Acly/comfyui-inpaint-nodes

What it solves

This project provides a set of specialized nodes for ComfyUI that improve the quality and control of image inpainting and outpainting. It addresses common issues like visible seams, color shifts, and the inability to blend existing content with SDXL inpaint models.

How it works

The toolkit expands ComfyUI's capabilities through several functional groups:

• Model Integration: It allows the use of Fooocus inpaint models to transform standard SDXL checkpoints into flexible inpaint models. It also integrates fast, specialized inpainting models like LaMa and MAT for object removal and outpainting.
• Pre-processing: Before the main diffusion process, it can fill masked areas using various methods (neutral grey, Telea, or Navier-Stokes algorithms) or blur the image into the mask to maintain color consistency.
• Mask Manipulation: It includes tools to expand, shrink, and stabilize binary masks to ensure smoother transitions and avoid numerical errors.
• Post-processing: It features a color-matching node to fix brightness or color shifts and a tool to convert denoise masks into compositing masks for better alpha blending.

Who it’s for

Digital artists and AI image generation users who use ComfyUI and want professional-grade control over filling, expanding, or editing specific parts of an image.

Highlights

• Fooocus SDXL Support: Seamlessly converts SDXL checkpoints into inpaint models.
• Advanced Fill Modes: Includes Telea and Navier-Stokes algorithms for high-quality pre-filling.
• Model-based Pre-filling: Integration of LaMa and MAT for fast, non-diffusion based filling.
• Efficient Conditioning: A dedicated node that combines VAE encoding and inpaint conditioning to reduce computational overhead.