Skip to content

Diffree

Add new objects to an image using only text

Visit official site

What Diffree is

Diffree is described as a text-guided image inpainting model that adds new objects to an image from a text description while keeping background consistency and object quality. The model is trained on the OABench dataset with Stable Diffusion and an extra mask prediction module.

What you can do with it

  • Add new objects to an image from a text description
  • Keep the background consistent while adding objects
  • Add several objects to the same image iteratively
  • Add objects across a range of natural scenes

Application scenarios

  • Add birds to a landscape photo to make it livelier
  • Place a virtual product in a promotional image for market testing
  • Add missing elements to a historical scene from a description

Getting started

  • Visit the online demo page for Diffree
  • Read the usage instructions and requirements
  • Enter the text description of the object to add
  • Upload an original image or pick an existing sample

Pros & cons

✓ What we like

  • Object addition needs only a text description
  • Background consistency is kept while adding objects

! What to watch out for

  • Adding an object starts from an uploaded original image or a chosen sample
  • The mask prediction module decides the position of the new object

FAQ

How are new objects placed?

The material states a mask prediction module predicts the position of the new object.

What data is the model trained on?

The model is trained on the OABench dataset generated with image inpainting techniques.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review