It's better to avoid using images generated by AI and filtered images for training. Just use no-filter selfies with natural light and Flux will do the magic for generating good images with high realism.

Hey, I just wanted to let you know I posted one more picture for you. I didn't have time to finish it for your bounty, but wanted you to have it anyway. It's an image made with all four of your LoRAs at the strengths you asked for. I call it "Gypsy." I hope you like it, and thanks again for a fun challenge.