Replies: 1 comment
-
|
Adetailer does not care where the image comes from since it uses image recognition to detect the target area. But it still needs a prompt to work with and a model that can actually replicate the style of what you see. So don't use an anime model for realistic images. I know there are some websites that let you do this, but they can also be very hit and miss, and often charge money. However, I would not recommend using img2img; you can get lucky, but most of the time, it looks like crap. From my experiance adetailer alone is just not enough. I have gotten the best results from using inpainting and adetailer together with the correct models, and inpainting actually does most of the work. But that obviously means more effort. you actualy have to write the prompt for what should be there. It also depends on the image and the censor type. Each censor type has its own issues to deal with, and it usually still takes a bit of work specialy if you don't want it to look bad. I could post an example, but all the images I have are adult content, and I'm not sure how GitHub would react to that; my guess is badly. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
The idea seems to be the same, autodetect a body area, process it and replace in the original. Although I'm not sure if Adetailer can work on random images instead of in-pipeline currently generated picture. But then it does have the "skip img2img" option so there's something.
Did anyone perhaps already adapt it for this task or there are better suited tools?
Beta Was this translation helpful? Give feedback.
All reactions