This is an experimental lora that I try to train a character with many cosplaying images by different people on Flux1 D.
I found many horrific images while I was finding the training dataset, they are removed from the dataset.
These are removed to prevent the result being too scary. Usable images are not many.
And I found out this can create anime images after I use the anime prompt.
Description
FAQ
Comments (3)
Thank you for removing the monsters. It will help get a cleaner more precise image. rather than allowing The Latent Space to generate bizarre Con larpers that will definitely ruin an image generation.
Actually, the base model of Flux 1D is able to generate photos of safe for work " thin Nordic teen," but this site bans all safe for work teen photos. And the people who post those horrific photos of themselves, usually have some friends who can help them harass the critics. The whole cosplay community is pretty terrible. The number of images I removed is more than those usable images.
Kind of fucked up to post images of real people and call them 'horrific', especially the second one whose sole "issue" is that the cosplayer is black