Jinx lora, best paired with a workflow with two reference images of the character. Strength 0.8-1.0, I used a GGUF Klein model with 5 steps and euler/simple sampler. Works with NSFW Loras.

Was training on 50 images with a mix of styles, anime/CGI, semi-realistic, and realistic. You need to add the preference in the prompt. Warning that some of the training images were NSFW.

She has default pink eyes but can be blue if added to the prompt.

My first LORA so please provide feedback and I can improve.