This LoRA model was created as a bit of a joke. I also wanted to see how far I could push the AI before it broke; if it could.

I created her using myself at a low denoise and prompting for certain features. Then using FaceApp, I made use of face transfer, blend, and reshaping tools to make her eyes as large as possible. The first training resulted in eyes within eyes, so I had to tag my image set for "glasses". After that, she worked fairly well, but the eyes weren't great. I then took a photograph of my eyes and used Affinity Designer to add them into the pictures and work out some details that weren't great.

I trained her again and she seems to be fairly consistent!

Flux:

Flux generations were in ComfyUI with a strength of 1!