How to Inpaint
Has anyone figured out how to use the inpainting method here? I tried with apple's stable-diffusion package passing in a mask but it just seems to place the mask itself into the final image
I'm in the same place as you. I tried passing it a mask (white, black, green, transparent) along with an image using the i2i argument, tried that reversed, and tried with a masked image (white, black, green, transparent masked area), and it never seemed to understand the mask. I think it needs a second input and some dedicated code added in Apple's package. I posted for help/info over there (https://github.com/apple/ml-stable-diffusion/issues/181) a few weeks ago with no response yet. Maybe you can do a "me too" there to bump that post to the top of a "recent" list, but I'm assuming the silence means t is just not supported. FWIW, the Python pipeline, if I am remembering right, uses a second input for the mask that it reads in as L,r,b,g. So it probably converts it to just the L channel and adds that into the r,g,b of the input image, and then goes from there.
Did a me too bump ☺️