Skip to main content
Higgsfield Genjutsu Motion Transfer rebuilds an existing video instead of generating a new one. The source clip keeps its motion, camera movement, timing, and audio, while the cast, location, wardrobe, and style are rebuilt from 1 to 8 reference images. You describe what should be different, and the rest of the shot stays as it was filmed. Higgsfield ships Genjutsu in two modes, Motion Transfer and Object Swap. ComfyUI exposes Motion Transfer through the HiggsfieldMotionTransferNode in the partner/video/Higgsfield category. It is a cloud API node: it needs a Comfy account with credits, downloads no model, and bills per output second, because the length of the result follows the source clip. The node is useful when the performance already exists and only the look has to change. Recast a phone clip with a different character, move a shoot to a location that was never available, restyle footage for another market, or produce several variants of one take without filming each one separately.
To use the Partner Nodes, you need to ensure that you are logged in properly and using a permitted network environment. Please refer to the Partner Nodes Overview section of the documentation to understand the specific requirements for using the Partner Nodes.
Make sure your ComfyUI is updated.Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.If nodes are missing when loading a workflow, possible reasons:
  1. You are not using the latest ComfyUI version (Nightly version)
  2. Some nodes failed to import at startup

​
Higgsfield Genjutsu: Motion Transfer

Recast a video from reference images while the source clip keeps its motion, camera movement, timing, and audio.

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “Higgsfield Genjutsu: Motion Transfer” in Template Library
Input materials Upload these files to the matching LoadVideo and LoadImage nodes:

rooftop_dancer.mp4

LoadVideo · source motion clip

snow_queen_ref.png

LoadImage · character reference · image_1

ice_palace_ref.png

LoadImage · scene reference · image_2

​
Node controls

The output length follows the prepared source clip, so a 12-second clip returns a 12-second video.

​
Prompting tips

  • Name what each reference replaces. The prompt is optional, but it is the only place to say which image belongs to which person or object. Write it as a short instruction, such as recasting the dancer as the character in the first reference image and using the second one for the new setting.
  • Keep the reference set narrow. One image per element is usually enough: a character, a location, a wardrobe change, or a style. Adding more images gives the model more to blend, which makes the result less predictable.
  • Match the framing you want. The output keeps the motion and camera of the source clip, so choose a source video whose framing already suits the result. A close-up performance stays a close-up.
  • Describe the change, not the whole scene. Everything not mentioned in the prompt is rebuilt from the references, so there is no need to restate the motion, the pacing, or the audio.
  • Expect variation between runs. The seed does not lock the result; the same settings can produce a different take.

​
Get started

  1. Update ComfyUI to version 0.40.0 or later
  2. Go to Template Library, search for Higgsfield Genjutsu
  3. Load the source clip in the LoadVideo node and the reference images in the LoadImage nodes
  4. Write the prompt, pick a resolution, then run the workflow