What generative models does Wan AI use?

The platform is built around the Wan family of generative models, specifically Wan 2.1 and Wan 2.2. These models cover text-to-image, image-to-image, text-to-video, image-to-video, and image editing.

Is there a public API for Wan AI?

No, the hosted web product does not offer an obvious public API. Developers requiring API access or self-hosting should use the openly published model weights via Hugging Face, ComfyUI, or their own inference stack.

Who is Wan AI best suited for?

It targets creators, marketers, and social video producers who want a one-stop shop for short-form AI clips and stills. It is ideal for users who want to try Alibaba's video models without spinning up their own GPU inference.

How does Wan AI pricing work?

The service operates on a freemium model. Specific pricing and quotas are not clearly documented on the marketing page, but it appears to run on a credit or subscription model typical of Alibaba's consumer AI products.

Can I use Wan AI for enterprise applications?

It is not recommended if you need documented enterprise SLAs, a first-party API, or tight programmatic control. For serious pipelines, it is better to use the open checkpoints directly rather than the hosted consumer app.

Wan AI

Alibaba's generative video and image playground, wrapping the open Wan model family in a consumer web app.

FreemiumVideoWan 2.1 / Wan 2.26.5 / 10

Visit website →

In short

Wan AI is a consumer web platform wrapping Alibaba's open Wan 2.1 and 2.2 models for text-to-video, image-to-video, and image editing. It serves as a frictionless way to test state-of-the-art video generation without self-hosting weights. It is best for creators wanting quick access to high-quality AI clips and stills.

Best for

Pick Wan AI if you want the fastest way to try Alibaba's Wan video models for short clips and stills without spinning up your own GPU inference.

Skip if

Skip it if you need documented enterprise SLAs, a first-party API, or tight programmatic control - go straight to the open weights instead.

Wan AI is Alibaba's public-facing creative platform built around the Wan family of generative models (Wan 2.1 and 2.2), covering text-to-image, image-to-image, text-to-video, image-to-video, and image editing in a single web app. It targets creators who want a one-stop shop for short-form AI clips and stills without stitching together separate tools for each modality.

The headline draw is that Wan's underlying video models are among the strongest open-weight video generators released to date, and the hosted site at wan.video is essentially the frictionless way to try them without running the weights yourself on a GPU. It fits creators, marketers and social video producers who want fast T2V/I2V generation with a polished UI, and it's a reasonable alternative to Kling, Hailuo, Runway or Sora-style closed platforms. Pricing on the web app isn't publicly disclosed on the landing page and appears to run on a credit/subscription model typical of Alibaba's consumer AI products.

Because the model weights themselves are openly published, developers who need API access or self-hosting can go around the web app entirely and run Wan through Hugging Face, ComfyUI or their own inference stack. The hosted product is best treated as the convenient front door, while serious pipelines will lean on the open checkpoints.

Editor's take

Wan is one of the more interesting hosted showcases for an open-weight video model, which is a rare combination. The web app is fine for exploration, but the real story is that the underlying Wan checkpoints are available to self-host, so power users should treat wan.video as a demo surface rather than the endgame.

— The AI Tool Bible editorial team

Pros

✅ Backed by Alibaba's Wan video models, which are competitive with top closed platforms
✅ Covers T2I, I2I, T2V, I2V and image editing in one interface
✅ Underlying model weights are openly released for self-hosting
✅ Lower barrier to try state-of-the-art video generation than running weights locally

Cons

⚠️ Pricing and quotas are not clearly documented on the marketing page
⚠️ No obvious public API on the hosted product itself
⚠️ Consumer app UX limits fine-grained control vs ComfyUI workflows

Use cases

text-to-videoimage-to-videotext-to-imageimage editingshort-form content

Frequently asked

What generative models does Wan AI use?: The platform is built around the Wan family of generative models, specifically Wan 2.1 and Wan 2.2. These models cover text-to-image, image-to-image, text-to-video, image-to-video, and image editing.
Is there a public API for Wan AI?: No, the hosted web product does not offer an obvious public API. Developers requiring API access or self-hosting should use the openly published model weights via Hugging Face, ComfyUI, or their own inference stack.
Who is Wan AI best suited for?: It targets creators, marketers, and social video producers who want a one-stop shop for short-form AI clips and stills. It is ideal for users who want to try Alibaba's video models without spinning up their own GPU inference.
How does Wan AI pricing work?: The service operates on a freemium model. Specific pricing and quotas are not clearly documented on the marketing page, but it appears to run on a credit or subscription model typical of Alibaba's consumer AI products.
Can I use Wan AI for enterprise applications?: It is not recommended if you need documented enterprise SLAs, a first-party API, or tight programmatic control. For serious pipelines, it is better to use the open checkpoints directly rather than the hosted consumer app.