Wan 通义万相
Alibaba's text-to-video model
Overview
Wan (通义万相, Tongyi Wanxiang) is Alibaba's text-to-video and image-generation product family. The Wan 2.1/Wan 2.2 video variants are competitive with Hailuo and Kling, with strong support for Chinese-prompt video generation. Available via Alibaba Cloud and the Tongyi consumer app.
---
通义万相是阿里巴巴旗下文生视频与图像生成产品矩阵。Wan 2.1/2.2 视频版本与海螺、可灵形成正面竞争，中文 prompt 的视频生成表现突出。可通过阿里云与通义 App 使用。
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 20 May 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
What you can produce with Wan 通义万相
- Generate text-to-video clips locally and offline by running the Apache-licensed Wan 2.2 weights in ComfyUI.
- Animate a still image into a short video with strong motion quality using the image-to-video pipeline.
- Produce videos from Chinese-language prompts, including legible Chinese text effects rendered inside the video.
- Fine-tune or stack community LoRAs on the open weights to lock in a specific character or visual style.
- Call the newer Wan 2.5/2.6 models through the Alibaba Cloud API to add cinematic video generation to an app.
- Transfer motion from a reference video onto a generated character for controlled, choreographed movement.
- Generate marketing images and video variants in the Tongyi consumer app without any local hardware.
ASEAN Perspective
Wan 通义万相 in Southeast Asia
阿里同时开源 Wan 2.1 权重，与混元视频共同构成国产开源视频模型双极。
Wan (Tongyi Wanxiang) is Alibaba's open-source video generation family and one of the most capable open video models available, with the Wan 2.x series widely regarded as among the most cinematic open options thanks to its Mixture-of-Experts diffusion design. Crucially, the weights are released on Hugging Face and ModelScope, so teams can self-host and fine-tune, while a hosted path exists via Alibaba Cloud Model Studio.
It suits developers, researchers and studios who want strong text-/image-to-video quality with the freedom of open weights, avoiding per-clip SaaS pricing. Caveats: self-hosting demands serious GPU resources and technical skill, the polished SaaS UX of closed rivals isn't matched out of the box, and as with all video AI, long and complex motion remains imperfect. Good ASEAN readiness via open weights and Alibaba Cloud's global infrastructure and API, though the consumer-facing site and docs lean China-first.
What people say
Wan (Tongyi Wanxiang) has become the open-source workhorse of AI video. Alibaba released Wan 2.1 and then Wan 2.2 (July 2025) under the Apache 2.0 licence, and Wan 2.2's mixture-of-experts design ships in a 5-billion-parameter variant that runs on consumer GPUs alongside a 14-billion-parameter variant for higher quality. The line has since advanced to Wan 2.5 and the Wan 2.6 series (which lets users insert themselves into generated videos), though those newer commercial variants are API-only via Alibaba Cloud — a shift some open-source users grumble about.
Community sentiment around the open models is among the most enthusiastic of any video tool. On Reddit, Wan 2.2 threads regularly top the AI-video communities — a motion-transfer showcase drew over 6,000 upvotes, and a thread on seamless 20-second 720p long-form generation drew 2,100+ upvotes — with users praising motion quality, prompt adherence and how well it composes with third-party tooling. Native ComfyUI support with official workflow documentation has made it the default local video model for hobbyists and VFX tinkerers, and Chinese-language prompting is notably stronger than in Western rivals.
The trade-offs are the usual open-source ones. Getting the best out of the 14B model requires serious GPU hardware and workflow literacy; the 5B model is accessible but visibly softer. There is no polished consumer editor around the weights themselves — casual users are steered to the Tongyi app or Alibaba Cloud's paid API, and the best community results come from users comfortable assembling node graphs, LoRAs and upscalers.
Wan fits developers, ComfyUI users and studios that want free, locally run, commercially usable video generation — especially with Chinese prompts — and teams happy to consume Alibaba's API for the newest models. Creators who want a one-click web tool with support will be better served by packaged rivals like Kling, Hailuo or Vidu.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Wan 通义万相's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Wan 通义万相 unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Wan 通义万相 directly →
Spotted something out of date? Suggest an update →
Alternatives to Wan 通义万相
More in Video & Audio
