ForserX/StableDiffusionUI

[User Question - Advice Needed] Inpainting seems not to work properly. How to use correctly.

ratunku opened this issue · 4 comments

Hello good people!

I was looking up in the wiki and was searchig the web, but could figure out a solution on my own. I hope you can help me Could somebody advice please, how to inpaint with XUI correctly. I tried with positive and negative masks but the changes are made to the whole image and never to the masked or unmasked parts of the image. I created the mask in paint.net: a png image with transparency/erased pixels with the eraser tool for the masked region. Then I uploaded the source image and dragged and dropped the mask, described the desired changes ("blue dress" - original is green) and clicked on "Make!". However, the changes seem to be always applied to the whole picture and never to the masked area. I have the feeling that I might be creating masks wrongly or there is something wrong with the models I am using, but this is only a guess and might be misleading. Any Ideas? As a model I used analogMadness_v50.

Thanks a lot in advance for your help and support!

Use black color for mask

wow, thanks a lot for the quick response.

Black was the first color I tried for this but it didn't work for me either. Just reproduced it with a fresh image and a new mask. Same PNG file, with black mask inpainted. However, still the whole image is being rendered from instead of just the masked area. Both PNG files have the same dimensions.

Are there any settings that may contribute to inpainting and masking? Is a specific inpainting model needed maybe? I tried multiple inpainting models, but they didn't convert from savetensor.

I also unzipped and started a fresh instance, but with no effect.

I am sorry for bugging you with this. Thanks a lot for your help and support!

I check it later