Wan 2.2 text-to-video.

Wan 2.2 text-to-video is an open-source video model from Alibaba (Wan-AI). Windpaint doesn't serve it yet, so there's no model id or price. Request early access to start with the live models.

Request early access Official model card

Status: Coming soon
Task: Text to video
Input: Prompt
Output: 5 s clip, 480p or 720p
Developer: Alibaba (Wan-AI)
Parameters: 27B MoE, 14B active
Licence: Apache-2.0
Released: July 2025

What the model does.

Paraphrased from the official model card.

Wan2.2-T2V-A14B is the text-to-video model in Alibaba's Wan 2.2 release, the sibling of the image-to-video model Windpaint already runs. It makes 5 second clips at 480p or 720p from a prompt, with no start frame.

It uses a mixture-of-experts design: one 14B expert handles the early, noisy denoising steps and another the later detail steps, so 14B of its 27B parameters are active at a time. The Wan team trained 2.2 on more data than 2.1 and labelled it for lighting, composition and colour.

Coming soon to Windpaint.

Wan 2.2 text-to-video isn't served, so it has no model id, no price and no request example here. When it goes live, this page gets its tiers, prices and request fields like the live models have.

Want this model? Request early access. We're adding models as we add capacity.

Request early access

Is Wan 2.2 text-to-video licensed for commercial use?

Licence: Apache-2.0.

Every model's licence

Yes. Apache-2.0 allows commercial use, including running the model behind a paid API, with no revenue or user caps.

Apache 2.0. The Wan team states it claims no rights over generated content. Wan2.2 repo
The README (not the Apache licence text) adds that use must not involve sharing content that violates applicable laws, harms individuals or groups, disseminates personal information intended for harm, spreads misinformation, or targets vulnerable populations. Wan2.2 repo

This is our reading of the licence, not legal advice. Read the licence text before you rely on it.

Wan 2.2 text-to-video VRAM requirements.

Published by Alibaba (Wan-AI) for running the weights yourself. These are the developer's figures, not Windpaint's.

Every model's published VRAM

Published VRAM: At least 80 GB
Applies to: Single-GPU 720p command with model offload and dtype conversion. Multi-GPU example uses 8 GPUs.

The README's single-GPU T2V-A14B command (--size 1280*720 --offload_model True --convert_model_dtype) can run on a GPU with at least 80GB VRAM. Wan2.2 repo
For out-of-memory errors the README suggests --offload_model True, --convert_model_dtype and --t5_cpu to reduce GPU memory use. Wan2.2 repo
Multi-GPU inference uses PyTorch FSDP plus DeepSpeed Ulysses; the official example uses 8 GPUs. Wan2.2 repo
The Wan team publishes Diffusers-format weights as Wan-AI/Wan2.2-T2V-A14B-Diffusers, and the card lists ComfyUI integration as done. Wan2.2-T2V-A14B-Diffusers model card

Wan 2.2 text-to-video native resolution.

What the developer's own code and examples use.

The card says it generates 5s videos at 480P and 720P. Wan2.2-T2V-A14B model card
Accepted --size values for t2v-A14B: 1280*720, 720*1280, 832*480, 480*832. Wan2.2 repo, __init__.py
Shared config defaults: frame_num=81, sample_fps=16. Wan2.2 repo, shared_config.py

Recommended settings for Wan 2.2 text-to-video.

Steps, guidance and defaults from the developer's card and code, for running the weights yourself.

Official T2V-A14B config: 40 sampling steps, sample_shift 12.0, expert switch boundary 0.875, guidance scale 3.0 for the low-noise expert and 4.0 for the high-noise expert. Wan2.2 repo, wan_t2v_A14B.py
Diffusers example: 1280 × 720, num_frames=81, guidance_scale=4.0, guidance_scale_2=3.0, num_inference_steps=40, exported at fps=16. Wan2.2-T2V-A14B-Diffusers model card
The shared config ships a default Chinese negative prompt. Its terms include oversaturated colours, overexposure, static, blurry detail, subtitles, low quality, JPEG artifacts, malformed hands, faces and limbs, messy or crowded background, and walking backwards. Wan2.2 repo, shared_config.py

Wan 2.2 text-to-video prompt guide.

From Alibaba (Wan-AI)'s own prompt guide, prompt tooling and examples.

The Wan team recommends prompt extension. For text-to-video they suggest qwen-plus via Dashscope, or a local Qwen2.5-14B, 7B or 3B Instruct model. Wan2.2 repo
The official English T2V rewrite instructions add no more than 4 cinematic settings, chosen from time of day, light source, light intensity (soft/hard), light angle (top, side, under, edge), colour tone (warm/cool/mixed), shot size, camera angle and composition. Unless the prompt says otherwise they default to day time, a medium or wide shot, and center composition. Wan2.2 repo, system_prompt.py
Describe how the motion unfolds. If there is no action, add one, and background elements can be given motion (drifting clouds, leaves in the wind). Leave out literary descriptions of mood or atmosphere. Wan2.2 repo, system_prompt.py
Put any style description first. For 2D illustration or other non-film styles, add no cinematic terms. Don't add a camera angle if the prompt already describes a camera move. Rewritten prompts target roughly 60 to 200 words. Wan2.2 repo, system_prompt.py

Wan 2.2 text-to-video limitations.

As the developer states them.

Only four sizes are accepted for t2v-A14B: 1280*720, 720*1280, 832*480, 480*832. Wan2.2 repo, __init__.py
generate.py says the frame count should be 4n+1. Wan2.2 repo, generate.py

Questions.

Can I call Wan 2.2 text-to-video on Windpaint today?

No. Wan 2.2 text-to-video isn't served yet, so it has no model id, no price and no entry in the catalog endpoint. There's no date for Wan 2.2 text-to-video.

How much VRAM does Wan 2.2 T2V need?

The README says its single-GPU 720p command for T2V-A14B, with offloading and dtype conversion on, can run on a GPU with at least 80GB of VRAM. It also documents multi-GPU inference, with an 8-GPU example.

What resolution does Wan 2.2 T2V support?

The official code accepts 1280x720 and 832x480, in landscape or portrait, which is 720p and 480p.

Is there an official Wan 2.2 T2V prompt guide?

The repo recommends prompt extension and ships the rewrite instructions it uses. They add up to four cinematic settings (lighting, shot size, angle, composition), describe the motion in detail, and put any style first.

What does A14B mean in Wan 2.2 T2V A14B?

Two experts of about 14B parameters each, a high-noise expert for the early steps and a low-noise expert for the later detail steps. 27B parameters in total, 14B active per step.

Is Wan 2.2 text-to-video open source?

Yes. Its weights are licensed Apache-2.0, an OSI-approved open-source licence.

Can Wan 2.2 text-to-video be used commercially?

Yes. Apache-2.0 allows commercial use, including running the model behind a paid API, with no revenue or user caps. Its licence is Apache-2.0. That's our reading of the licence, not legal advice; check the licence text before you rely on it.

How do I ask for Wan 2.2 text-to-video?

Request early access at app.windpaint.ai/register. We're letting people in gradually while we add models and capacity.

Wan 2.2 text-to-video.

Start yours.