Six steps, one edit

Genjutsu AI Tutorial: From Clip to Finished Edit

To use Genjutsu AI, pick a built-in clip or upload an MP4/MOV (5 seconds to 5 minutes, up to 50 MB), choose a 5, 10, 15 or 30-second window, select what to change, add up to nine reference images, optionally describe the change, then pick a model and resolution and generate. The credit cost is shown first, and failed jobs refund. Genjutsu AI is an independent Genjutsu-style editor, not Higgsfield.

  • Nothing to install
  • Built-in clips or your upload
  • Credit quote before you start
  • MP4 download
What do you want to change?

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

Reference images0/9

Image 1 replaces the lead in this clip. Up to 9 images; extras add angles or outfits. JPG, PNG or WebP · up to 10 MB · 300–6000 px per side.

0/800

Seedance 2.5 edits your clip directly, so motion, camera and framing stay closest to the source. Up to 30 seconds.

视频时长
分辨率

Output: 16:9 · 1280×720

消耗 175 积分

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

Clip start time00:00 – 00:10
00:0000:10

Motion-transfer reference footage, not results from this tool. Pick a window, add your photos and change what you like.

How to use Genjutsu AI

  1. 01

    Pick the source clip

    Choose one of 28 built-in clips or upload your own MP4 or MOV, 5 seconds to 5 minutes, up to 50 MB and 300–4096 px per side.

  2. 02

    Set the window

    Choose 5, 10, 15 or 30 seconds and drag the window along the timeline to the moment you want to edit.

  3. 03

    Choose what to change

    Select Character, Outfit, Product, Background or Style. Everything you do not select is kept as close to the source as possible.

  4. 04

    Add reference images

    Add up to nine JPG, PNG or WebP images (10 MB each). On a built-in clip, Image 1 replaces the first lead and Image 2 the second.

  5. 05

    Describe the change

    Write one or two sentences naming the change and the image to use. On a built-in clip this is optional once an image is added.

  6. 06

    Pick a model, generate and download

    Choose a model and 480p, 720p or 1080p, check the credit quote, generate, then download the MP4 when it is ready.

The tutorial in one minute

  1. Pick a built-in clip or upload your video.
  2. Set a 5, 10, 15 or 30-second window.
  3. Choose Character, Outfit, Product, Background or Style.
  4. Add up to nine reference images.
  5. Describe the change in a sentence (optional on built-in clips).
  6. Pick a model and resolution, check the credits, generate and download.

The editor above is already open on the Diner walk clip, a 10-second shot with one lead, so you can follow each step as you read.

Step 1: built-in clip or your own video

Built-in clips are reference footage in five categories (duo, street, fashion, cinematic and animated), 5.7 to 30 seconds long. Each one knows who its leads are, so you can add a photo and generate without writing anything. They are source material, not results from this tool.

Your own video must be an MP4 or MOV, 5 seconds to 5 minutes, up to 50 MB and 300–4096 pixels per side. The upload stays private and is only sent for processing when you generate.

Step 2: choose the window

The edit covers one continuous section of 5, 10, 15 or 30 seconds, starting at any whole second. Drag the window along the timeline strip to the moment you want. The 30-second option needs Seedance 2.5 and a source at least 30 seconds long; among built-in clips, only Hotel lobby and Pier walk are that long.

Step 3: choose what to change

ChangeWhat changesWhat staysGuide
CharacterThe person: face, hair, body, costumeMotion, camera, timing, backgroundMotion transfer
OutfitClothing and accessoriesFace, hair, pose, movementOutfit change
ProductThe featured objectHands, camera path, timingObject swap
BackgroundThe place and lightingPeople and their actions

Background change

StyleThe rendering: anime, 3D, claySubjects, composition, timingVideo to anime

Step 4: add reference images

Add up to nine JPG, PNG or WebP images, 10 MB each and 300–6000 pixels per side. They are numbered Image 1, Image 2 and so on, in the order you add them.

  • Character: a full-body photo with the face visible, plus a 3/4 view for turns.
  • Outfit: the whole garment, front and back if the person turns.
  • Product: plain background, two or three angles, label readable.
  • Background: a wide shot of the new place.
  • Style: optional frames of the look you want.

On built-in clips, Image 1 replaces the first lead (left at the start) and Image 2 the second; if you add fewer images than leads, the others stay as they are. Images of people go through the provider's asset review before Seedance 2.5 generation.

Step 5: describe the change

Write 3–800 characters naming the change and the image, such as "Replace the man in sunglasses with the person in Image 1." You do not need to ask for the motion, camera or timing to be kept; the editor already does that. The prompt guide has copyable templates for every change type.

Step 6: model, resolution and credits

ModelHow it uses your clipMax windowResolutions10 s at 720p
Seedance 2.5Edits the selected cut30 s480p, 720p, 1080p175 credits
Seedance 2.0Follows the cut as a reference15 s480p, 720p, 1080p116 credits
MiniMax H3Follows the cut as a reference15 s720p, 1080p77 credits
Seedance 2.0 FastFollows the cut as a reference15 s480p, 720p69 credits
Seedance 2.0 MiniFollows the cut as a reference15 s480p, 720p19 credits

Start with Seedance 2.5 when the motion should follow the source most closely. Use Seedance 2.0 Mini or Fast for cheap drafts, and MiniMax H3 when you want a bigger transformation and can accept looser timing. The output keeps your clip's shape and restores the selected section's audio.

Generate, wait, download

Press Generate. If you are not signed in, a sign-in window opens and your clip, images and instruction are kept. The exact credit cost is charged once when the job starts. Generation usually takes a few minutes; you can leave the page, because the task is saved to your account. When it finishes, play it in the result panel and download the MP4.

Troubleshooting

ProblemWhat to do
Rejected material

The provider declined an image or the cut; credits are returned. Use a different photo or a section without the flagged content.

Failed generationCredits are returned automatically; try again or switch model.
It is taking long

Leave the page open or come back later; the task keeps running and stays in your history.

The wrong person changed

Name the person in the instruction ("the man in the blue jacket"), or use a built-in clip where Image 1 maps to a lead.

The face drifts on turns

Add a full-body photo and a 3/4 view, use a shorter window, and use Seedance 2.5.

最近更新: 2026-10-06

Genjutsu AI tutorial questions

How long can a Genjutsu AI clip be?

The edited window is 5, 10, 15 or 30 seconds. Seedance 2.5 supports all four; Seedance 2.0, Seedance 2.0 Fast, Seedance 2.0 Mini and MiniMax H3 go up to 15 seconds. Uploaded source videos can be 5 seconds to 5 minutes long, and you choose where the window starts.

Do I need to install anything?

No. The editor runs in a web browser, including on a phone. There is no app, APK, plugin or ComfyUI workflow to set up.

How many photos should I upload?

One clear full-body photo per person is the minimum. Adding a 3/4 view or a second angle helps the face and hair stay consistent through turns. Up to nine images are accepted; on built-in clips images beyond the number of leads are used for detail.

Can I use a vertical phone video?

Yes. The output keeps your clip’s shape, snapped to 16:9, 4:3, 1:1, 3:4 or 9:16, so a vertical phone clip at 1080p comes back as 1080×1920. Rotated phone recordings are read at their displayed size.

How long does generation take?

Usually a few minutes; longer windows and higher resolutions take longer. The task is saved to your account, so you can leave the page and come back to download it.

Why did my generation fail or get rejected?

Before Seedance 2.5 generation, images of people and the selected cut go through the provider’s asset review; rejected material stops the job. Provider errors can also fail a job. In both cases the credits are returned automatically.

How do I download the video?

When the job finishes, download the MP4 from the result panel or your history. Accounts without a paid order get a moving watermark; after any purchase the same video downloads clean.

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