Overview

"✂️ Inpaint Crop" is a node that crops an image before sampling. The context area can be specified via the mask, expand pixels and expand factor or via a separate (optional) mask.

"✂️ Inpaint Stitch" is a node that stitches the inpainted image back into the original image without altering unmasked areas.

Parameters

context_expand_pixels: how much to grow the context area (i.e. the area for the sampling) around the original mask, in pixels. This provides more context for the sampling.
context_expand_factor: how much to grow the context area (i.e. the area for the sampling) around the original mask, as a factor, e.g. 1.1 is grow 10% of the size of the mask.
invert_mask: Whether to fully invert the mask, that is, only keep what was marked, instead of removing what was marked.
grow_mask_pixels: How many pixels to grow the mask to provide a bit of a border, to be used in combination with blur_radius_pixels.
fill_holes: Whether to fully fill any holes (small or large) in the mask, that is, mark fully enclosed areas as part of the mask.
blur_radius_pixels: Whether to blur the mask, to be used in combination with grow_mask_pixels. Some models prefer blurred masks, some don't.
adjust_to_preferred_sizes: This will try to have width and/or height of the context area match any of the sizes in preferred_sizes by growing the context area (not upscaling), e.g. 512, 1024. Some models prefer this.
preferred_sizes: Comma-separated list of preferred sizes, e.g. "512,1024". Default is 1024 because it fits most of the use cases of this feature.
prefer_square_size: This will try to have width=height (if possible). Some models prefer this.
internal_upscale_factor: Upscale the image and mask between the crop and stitch phases. This means the sampling happens only around the context area but at a higher resolution (e.g. 2 for x2), then it is downsampled and merged with the original image. This in practice gets more details from models. If you want to obtain a higher resolution image, please upscale it before cropping/sampling/stitching. This can also be lower than 1 for the rare cases you need it, such as the original image too large and the model generating double heads.
padding: Pad the cropped image and mask to a certain number of pixels in order to avoid artifacts. 8, 16, 32 are reasonable values.

Simple example

This example inpaints by sampling on a small section of the larger image. It runs ~20x faster than sampling on the whole image.

Download the following example workflow from here or drag and drop the screenshot into ComfyUI.

Advanced example

This example inpaints by sampling on a small section of the larger image, but expands the context using a second (optional) context mask. It runs ~10x faster than sampling on the whole image but allows navigating the tradeoff between context and efficiency.

Download the following example workflow from here or drag and drop the screenshot into ComfyUI.

Installation Instructions

Install via ComfyUI-Manager or go to the custom_nodes/ directory and run $ git clone https://github.com/lquesada/ComfyUI-Inpaint-CropAndStitch.git

Changelog

Upcoming!

Make adjust_to_preferred_sizes scale the section rather than grow the context area to fit preferred_sizes, to be used to e.g. force 1024x1024 for inpainting.

2024-05-13

Enable internal_upscale_factor to be lower than 1 (that is, downscale), which can be used to avoid the double head issue in some models.
Added padding on the croppedp image to avoid artifacts when the cropped image is not multiple of (default) 32

2024-05-12

Add internal_upscale_factor to upscale the image before sampling and then downsizes to stitch it back.

2024-05-11

Initial commit.

Acknowledgements

This repository uses some code from comfy_extras (https://github.com/comfyanonymous/ComfyUI) and from KJNodes (https://github.com/kijai/ComfyUI-KJNodes), both licensed under GNU GENERAL PUBLIC LICENSE Version 3.

License

GNU GENERAL PUBLIC LICENSE Version 3, see LICENSE

4.4 KiB Raw Blame History