Wan

Wan (/media/wan) is Alibaba's Wan creative platform on Model Studio (DashScope). It generates images and videos with the Wan2.x model family, including text-to-image, text-to-video, and image-to-video.

Where to configure

Configure a dedicated Wan key under Settings → Media. Wan is an own-key provider — it does not reuse the Qwen AI credential, even though both use DashScope. See Settings for provider setup.

Tabs / operations

Tab	Operation	Key fields	Produces
Text → Image	Image	Model, prompt, negative prompt, size, number of images, prompt rewrite	Still image(s)
Text → Video	Video	Model, prompt, negative prompt, size, duration, prompt rewrite	Video clip
Image → Video	Video	Model, source image, prompt, negative prompt, resolution, duration	Video clip

Generation flow

Image generations complete synchronously via bounded internal polling, so the result appears in the render queue almost immediately. Video generations submit an async DashScope task and complete through the shared poll cron. Finished artifacts land in /files and can be opened in the Designer or posted directly. See Media Studios overview for the common flow.

Caveats

• Wan points at the international DashScope host (dashscope-intl.aliyuncs.com).
• There is no completion webhook; video jobs rely on the media-jobs-poll cron, like Runway and Qwen video.
• The model dropdown is curated because DashScope has no clean per-modality catalog; the combobox also accepts any typed model id.
• Field names are native DashScope params: negative_prompt and img_url route into input, while size, resolution, duration, n, and prompt_extend ride into parameters.
• Wan does not support audio or avatar generation.

---

Verified against v1.0.0