Usage Guide
TODO: write this guide
For now, watch the (Outdated) Full Usage Demo:
If coming from AUTOMATIC1111's webUI, you may be used to:
- Using white to color the inpainting mask.
- Using full resolution inpainting and mask blur.
However, auto-sd-paint-ext's inpainting is more sophisticated:
- The alpha/transparency channel is used for the inpainting mask (any color can be used).
- Use the selection area to select what region the model sees.
- The actual blur in the inpainting mask will be used as the mask blur.
NOTE: You have to create a paint layer to draw the inpainting mask & ensure the inpaint mask layer is selected.
Quick Jump:
In AUTOMATIC1111's webUI, usually the whole image is given to the Stable Diffusion model when inpainting an area. Turning on Full Resolution Inpainting causes the webUI to crop the image to only the region being inpainted plus some padding. This allows the model to inpaint at full resolution, before pasting the generated inpaint back into the original image.
In Krita, cropping the image given to the model is more intuitive, you just need to select the region you want the model to see.
This does the exact same thing as Full Resolution Inpainting, but with much more control over what the model sees.
In AUTOMATIC1111's webUI, mask blur applies Gaussian blur to your inpaint mask before inpainting.
In Krita, you can also apply a Gaussian blur to your inpaint mask:
Or you may use a softer brush that has blur/fade effect (for example, airbrush) when drawing the inpaint mask:
