Isn't stable or very good yet. Posting for buzz as to train more with a better dataset.
DO NOT USE ADETAILER! IT WILL BREAK IT!
Prompt
glowing eyes, multiple eyes, wrinkles, disfigured
Neg Prompt
Robot, Bald
I'm thinking about just running the training locally, but I need to get more screenshots from cyberpunk 2077. I'm also thinking about splitting this into two versions, one from art of the characters and another from in game screenshots, which should allow very different styles.
I could keep them together which might allow more styles with the lora and more flexibility, but it could just turn out poorly. I'm still new to this, so any feedback is welcomed!