Sign In

How to use potatcats MOSS inpainting workflow

1

How to use potatcats MOSS inpainting workflow

How to use potatcats MOSS inpainting workflow

kuva.png

I recommend setting browsing level to PG-13, for some reason comfyui nodes are too sexy and get that rating automatically. I dispute them, but that can take days.

For cleaner look i recommend disabling links from bottom right corner on the comfyui

kuva.png

Fast showcase

1. Load image

  1. Drag and drop image to Load image or choose file.

  2. Draw mask from right click -> open in mask editor (Press clear mask before creating new mask, mask editor might leave some residue if you use erase tool)

kuva.png

2. Choose mode

Choose either Cut gen or Full gen

  • Cut gen cuts area around the mask for generation

  • Full gen throws the full image to the generation

kuva.png

3. Adjust settings

Load correct models

kuva.png

Positive prompt

kuva.png

Negative prompt

kuva.png

Mask settings

kuva.png
  • Pixel size: Size of the latent for the generation. Will always be square. 1024 will be 1024x1024.

  • Expand mask: How much the mask is expanded beyond the masking.

  • Padding: How much image around the mask is included in the generation. Default it will add 32 pixels around the mask. If the inpainted pixel area is low, raising this can help with inconsistency or too sharp results.

  • Mask blur radius: Radius of blur around the mask. Helps with consistency of image and possible visible "lines" of inpainting.

  • Mask opacity: 1 sends the mask with 100% opacity and 0 sends it with 0% opacity. Better than denoise when you want subtle changes or just bump up the quality in the image and want to keep the composition. When using this i recommend using 1 denoise.

  • Color match: Tries to match the generated image color to original. If you generate with same model as you made the image with, this is not usually necessary.

  • Fill mask holes: Makes masking easier. Enables encircling area to mask it.

    kuva.png

Sampler settings

Sampler settings are set from the sampler.

  • With inpainting you want to adjust denoise depending on how much change you want in the image. Rest of the settings are preference at this point.

kuva.png

Bounding boxes

With bounding boxes you can specify area where you want the prompt in. In this example i am adding top hat to the cat and making it smile.

  • Tip: When you load new image, you can click the node and press the blue "play" button to get the new image as the background. This makes it easier to adjust the bounding boxes to right place. The image will show cut or full image depending on which mode is selected.

It is not necessary to use bounding boxes. If you only wish to use positive prompt, you can do that.

kuva.png

Rest of the workflow

The 100 Previews

  • At the top you can see before and after review for the generated area.

    kuva.png
  • On top of bounding boxes you can see review for the bounding boxes on the image

    kuva.png
  • Next to the load image there are previews for Cut mask, Original image and inpainted image.

    • Cut mask shows all the settings applied. Blur, expand mask etc..

      kuva.png

Model patches

At left side of the workflow there is are for all the nodes that require model patching. These are things you can remove or add for the generation.

kuva.png

The monster in the room

Some day i might do something about this. (~ ̄▽ ̄)~

kuva.png

Example

Just a simple example of adding tophat to cat and making it smile

kuva.png

Added masks for area i wanted to edit

kuva.png

Described mostly only the style in the positive prompt.

kuva.png

Added bounding boxes for the areas i wanted to edit.

kuva.png

Adjusted denoise to fitting 0.75 on sampler

kuva.png

Generated the image

ComfyUI_00197.jpg

1

Comments