A good Wan 2.2 image-to-video prompt explains the motion missing from the starting image. If the picture already establishes a creator beside a window, the prompt should say what the creator and camera do next.
Use the image as the scene’s starting point
Wan’s official repository supports image-conditioned generation and describes its image-to-video variants. Your host or local workflow determines which variant and settings are available. For installation, use the separate Wan ComfyUI setup guide.
Before prompting, inspect the starting image. A cropped-off hand cannot provide a clean reference for a complicated product demonstration. Choose a source that already contains the important subject, wardrobe and product geometry.
A simple prompt structure
Subject motion: one visible action.
Camera: fixed, or one deliberate movement.
Environment: only the movement needed for this shot.
Continuity: the identity, clothing and objects that stay stable.
Ending: the pose or composition where the shot settles.This is an editorial planning template, not a special syntax required by Wan. Convert it into a concise natural-language prompt for the tool you use. The six examples below are original briefs to test, not benchmarked outputs.
1. A restrained creator introduction
The adult creator looks toward the camera, makes a small friendly nod and settles into a relaxed expression. The camera stays fixed at chest height. Keep the jacket, hairstyle and background arrangement consistent with the starting image.
Use a clear chest-up portrait. Review the eyes, mouth and final expression. If the face changes during the nod, reduce the motion and compare again.
2. A product-adjacent glance
The creator briefly glances at the bottle resting on the table, then returns their gaze toward the camera. Their hands remain resting beside the bottle. The camera and bottle stay still; preserve the package shape and label placement.
This avoids asking a generated hand to rotate and open the product in the first test. For a real advertisement, verify product accuracy against the authentic source asset.
3. A slow environmental reveal
A slow camera move reveals more of the reading corner while the seated creator remains relaxed and nearly still. Soft light continues from the window on the left. Preserve the face, chair and clothing throughout the move.
Choose an image with enough visible scene context. If newly revealed areas become inconsistent, use a smaller camera move or a separately composed wider starting frame.
4. A simple walking moment
The adult creator takes a few natural steps along the clear path already visible in the image. The camera follows gently from the same angle. Keep the outfit and hairstyle unchanged, with no new objects entering the path.
Start with the lower body visible and a plausible walking pose. Inspect feet, knees and clothing motion. Do not accept a clip merely because the face remains recognizable.
5. A quiet fashion detail
The creator turns their torso slightly to show the jacket’s side profile, then pauses. The camera stays still. Preserve the jacket seams, sleeve length and color while the fabric moves naturally with the turn.
Keep the turn modest. A dramatic spin asks the model to invent large parts of the garment and increases the review burden.
6. Background motion with a stable subject
The creator holds a relaxed pose on the terrace while a light breeze moves a few loose strands of hair and the nearby leaves. The camera remains fixed. Keep the facial features, clothing and furniture placement stable.
Use this when the shot needs life but not a new action. Review whether the “subtle” motion actually stayed subtle throughout the clip.
Diagnose failures by changing one instruction
| Failure | Next controlled test |
|---|---|
| Identity changes during movement | Reduce the subject motion with the same source image |
| Camera motion dominates | Use a fixed-camera version |
| Product geometry changes | Keep the product stationary and simplify nearby movement |
| Too many events compete | Split the brief into separate shots |
Save the prompt, source image, variant and settings beside each accepted clip. For a recurring persona, use the identity review sheet. To create that persona and source content in a hosted studio, start your AI creator in Clout.
