Before I go in and tackle the other Kamen Riders, I'm going to experiment with a larger dataset, since we're gonna be dealing with a ton more Kamen Riders.Enter the Hitotsuki from Avataro Sentai Donbrothers. The dataset for the Hitotsuki are 50 images trained on Stable Diffusion 1.5, and they're perfect for this experiment because they all have the same lower body (barring a purple or red variation) with their top being different. I also included their giant versions, the Hitotsukings (also trained on SD 1.5 but with 27 images instead, omitting Resshaki and Kyukyuki due to them being CGI and potentially throwing off the whole dataset), as they have even more of a uniform fit to them.In the end, I conclude that the Hitotsuki LoRA has no overcook, since it was able to produce a human without its face melting off (same holds true for Hitotsukings, but I wanted to give Monaking a monster face so I upped the weight for Hitotsuking). This proves that while the slow simmer method is still perfect for larger items, I'm a little nervous for the cook times, so the larger steps may need to necessitate a batch of 3 or 4, or I need to reduce the epochs to 15 for the sake of brevity and reducing the risk of overcook.Also, the activation tags aren't programmed into the LoRA, but they seem to be what cause them to activate. I found you can use either one of them or both of them.

Files with SHA256: 76fbd30889225e89720691b46d095e56b26dab392e735afd5e8bbb15719d57fa