Have you ever tried to use inpainting to remove an object from an image, only to have the AI generate something completely different from what you wanted? You’re not alone.
I’ve seen this problem pop up time and time again. Let me give you an example. Imagine you have an image of a character wearing white bunny ears, a bunny tail, some cufflinks, and a black latex outfit. You want to remove the black latex outfit, so you cover it up with the inpainting tool and set the strength to full. You double-check to make sure you haven’t missed a single pixel.
But when you generate the new image, the AI decides to create its own black outfit in that space, no matter what you put in the prompt or negative prompt. It’s as if the AI has a mind of its own and is determined to do what it thinks is right, rather than what you want.
## Why Does This Happen?
The reason this happens is that the AI is using contextual clues from the rest of the image to decide what should be in that space. In this case, it sees the bunny ears and cufflinks and decides that a black outfit is the most logical choice.
## The Frustrating Part
The frustrating part is that the AI is ignoring your explicit instructions. You’re telling it to generate a bare or unclothed character, but it’s deciding to do the opposite. It’s like the AI is saying, ‘I know better than you, human. This is what should be here.’
## A Possible Solution
One way to get around this problem is to mark the other clothing items in the image. This seems to help the AI understand that you want to remove the entire outfit, rather than just replacing one piece with another.
## The Bigger Question
But this raises a bigger question: can we ever get AI to truly listen to our instructions? Or will it always be trying to outsmart us and do what it thinks is best?
—