Wan

Our verdict

The permissive end of open-weight video, and a useful yardstick for everything else in the category. Apache-2.0 on the models, no territorial carve-out, no monthly-active-user ceiling, and an explicit statement that the project claims no rights over what you generate. Set that beside HunyuanVideo's licence and the contrast is stark. The cost is simply hardware: there is no hosted tier, no support, and no fallback if your GPUs are not up to the job.

Key features

  • Open-weight video generation with weights published for local running
  • Apache-2.0 licence covering the models in the repository
  • The project states it claims no rights over generated content
  • Published on both GitHub and Hugging Face
  • Runs on your own hardware with no metering of any kind

Strengths

  • Apache-2.0, a genuine open-source licence, read from the LICENSE.txt in the project's own repository
  • The README states it plainly: "We claim no rights over the your generated contents"
  • No territorial restriction, no monthly-active-user ceiling and no acceptable-use gate on the grant itself
  • Weights published openly rather than behind a request form
  • A clean yardstick for judging how conditional other open-weight video licences actually are

Limitations

  • Running a large video model well needs serious GPU capacity, which is the real cost
  • The project still sets conduct expectations: you are "fully accountable" for your use of the models
  • No hosted option from the project itself, so there is no fallback if you lack hardware
  • Quality and speed depend heavily on which checkpoint and resolution you run
  • No vendor support, SLA or commercial contact behind it

Who should skip Wan

Teams without GPU capacity or the appetite to operate it, since there is no hosted tier at all. Also anyone who needs vendor support standing behind their video pipeline.

Pricing details

Checked on the official pricing page on . Prices change often, so confirm before you buy.

What the free plan includes

  • No paid tier and nothing withheld. Apache-2.0 covers the models in the repository, and your hardware is the only limit.

Worth knowing

Wan has no pricing page and no commercial tier. It is published on GitHub and Hugging Face under Apache-2.0, so the cost is hardware and electricity rather than licensing. Checked 23 October 2026.

Commercial use

Allowed. The project's README states that "The models in this repository are licensed under the Apache 2.0 License", and the repository ships the Apache License 2.0 as LICENSE.txt. Apache-2.0 permits commercial use, modification, redistribution and running the model as a service for clients, carries an express patent grant, and requires only that notices are retained; it grants no trademark rights. On output the project is unusually direct: "We claim no rights over the your generated contents, granting you the freedom to use them while ensuring that your usage complies with the provisions of this license." That is paired with conduct expectations rather than scope limits - you are "fully accountable" for your use, which must not involve content violating applicable laws, harming individuals or groups, spreading misinformation or targeting vulnerable populations. Those are obligations about how you behave, not territorial or scale restrictions on the grant, which is what separates this licence from the more conditional open-weight video models. Checked 23 October 2026.

Alternatives to Wan

  • HunyuanVideo

    Tencent's open-weight video generation model. Free to download and run, but the licence excludes the EU, the UK and South Korea entirely, output included.

  • ComfyUI

    A node-based local inference engine for diffusion models, letting you wire up reproducible image and video workflows on your own GPU. GPL-3.0 licensed.

  • Runway

    Runway is a browser-based AI video creation suite for generating, transforming and editing short-form visual content from text, images and video.

Guides