This is trained on a dataset of 51 images from AIbooru. All images are low quality. It is intended to remove AI artifacts at negative values. This is an experimental model and has not been tested. recommended strengths [-2], [-0,8]. Only use illustrious models.
Description
Added version for Animagine. Unlike illustrious, works down to -0.3.
Comments (1)
Results?
It might need more diverse data and more data to properly pick up AI artifacts. Currently, it seem s a bit odd, and makes images noticeably worse on Animagine 3.1.
Best to just use something like '[[[worst quality AI-generated, bad hands ai-generated, bad anatomy AI-generated]]]` in your prompt, since animagine already has tagged AI data. Same can't be said for illustrious.