Wan
: How to use it, features, and the business problems it solves
Add bookmark
What is Wan?
Wan (Tongyi Wanxiang) is Alibaba's family of AI video and image generation models with an official online service. Wan2.1/2.2 are open models under Apache 2.0, and the latest Wan 3.0 (beta) supports native 30-second generation. Free daily credits on wan.video, with paid plans from $5/month (billed yearly); billing is in US dollars.
Business problems it solves
About "Wan"
What is Wan?
Wan is a family of AI video and image generation models developed by Alibaba (China), together with its official online service. Its Chinese name is Tongyi Wanxiang. Wan2.1 and Wan2.2 are published as open-weight models under the Apache 2.0 license and are widely used for local generation by creators worldwide, while the latest generation, Wan 3.0 (beta, released August 6, 2026), supports native 30-second generation and up to 20 reference assets including documents and web pages.
There are three main ways to use it: (1) online generation on the official site wan.video (the web app is create.wan.video), (2) the Alibaba Cloud Model Studio API, and (3) local execution of the open models published on GitHub / Hugging Face (via community tools such as ComfyUI). Being able to "try for free first, then scale via API or run locally" is the big difference from closed competitors.
How to use
-
Create an account
Go to wan.video (the app itself is create.wan.video) and log in. A daily check-in system grants free credits, so you can try it without paying.
-
Choose a generation mode in Generate
From "Generate", pick Text to Video / Image to Video, then set the model (e.g., Wan 3.0), resolution, and aspect ratio. Image generation and editing are available from the same interface.
-
Enter a prompt and reference assets
Describe the video you want. Wan 3.0 accepts documents and web pages as references in addition to images, audio, and video (up to 20 assets), and can preserve a referenced person's appearance and voice.
-
Generate and edit
After generation, use instruction-based and reference-based video editing (Precision Video Editing) for partial fixes. Timeline and Canvas views are provided for editing work.
-
Export, or move to API / local
Paid plans allow watermark-free downloads. For volume production move to the Alibaba Cloud Model Studio API; to fix costs, run the open Wan2.2 locally.
Key features
- Text to Video / Image to Video — generate moving footage from words or a single image. Wan2.6 and later support multi-shot storytelling.
- Reference-to-Video (R2V) — pass a reference video containing a person's appearance and voice, and generate new scenes with that person preserved (officially announced with Wan2.6 as China's first).
- Native 30-second generation (Wan 3.0) — 30 seconds of continuous video without stitching, doubling the 15-second ceiling of Wan2.6/2.7.
- Video editing — instruction-based and reference-based partial edits, plus pixel-level box-selection editing.
- Image generation and multilingual text rendering — long-form text rendering in 12 languages, charts and formulas, and fusion of up to 9 images.
- Open weights — Wan2.1/2.2 are published under Apache 2.0, with over 34k combined GitHub stars (as of August 19, 2026). The license permits commercial use and local execution.
Pricing
The online service (create.wan.video) uses USD-based credit plans. Based on the official pricing page as of August 19, 2026.
| Plan | Monthly (billed yearly) | Monthly (billed monthly) | Credits/month | Highlights |
|---|---|---|---|---|
| Free | $0 | $0 | Daily check-in credits | 1 concurrent generation, 6 image styles |
| Pro | $5 | $10 | 300 | Up to 60 videos or 1,200 images, watermark-free downloads, 1080p, 10–30s videos |
| Premium | $20 | $40 | 1,200 | Up to 240 videos or 4,800 images, 8 concurrent generations, all Pro features |
Yearly billing is 50% off the monthly price. Subscriptions auto-renew and can be canceled anytime, but the first month/year is non-refundable after activation. These plans cover generation on create.wan.video only — API use is billed separately (pay-as-you-go on Alibaba Cloud Model Studio). Gift cards and one-off credit purchases are also available.
API pricing is pay-as-you-go and varies by model, resolution, and duration (e.g., Wan2.6 is billed per second). Check the Alibaba Cloud Model Studio pricing page for current rates.
Free plan and trial
On the free plan you can generate videos and images within the free credits earned by daily check-in (1 concurrent generation each; 6 image styles). Watermark-free downloads, 1080p, and 10–30 second videos are paid-plan features, so treat the free tier as a way to evaluate quality and prompts. No credit card is required to start.
Japanese language support
- UI — The wan.video UI is in English; no Japanese UI was confirmed.
- Input (prompts) — No official statement on Japanese prompt support was found. In-image text rendering is announced as supporting 12 languages, but the language list is not published.
- Support — Contact is by email (wan_ai@service.alibaba.com); no Japanese-language support is stated.
Reputation and track record
- The open-source repositories have about 16.9k stars (Wan2.1) and 17.2k stars (Wan2.2) as of August 19, 2026 (measured), and the models are staples of local video generation in the ComfyUI ecosystem.
- The official Wan2.6 announcement (December 16, 2025, Alibaba Cloud official blog) promoted it as "China's first reference-to-video model", preserving a person's appearance and voice from a reference video.
- No Japanese enterprise case studies have been published.
How it differs from other tools
Wan's defining trait is that it runs on both open models and a commercial online service. Unlike closed competitors, one model family covers evaluation on the web, production via API, and fixed-cost local execution.
| Aspect | Wan | Kling | Hailuo | Sora |
|---|---|---|---|---|
| Provider | Alibaba (China, listed) | Kuaishou (China, listed) | MiniMax (China) | OpenAI (US) |
| Open models | Yes (2.1/2.2, Apache 2.0) | No | No | No |
| Max length | 30s (Wan 3.0) | Extend feature | Short clips | Short clips |
| Entry price | Free + Pro $5/mo (yearly) | Free + from US$8.8/mo | Free + from $8.40/mo (yearly) | ChatGPT paid plans |
In video generation, Kling, Hailuo, Sora, and Veo are the main comparison targets. If local execution or self-hosted operation is on your roadmap, Wan with its open versions is practically the only choice. It belongs to the same Alibaba AI family as the Qwen LLM, and Wan video generation is also integrated into the Qwen app.
FAQ
Q. Is Wan free to use?
Yes. After signing up at wan.video, free credits are granted via daily check-in. Watermark-free downloads and 1080p output require a paid plan (Pro or higher).
Q. Can I use it commercially?
The open Wan2.1/2.2 models are Apache 2.0 licensed and can be used commercially within the license terms. For videos generated on the online service (create.wan.video), no explicit commercial-use terms were found on public pages, so we recommend reviewing the Terms of Service (paid plans do allow watermark-free downloads).
Q. Can I pay in Japanese yen?
No JPY display or billing. Prices are in US dollars, converted at your credit card's exchange rate.
Q. Can I cancel? Are refunds available?
Both monthly and yearly subscriptions auto-renew and can be canceled anytime. However, the official pricing page states the first billing period (first month/year) is non-refundable after activation.
Q. Which model is the latest?
As of August 2026, the latest is Wan 3.0 (beta). The lineage runs: open Wan2.1 (February 2025) and Wan2.2 (July 2025), then closed Wan2.5 (audio sync), Wan2.6 (reference video, 15s), Wan2.7 (Thinking Mode), and now Wan 3.0.
Q. Can I run it locally?
Wan2.1/2.2 weights are available on GitHub and Hugging Face and are widely run locally with ComfyUI and similar tools. GPU memory requirements depend on model size (1.3B–14B, etc.).
Provider
Alibaba Group, headquartered in Hangzhou, China, is a major e-commerce and cloud company listed on the New York Stock Exchange and the Hong Kong Stock Exchange. Wan is developed by its AI research arm Tongyi Lab, with Alibaba Cloud providing the service infrastructure. Related products include the Qwen LLM (Tongyi Qianwen), and Wan video generation is integrated into the Qwen app. The official site's footer states "Wan © 2026 Powered by Alibaba Cloud", and a Training Data Summary is also published.
The information on this page is based on content published on the official website as of August 2026. Features and pricing may change; please check the official website for the latest information.
