trained on 4 headshots of y0 kiryu at 512x512 resolution (because i trained a sd1.5 dreambooth kiryu model with the same 4 input images back then and i wanted to test how sd3.5 large would do with only 4 512x512 headshots) with y0kiryu as the only caption and triggerword it barely looks like him bruh and he has unwanted necklaces earrings and ties even though none of these are present in the dataset + his eyes look melted and why is his hair going up