Skip to content

HOI-Swap

Swap video objects with hand interaction in mind

Visit official site

What HOI-Swap is

The framework is described as a diffusion-based video editing approach that handles hand and object interaction, starting with an object swap in one frame and then extending it to a full sequence.

What you can do with it

  • Swap an object using a supplied reference image
  • Adjust hand interaction to the changed shape and function
  • Learn from self-generated training data

Application scenarios

  • A video maker replaces an object for a more realistic scene
  • A post-production team tunes hand motion after a swap
  • A VR developer builds more natural hand and object interaction

Pros & cons

✓ What we like

  • Grip patterns adapt to object shape and function
  • A second stage rebuilds the full video sequence

! What to watch out for

  • Editing starts from a single frame
  • A reference image of the object is required

FAQ

How many stages are there?

The card describes a single-frame stage and a stage that extends the edit to the video sequence.

Does it need labels?

The material states training is self-supervised with no external annotation data.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review