Open ARandomUserFromGithub opened 1 year ago
I'm not positive, but I think it's similar to diffusion. Say you got little fish floating around a person with the absence of water, they may come out as birds, so you could negative weight "flying birds" in this particular image to ensure the "fish" aspect remains true.
Tho can it read negative prompt from the positive filewords? Not sure how to train negative during positive training.
I'm not too sure how to get the negative prompt through besides txt2img which only generates previews. I tried to follow the same pattern than with generation parameters txt fiels coming along generated images:
I'm not sure if the negative prompt is getting through or not.
As there's a checkbox saying "positive filewords only", I assume there's such a thing as negative filewords. So I try to tinker around but there's not any information online about that subject.