ali-vilab/MimicBrush

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1311 stars · 90 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MimicBrush is the official implementation of a zero-shot image editing method that lets users mask a region in a source image and provide a reference image showing the desired appearance, with the model automatically capturing semantic correspondence between them. It is built on a diffusion prior (two U-Nets over Stable Diffusion 1.5) and ships with inference code and a Gradio demo.

Use cases

• edit part of an image using a reference photo
• transfer texture from one image to another
• replace a masked region guided by an example image
• zero-shot image editing without text prompts
• run a local gradio demo for reference-based inpainting

When to choose

• you want reference-image-guided editing of masked regions without writing prompts
• you need texture transfer or post-processing refinement backed by a diffusion model
• you want to reproduce or build on the MimicBrush research paper

When to avoid

• you need simple text-prompt-only image editing
• you lack a GPU or cannot download large model checkpoints
• you need a production-ready, actively maintained product rather than research code

Facets

application · maturity active

image-processing machine-learning deep-learning artificial-intelligence image-processing computer-vision python diffusion-model stable-diffusion image-editing reference-imitation inpainting gradio-demo research-paper linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ali-vilab/MimicBrush")