Text-to-video
T2V-A14B turns written prompts into video and supports 480P and 720P output.
Open video models
Wan 2.2A foundation for creative control.
Explore Wan 2.2: an open model family for turning written scenes and still images into moving stories.
The Wan3.video workspace currently offers W3.0 and W3.0 Pro. This page is a model guide.
- Open weights
- Build your own workflow
- Text + image
- Two starting points for a scene
- 720P · 24 fps
- With the TI2V-5B model
Wan 2.2
Understand the Wan 2.2 family.
Pick the model for your input and your workflow. The family includes distinct models, each with its own requirements.
Image-to-video
I2V-A14B starts from an image. The smaller TI2V-5B combines text and image inputs in one model.
A workflow you can build on
Official model weights and inference code are available, with ComfyUI and Diffusers integrations for custom workflows.
Start with a clear creative direction
Choose your starting point
Write a scene from scratch, or prepare a reference image when the model supports it. Decide what should move and what should stay consistent.
Describe one focused shot
Name the subject, action, setting, lighting, and camera movement. Start with one clear moment before attempting a more complex sequence.
Review and refine
Check motion, composition, and subject consistency. Change one instruction at a time so you can see what improves the result.
Wan 2.2, explained.
Can I select Wan 2.2 on Wan3.video?
Not currently. The create button opens our W3.0 and W3.0 Pro workspace. For Wan 2.2 itself, use the official model resources linked below.
Is Wan 2.2 open source?
The official repository publishes code and model weights under Apache 2.0. Check the license and requirements for the checkpoint you plan to use.
What hardware does Wan 2.2 need?
Requirements depend on the model. The official TI2V-5B example runs on a 24 GB GPU with offloading; larger A14B models require more resources.
Does every Wan 2.2 model create audio?
No. The family has a separate S2V model for audio-driven video. Do not assume the base text-to-video and image-to-video checkpoints generate a soundtrack.
Model details: official documentation. Options vary by model and service.
Put your next idea in motion.
Ready to create online? Open the Wan 3.0 workspace, choose your settings, and build your first shot.