How I Motion Swap Any Video with Higgsfield's Genjutsu Tool

Motion swapping used to mean hours in an editor and a result that still looked slightly off. Higgsfield's Genjutsu tool does it with two reference photos and a prompt, and it keeps the original motion, lighting, and background completely intact. Here's exactly how I set it up, step by step, including the prompt that actually got it working.

What Genjutsu Actually Does

Genjutsu is a motion transfer tool inside Higgsfield. You start with a base video, either one from the community library or your own upload, then hand it two reference images of the people you want in the scene. Genjutsu keeps everything else the same, the set, the camera work, and the original movement, and just swaps your characters into that motion. No rigging, no manual masking, just a prompt.

Step by Step: Motion Swapping with Genjutsu

Here's the exact workflow I used to swap two characters into a community video.

  1. Head to Higgsfield and open the Genjutsu tab in the top navigation.
    Higgsfield top navigation with the Genjutsu tab circled
  2. Click Community to browse existing motion clips, or upload your own four to thirty second video under Motion Transfer.
    Higgsfield Genjutsu Community tab circled showing sample motion transfer clips
  3. Pick a clip and hit Recreate to load it into the editor.
    Genjutsu video preview with the Recreate button circled and the default prompt shown
  4. Under "Add your characters, products, or clothes," upload the two reference images of the people you want swapped in.
  5. Select both images from your uploads and confirm them, you'll see a checkmark appear on each.
    Uploads panel in Genjutsu with two character reference images selected and checked
  6. Write your prompt. Genjutsu drops in a default line, but I got the cleanest result with this: "keep everything the same in the video, just replace the main subjects with the two characters provided in @Image 1 and @Image 2." Tagging your images directly with @ instead of just describing them as "the 2 images provided" made the swap noticeably more accurate.
    Genjutsu prompt field showing the refined prompt tagging @Image 1 and @Image 2
  7. Set your output quality, 480p, 720p, or 1080p, and hit Generate.
    Genjutsu quality dropdown showing 480p, 720p, and 1080p options above the Generate button

That's the whole process. The part worth paying attention to is the prompt, being specific about "main subjects" rather than just "faces," and tagging each reference image by name, is what got Genjutsu to lock onto the right people instead of guessing.

Tools Used In This Guide

  • Higgsfield AI, home to Genjutsu and the rest of their video and image tools.

Want My Exact Prompts?

Every Claude skill and prompt I use is inside my Vault. Join The Viral AI Vault for step by step tutorials, prompts, and a community built around going viral with AI content.

Follow Along

New comparisons and breakdowns drop regularly, follow for more:

Some of the links above are affiliate links, meaning I may earn a small commission if you use them, at no extra cost to you. It just helps support the channel. Thanks for the support!

Back to blog