comfyanonymous / ComfyUI

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
https://www.comfy.org/
GNU General Public License v3.0
58.56k stars 6.21k forks source link

(Feature Request) Soft Inpainting #2840

Open CognitiveDiffusion opened 9 months ago

CognitiveDiffusion commented 9 months ago

I may be a bit too much to ask on the main repo, but hopefully it'll at least get exposure this way.

Basically it's a new A1 feature that seams (heh) to be quite amazing, utilizing actually Photoshop-esque instead of binary masks for inpainting, getting rid of most inpainting problems.

https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14208

LuLmaster69 commented 9 months ago

I too have tried to ask for this feature, but on a custom node repo https://github.com/Acly/comfyui-inpaint-nodes/issues/12 There are even some details that the other posters have uncovered while looking into how it was done in Automatic1111. Hopefully it would be of help to anyone who may be interested in implementing it in ComfyUI.

BenDes21 commented 2 months ago

still no node with settings like Schedule bias, Preservation strength, Transition contrast boost etc?