize it yourself
- Study the architecture to learn how modern video diffusion works
Who it is for
- Independent filmmakers who want an open, customizable tool
- AI researchers studying and modifying video diffusion models
- Content creators who need 720P output without paying subscription fees
- Video studios that want to add AI generation to their pipeline
What to watch out for
Wan2.2 is open source, but you still need your own compute and some technical setup to download, configure, and run the model, and high-quality output can be demanding on hardware. The license and model weights are provided by the developers, so the exact terms and future updates can change, and you should review them before any commercial use. As with any generative video model, results vary with your prompts and settings, and you are responsible for how you use the generated footage, including respecting rights and avoiding misleading content.
FAQ
Is Wan2.2 free?
It is open source with complete model weights, so you can use it without subscription fees, though you supply your own compute.
What resolution does it output?
It supports 480P and 720P video at 24fps.
Who made it?
It was created by Alibaba Tongyi Laboratory and released as an open-source MoE video model.
Why it is free
What the official page says： for free
Free limits, trial length and features are decided by the official site and can change at any time.
Product preview
Screenshot archived from the listing page, for preview only.