Overview

I recently published a couple of nodes that automate and significantly improve inpainting by enabling the sampling to take place only on the masked area.

The main advantages these nodes offer are:

• They make it much faster to inpaint than when sampling the whole image.

• They enable setting the right amount of context from the image for the prompt to be more accurately represented in the generated picture.

• They enable upscaling before sampling in order to generate more detail, then stitching back in the original picture.

• They enable downscaling before sampling if the area is too large, in order to avoid artifacts such as double heads or double bodies.

• They enable forcing a specific resolution (e.g. 1024x1024 for SDXL models).

Tutorial

Download

The nodes can be downloaded using ComfyUI-Manager, just look for "Inpaint-CropAndStitch".

You can also get them, together with several example workflows that work out of the box from https://github.com/lquesada/ComfyUI-Inpaint-CropAndStitch