Wan 3.0 Pro Prime reference-to-video locks a face, garment, set, movement or voice using up to 10 image, 5 video and 5 audio references. It renders 2 to 30 seconds at 1080p, 2K or 4K, enlarged from 1080p above that.
Workbench
These examples were written for this endpoint. Replace the prompt and sample media URLs with your own accessible inputs, then check the current quote and model availability before submitting.
- Output5s
two-character-plates Two `reference_image_urls` plates, told apart only by order in the array — the prompt calls the first 'the woman' and the second 'the man'. Both faces and haircuts carry into one frame at `resolution` `1080p`: she looks up from the workbench as he leans over her shoulder, lit by one hanging bulb. `seed` `220913` at `aspect_ratio` `16:9` keeps that framing reproducible.
- 220913
- 1080p
- 16:9
- Output2s
dancer-three-stages-4k A third stage for a performer and a costume that already exist, rendered at 1080p, the lowest tier here: layered chiffon under hard backlight is the hardest texture there is to hold. Three stills carry the dancer and the garment, so the prompt describes only this lighting set-up. This run sent images only; clips and audio can go in alongside them, which means movement can be referenced instead of written.
- 1080p
- 2s
- Output5s
set-continuity-4k One `reference_image_urls` plate — the ropes, the lockers, the single hanging bulb — rendered at `resolution` `4k` with the camera locked off and no people in frame. The prompt asks only for dust drifting through the light and a slow swing on the bulb; nothing else moves, which makes the plate reusable as a set anchor. `generate_audio` is `false` and `duration_seconds` is `5`.
- 4k
- 5s
This example starts at $0.234 per output second. The Playground updates the estimate when you change the settings.