Thirty seconds in one pass
A single generation runs from 2 to 30 seconds at 1080p, with audio produced alongside the picture. Leave duration unset and the model picks a length to suit the prompt.
Alibaba (Tongyi Lab)
Wan 3
30s 1080p single pass, thinking mode, document and web references. Thirty seconds of 1080p with native audio in a single pass, plus a thinking mode that reasons about composition before it renders and can read documents and web pages as reference material.
- Duration
- 2–30s
- Resolutions
- 480p, 720p, 1080p (default)
- Audio
- Native, on by default
- Aspect ratios
- 16:9, 4:3, 1:1, 3:4, 9:16, model-picked
- Thinking mode
- Optional, required for doc and web refs
Features
What Wan 3 does differently.
Thinking mode
The model reasons about composition before it starts rendering. It is also the mode that unlocks document and web references, so complex briefs are planned rather than improvised.
Rich reference inputs
Up to ten images, five videos of 15 seconds each and five audio clips of 15 seconds each. With thinking mode on, documents and web pages count as references too.
Full-body motion at speed
Dance battles, crowds and neon-lit movement hold together rather than smearing, so the model suits shots where the whole body is in frame and moving fast.
Examples
Prompts that show its range.
Woman late twenties bright bedroom mint-green serum bottle talking to camera, vertical handheld phone, natural skin texture warm voice
Try this promptTwo hands open matcha latte box, dip fingers into powder jar fine stream, vertical overhead soft light, quiet room tone
Try this promptChef's hands finish ramen bowl, broth steaming chashu chilli oil red rings, vertical handheld 35mm warm tungsten
Try this promptWoman walks toward camera golden hour, camel coat catching wind, boots wet cobblestone, vertical 50mm backlit flare
Try this promptThese prompts and clips are reproduced from Alibaba's own Wan 3 material. Run them in the Studio to see your own render.
Workflows
Every way to run Wan 3.
- Text to videoPrompt to video, audio in the same passFrom 50 credits
- Image to videoAnimate a still into a clip of up to 30 secondsFrom 50 credits
- Reference to videoImage, video, audio, document and web referencesOpen workflow
Cost is charged per second and steps up with resolution, so 480p drafts are far cheaper than final 1080p renders. Credits come from one shared balance across image, video and speech, and the exact estimate is shown in the Studio before you generate.
Built for
- Spokesperson clips in vertical phone framing with natural skin texture
- Product shots such as unboxings, powder pours and smudged jars
- Food footage with steam, chilli oil rings and kitchen ambience
- Fashion work with wind-caught coats, wet cobblestones and backlit flare
FAQ
Wan 3, answered.
Limits, audio, licensing and price — the things worth knowing before you spend a credit.
Half a minute in a single take.
Open the Studio, pick Wan 3, and see the credit estimate before you spend anything.