EyesGen - Lora - WIP!
![1.png](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="450" height="450"><rect fill-opacity="0"/></svg>)
This is a tool that I train to change the style and eye colors of my generated characters. I use the Photoshop tool since inpainting gives most of the time a problem. (if someone can teach me how to train an inpainting model I would appreciate it)
<lora:EyesGen:0.4> works quite well between 0.1 and 0.5.
Prompt: <lora:EyesGen:0.4> Masterpiece, best quality, (red eye:1.3), 1girl, close up eyes, looking a viewer, intricate iris, 1 eye solo
Negative Props: bad-artist, bad_prompt_version2, easynegative, low res,
Discussion
![Le_Fourbe's Avatar](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="96" height="96"><rect fill-opacity="0"/></svg>){kind=avatar}
interesting : inpainting works for me WITHOUT inpainting model. give it a try but don't use "send to" int the img2img tab it's bugged sometimes, drag and drop is no issues.
alternatively : use checkpoint merger tool to get an inpainting version (didn't make one myself but still...)
set model A to the inpainting model of your choice
set model B to your own custom model
set model C as the base model that you used for training model B
set interpolation to add difference and multiplier M to model A to 1.
you have now an inpainting version,
to get a Lora variant try this simple standalone script : https://github.com/sashaok123/LoRA-EXTRACTOR
I like this model very much. Can the author make the sharingan and white eyes in the Naruto anime?
