search

comfyanonymous / ComfyUI

The most powerful and modular stable diffusion GUI, api and backend with a graph/nodes interface.
GNU General Public License v3.0
41.12k stars 4.38k forks source link

Conditional masking does not affect image. #2567

Open NeoTech opened 5 months ago

NeoTech commented 5 months ago

After experimented quite a bit with the masking, and using Conditioning SetMask - im finding that mask doesnt seem to affect the tensor output regardless.. The combined text clips can achieve the same output from a straight up description.

If the clips are concated it will even produce identical images, but if one uses combine one get "slight" difference with the setmask node but more of an alteration rather than actually "control" over what gets genrated in the mask.

Test workflow attached.

workflow

ttulttul commented 5 months ago

I see similar results. I suspect that the Conditioning (Set Mask) node is not doing what we think it is doing.

Pos13 commented 3 months ago

@NeoTech I edit your workflow in a way I expect it to work, and got it first try:

CondCombine

Pos13 commented 3 months ago

Btw, max strength is 10.