Core Parameters
Wan3.0 video generation and reference asset parameters
Before generating, it's best to nail down your creative intent using the table below. Specific API fields may vary by version.
| Dimension | What to Define |
|---|---|
| Duration | 10 seconds, 30 seconds, or smart-length |
| Aspect Ratio | Landscape/portrait and target aspect ratio |
| Subject | Character, animal, product, or scene subject |
| Reference Assets | The purpose and reference name of each asset |
| Camera | Shot size, angle, movement, cuts, and sequencing |
| Motion | Starting state, motion process, force, and ending state |
| Style | Realistic, animated, cinematic, advertising, retro, etc. |
| Sound | Dialogue, voice timbre, ambience, sound effects, and music |
| Constraints | Prohibitions like subtitles, deformation, clipping, and drifting |
Parameter Ordering
Duration/Aspect Ratio → Subject & Reference Assets → Scene & Style → Shot & Motion → Dialogue/Sound Effects → Exclusions → TimelineMultimodal Reference Principles
- Each asset should carry exactly one clear responsibility — such as character, scene, motion, voice timbre, or shot composition.
- Specify which details must be preserved and which are allowed to vary.
- Video references should describe pacing, camera logic, and subject movement — not just "reference this video."
- Audio references should be bound to a specific character or sound layer, with the relationship between dialogue and ambient sound spelled out.
Wan 3.0 Docs