Skip to content

Emu Edit

One multi-task model for precise image editing

Visit official site

What Emu Edit is

The model is described as a multi-task image editing model that performs precise edits through recognition and generation tasks, with an architecture optimised for multi-task learning.

What you can do with it

  • Edit a chosen region of an image
  • Apply free-form edits
  • Run detection and segmentation tasks
  • Work on super-resolution, contour detection and background modification

Model notes

  • The architecture is trained on many tasks including region-based editing, free-form editing, detection and segmentation
  • Learned task embeddings guide the generation process to follow edit instructions
  • Task inversion adapts to unseen tasks with few samples while model weights stay unchanged

Reference cases

  • Repair photo defects through precise region editing
  • Change the style of an image for a personalised result
  • Use segmentation to separate different elements of an image

Pros & cons

✓ What we like

  • Several task types are handled by one model
  • Unseen tasks can be approached with few samples through task inversion

! What to watch out for

  • Output follows the listed editing and vision tasks
  • Adapting to unseen tasks relies on task inversion

FAQ

Which tasks are listed?

The feature list names region editing, free-form editing, detection and segmentation, super-resolution, contour detection and background modification.

How does it handle unseen tasks?

The material describes task inversion for few-sample learning, with model weights unchanged and only task embeddings updated.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review