How to prompt Wan AI

Photorealistic portrait of a woman in her late 20s, shoulder-length brown hair, light freckles, warm genuine smile, plain sage-green t-shirt, facing camera, chest-up framing, soft even daylight, neutral apartment background, sharp focus on the face.

This is the character-sheet prompt: generate the still with an image model in Media Forge, then attach it as the Avatar alongside your real UGC clip in WAN Replace. The original video's framing, room and lighting survive; only the presenter changes, and the scene's light is re-applied to her so she looks shot, not pasted.

3D character render of a friendly round robot mascot, glossy white shell with orange accents, large expressive blue eyes, simple articulated arms and legs, full-body, standing centered, neutral grey studio background, soft even lighting.

Stylized characters ride the same skeleton as photoreal ones — the model maps the performer's motion onto whatever body you provide. Keep the design simple and fully visible in the still: clean silhouettes retarget reliably, ornate costumes with floating parts do not.

Full-body photo of a young male dancer, athletic build, black joggers and a white hoodie, white sneakers, standing relaxed facing camera, arms at his sides, full figure visible head to toe, bright even studio light, plain background.

WAN Move drives your character with the source clip's choreography — spins, footwork and timing carry over via skeleton signals. Full-body-to-full-body is the rule: feed a full-figure still against a full-figure dance video or limbs get invented at the crop line.

Chest-up portrait of a confident businesswoman in her 40s, silver-streaked dark hair in a low bun, navy blazer over a white blouse, looking directly at camera, neutral pleasant expression, mouth closed, soft office lighting, blurred modern office background.

Facial expression is captured separately from body motion, so a talking-head source transfers the delivery — head tilts, brow raises, mouth movement — onto your character. Record (or pick) a clip with the exact energy you want; the neutral closed-mouth still gives the face room to perform.

Full-body photo of a medieval knight in weathered steel plate armor, no helmet, short grey beard, crimson cloak, gauntlets, standing centered facing camera, full figure head to toe, overcast diffuse daylight, plain stone wall background.

Replace shines when the motion is expensive to reshoot: stunts, sports, choreographed action. The knight inherits the performer's exact movement while the footage's world stays intact. Armor is detail-heavy — raise steps and quality so plate edges stay coherent frame to frame.

Clean anime-style illustration of a cheerful girl with short teal hair and amber eyes, orange hoodie and denim shorts, standing in a relaxed pose facing viewer, full body, flat cel shading, simple pastel background, crisp linework.

Motion transfer works on illustrated casts too — the performance is skeletal, not stylistic, so the cel-shaded look persists while the body follows your source video. Flat, consistent shading in the still keeps the style stable; painterly noise flickers.

Photorealistic portrait of a man in his early 30s, tousled dark blond hair, trimmed stubble, hazel eyes, charcoal crew-neck sweater, friendly direct gaze, chest-up, soft window light from the left, minimal beige studio background.

A consistent synthetic presenter is a reuse discipline: generate this character once, then attach the SAME image to every new source clip, keeping resolution and quality settings fixed and pinning the seed. Ten videos, one face — identity holds because the cast never changes.

Photorealistic full-body photo of a barista in her 20s, curly black hair tied back, dark green apron over a black t-shirt, jeans, standing naturally facing camera, warm cafe-toned lighting, full figure visible, simple neutral background.

When the b-roll is right but the on-camera person can no longer appear in it, Replace swaps them while keeping every camera move and the cafe's light. Match the still's lighting temperature to the footage — the relighting step closes the gap, but starting close ends closer.

A courier cyclist weaves through narrow neon-lit streets at night, rain-slick asphalt reflecting pink and cyan signs, camera tracking alongside then drifting behind, light rain falling, moody cyberpunk atmosphere, cinematic realistic style.

The Wan family's text-to-video lives in Workflow Studio as Wan 2.1 — a fast drafting model. Its native prompt formula is subject + scene + motion + camera language + atmosphere + style, in that order; every slot filled beats a long freeform paragraph.

A matte black smartwatch on a rotating pedestal, minimalist white studio, the strap unclasps and floats apart into components, slow orbit camera, clean bright lighting, calm premium atmosphere, sleek commercial style.

Use Wan 2.1 to storyboard motion ideas cheaply before committing a flagship model to the final render. The formula still applies to objects: the "motion" slot (unclasps, floats apart) is what separates a video prompt from an image prompt.