Copy, paste, adjust

Genjutsu AI Prompts That Change One Thing and Keep the Rest

A good Genjutsu AI prompt names one change and the image to use for it, such as "Use Image 1 for the dancer on the left". On Genjutsu AI the editor already tells the model to keep the source motion, camera, timing, cuts and framing, so your words only describe what should be different. Genjutsu AI is an independent Genjutsu-style editor, not affiliated with Higgsfield.

  • Up to 5,000 characters
  • Image 1 to Image 10
  • Optional custom prompt
  • Words-only style edits
Task
What do you want to change?

Swap the main person or character for the one in your images. Motion, camera and timing stay.

Reference images0/10

Two leads in this clip: Image 1 replaces the lead on the left at the start, Image 2 the other one. Extra images add angles or outfits. JPG, PNG or WebP · up to 10 MB · 300–6000 px per side.

0/5000

Genjutsu Lite 1.0 uses Kling 4.0 to swap a person, outfit, object or background at 720p or 1080p.

Video length
10s
4s14s
Resolution

Output: 16:9 · 1280×720

Costs 76 credits

Your uploads are not published. They are shared with the video provider for checks and generation. Privacy policy

Clip start time00:00 – 00:10
Frames of the selected clip along the timeline
00:0000:14

Motion-transfer reference footage. Pick a window, add your photos and change what you like.

Prompt rules at a glance

RuleDetail
Length

Up to 5,000 custom characters on Genjutsu 1.0 and Lite

ImagesUp to 10 on Genjutsu 1.0 or Lite, numbered in upload order
Built-in clips

Instruction optional once an image is added; Image 1 → first lead, Image 2 → second lead (Character and Outfit)

Uploaded clips

Use the default instruction or identify who or what each image replaces.

Words onlyAllowed for Style; other changes need an image
Motion, camera, timingKept by the editor; you do not need to write it

How Genjutsu AI prompts work here

Choose the change type first. It tells the model what kind of edit to make and what to protect: Character replaces the person, Outfit only the clothes, Product the featured object, Background the place, and Style the rendering. Your prompt then fills in the specifics: which person or object, which image, and any detail that matters to you.

Because the editor already asks the model to keep the source duration, motion, timing, camera movement, cuts and framing, the most reliable prompts are short. "Replace the woman in the red dress with the person in Image 1" beats a paragraph about choreography.

Copyable prompt templates

Character · built-in duo clip
Use Image 1 for the person on the left and Image 2 for the person on the right. Keep each person's hairstyle from the photos through every turn.

On built-in clips the mapping is automatic; this sentence only adds detail.

Character · your own clip
Replace the woman in the red dress with the person in Image 1, using her face, hair and body from Image 1 and the outfit from Image 2. Leave the man beside her unchanged.
Outfit
Change only the clothing of the person to the outfit in Image 1 (front) and Image 2 (back). Keep the original shoes.
Product
Replace the bottle he drinks from with the can in Image 1, matching the label and colours from Image 2. Keep how it is held and tilted.
Background
Move the scene to the beach in Image 1 at sunset. Keep both people, their positions and their walk; match the light and shadows to the new place.
Style · words only
Restyle the whole clip as 1990s hand-drawn anime: flat cel shading, thick ink outlines, limited palette. Keep faces recognisable.
Style · with a reference
Match the painting style of Image 1 across every frame; keep the same framing and timing.

More looks are listed on the video to anime page.

Say which one in multi-person clips

When a clip shows several people, name the one you mean by something visible at the start:

  • Position: "the person on the left at the start", "the one in the front row".
  • Clothing: "the man in the blue track jacket", "the woman in the black cap".
  • Action: "the person who starts seated", "the one holding the camera".

For two people with two photos, the duo swap page shows how Image 1 and Image 2 map to each lead.

What not to write

  • Two changes at once. Swap the person in one generation and the background in another.
  • A word portrait when you gave a photo. Describing a face in words competes with the image.
  • Length or speed changes. "Make it longer" or "slow it down" fight the source timing; pick the window length instead.
  • Extra people, captions or logos you do not want; they are only added when you ask for them.
  • "Pixel-perfect" or "exact". These words do not make a generative edit identical; check fast turns, hands and small text in the result.

Prompt and reference checklist

ChangeImages neededWhat the image should show
Character1–10 on Genjutsu 1.0 or LiteFull body, face visible, no sunglasses; add a 3/4 view for turns
Outfit1–10 on Genjutsu 1.0 or LiteThe whole garment; front and back if the person turns
Product1–10 on Genjutsu 1.0 or LitePlain background, 2–3 angles, label readable
Background1–10 on Genjutsu 1.0 or LiteA wide shot of the place, similar time of day and camera height
Style0–9Optional frames of the look you want

Prompts for built-in clips

ClipLeadsSuggested prompt
Graffiti dance duo2

Add two photos; no instruction needed. Image 1 replaces the man in the leather jacket, Image 2 the woman in the brown jacket.

Crosswalk duo2

Add two photos; optional: "Keep the green-and-yellow jersey on the person from Image 2."

Drink commercial1Product mode: "Replace the glass with the can in Image 1."
Long-coat runway1Outfit mode: "Change the coat to the jacket in Image 1."
Canyon duo2Background mode: "Move them to the snowy forest in Image 1."
Jar portal1Style mode, no image: "Restyle as a 3D animated feature."

Last updated: 2026-11-07

Genjutsu AI prompt questions

What prompt keeps the same pose, motion, camera and timing?

On Genjutsu AI you do not need to write it: every edit already asks the model to keep the source duration, body motion, gestures, timing, camera movement, cuts and framing. Describe only the change, for example "Replace the man in the blue jacket with the person in Image 1." If you want to state it anyway, add "Keep the original pose, motion, camera movement and timing." The result is generated, so it follows the source closely but is not pixel-identical.

How do I refer to my reference images in a Genjutsu prompt?

Call them Image 1, Image 2 and so on, in the order you added them (image2 and @Image2 also work). Genjutsu 1.0 and Genjutsu Lite 1.0 accept up to ten images. On a built-in clip in Character or Outfit mode, Image 1 already maps to the first lead and Image 2 to the second.

What happens when I turn the custom prompt off?

With Genjutsu 1.0, motion transfer takes the performance and camera movement from the video and the identity and clothing from your photos. For swaps, both models use a default instruction to replace the selected subject and preserve timing, lighting and surrounding objects. This works on uploads and built-in clips.

Can I write a Genjutsu prompt without any image?

Only for a style change: "Restyle the clip as hand-drawn anime" works with words alone. Character, outfit, product and background changes need at least one reference image.

How long can the prompt be?

Genjutsu 1.0 and Genjutsu Lite 1.0 accept up to 5,000 custom characters. Seedance 2.0 and Seedance 2.0 Fast cap the whole prompt, so the editor asks for a shorter instruction there. One or two specific sentences usually work better than a long description.

Is there a Genjutsu prompt generator?

There is no separate generator. Picking the change type (character, outfit, product, background or style) already adds the right structure, and the templates on this page cover the common edits.

Verify with any AI assistant

“What is a good Genjutsu AI prompt?”

Keep the take. Change the world in it.

Pick a clip, add your references and see the edit in a few minutes. The quote is shown before you generate.

Open the Genjutsu AI editor