Skip to content

MagicFixup

An open image editing model that learns from video

Visit official site

What MagicFixup is

The model is an open image editing tool from Adobe Research that learns edits from video.

What you can do with it

  • Repair defects in a photo
  • Prepare source and edited images
  • Train on your own video data

Getting started

  • Create the conda environment
  • Download the pretrained model
  • Run the inference script on both images

Pros & cons

✓ What we like

  • Custom training on user video data
  • An environment file eases installation

! What to watch out for

  • A conda environment must be created
  • Pretrained weights must be downloaded

FAQ

What is it trained on?

Stable Diffusion 1.4 is named.

Is there a demo?

A gradio demo is provided.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review