Create a video with reference assets
Give Wan 3.0 a visual starting point, a motion reference or an audio reference, then tell it how those materials should be used.
Updated 10/29/2026
Start with one explicit act
Choose one adult performer, one action and one camera position. Name the body parts and contact you want to see, then stop. A five-second shot cannot carry undressing, foreplay, a position change and a climax at the same time.
Read the result before adding more. If the action is vague, replace the metaphor with a physical verb. If anatomy drifts, give the camera a simpler angle. The first test is learning how the model reads your wording, not compressing a whole scene into one request.
Every example assumes consenting adults aged 18 or over. Describe people, not real individuals; do not use a likeness without permission.
Text to video: one readable reveal
The performer, the action and the camera each have one clear job. Replace the performer and room with your own idea before adding a second action.
Inputs to prepare
- One approved adult reference or a clear text description.
- One action that can be seen in the selected duration.
- One camera angle and one lighting setup.
Keep the shot stable
Use a locked camera while testing identity and anatomy. Change only one variable per retry, and keep the approved frame so you can compare drift without relying on memory.