Another model made using Seaart Ai's lora training feature. This time of Yumi Yoshimura from Hi Hi Puffy AmiYumi.

I still got good results and though: Recommended weights are from 0.6-1. I suggest using words in your prompts that resemble Yumi's look in the show like purple hair, spiked collar, spiked bracelets, blue eyes, purple skull shirt and also Yumi Yoshimura. I think also using "hiyhi" might have some influence since I added that to the tagging part of the training process but I can't say for sure that had any real beneficial effects on the outcomes. Look at the images I've included for reference and experiment form there.