← All tools

AI Video Generation & Avatars

Wan (Alibaba)

Alibaba's Wan video models: open-weight Wan2.2 under Apache-2.0, plus the wan.video site for hosted generation.

Visit Wan (Alibaba)

No affiliate links and no sponsorship on this page. Our disclosure policy.

Best for
Pro: editors and motion designers
Free plan
Yes
Agent-ready (MCP / API)
Partly
Platforms
Web, Windows, macOS, Linux
Last verified
Oct 15, 2026

Overview

Wan is Alibaba’s family of video generation models. The Wan2.2 release is open: its code and weights are published under the Apache-2.0 license, so anyone can run it on their own hardware. The wan.video site offers hosted generation for people who would rather not set up a GPU.

Who it fits

It suits technical creators, studios and ComfyUI users who want control and no per-clip fee beyond their own compute. The repository describes variants for text-to-video, image-to-video, audio-driven video (S2V) and character animation or replacement (Animate), and it notes integrations with ComfyUI and Diffusers.

What the plans include

The open weights are free. A smaller 5B model is described as running 720p, 24 fps video on a consumer card such as an RTX 4090. Hosted plans on wan.video were not readable without JavaScript, so check the site for prices.

Limits and alternatives

Local runs take setup and a strong graphics card. ComfyUI is the workflow tool most people use to run open models like Wan, Kling is a hosted generator with a free tier, and Runway offers a full browser editor.

Who it's for

A good fit if you are…

  • Technical creators and studios who want open video models they can run on their own GPUs
  • ComfyUI users who want text-to-video, image-to-video, speech-to-video or character animation nodes

Another tool may suit you better if you are…

  • Beginners who want a click-and-go editor
  • Anyone who needs a published hosted price: wan.video plan details were not readable

Pricing

Wan2.2 weights are free to download under Apache-2.0. Hosted use on wan.video: pricing varies, see the official page. As of Oct 2026, verify.

  • The Wan-Video/Wan2.2 GitHub repository is published under the Apache-2.0 license, with weights on Hugging Face and ModelScope.
  • Running the models locally needs your own GPU. The Wan2.2 TI2V-5B model is described as able to run on a consumer card such as an RTX 4090 at 720p, 24 fps.
  • wan.video's hosted plans were not readable without JavaScript, so no hosted price is quoted.

Prices as of Oct 2026 (checked Oct 15, 2026), verify before buying. Source: vendor pricing page.

Key features

  • Text-to-video and image-to-video with Mixture-of-Experts Wan2.2 models
  • Wan2.2-Animate-14B for character animation and replacement
  • Wan2.2-S2V-14B for audio-driven video
  • Integrated into ComfyUI and Diffusers
  • Open Apache-2.0 weights plus the hosted wan.video site

Integrations: ComfyUI, Hugging Face Diffusers, ModelScope and the wan.video web app.

Agent-ready? Partly. Open code and weights run from scripts, ComfyUI and Diffusers, so agents can drive them. No official MCP server was confirmed.

Pros and good to know

Drawn from the vendor's own documentation and public sources.

Pros

  • Open Apache-2.0 weights you can run yourself
  • Character animation and speech-to-video variants
  • Supported in ComfyUI and Diffusers

Good to know

  • Local use needs a capable GPU and setup time
  • Hosted wan.video pricing was not readable
  • Hosted wan.video features may differ from the open Wan2.2 release

Alternatives

  • ComfyUI (Comfy Cloud): Node-based workflow tool for image, video and audio models: free to run locally, or on Comfy Cloud with an agent that builds workflows.
  • Kling AI: Kuaishou's AI video generator for text-to-video and image-to-video, with credit plans and a free non-commercial tier.
  • Runway: Generate and transform video, images and audio with AI models, paid by monthly credits.

Sources

Spotted a mistake or out-of-date price? We want to fix it, see About for how to reach us.