This lora was trained on 25-30+ images from what I believe know to be Himukai Yuuji's artwork. These are images I fount while searching for resources to train my Etrian Odyssey lora. I really liked the style so I attempted to recreate it with the small available resources. The good news is, I captured the style. The bad news is, (the pony version) is notoriously bad with hands and feet. Nonetheless, I like it.
The Pony version was trained on Autism Pony. The illustrious version was trained on Zootlustrious V2. The Noob version was trained on Noob 1.0 V-Pred. The Anima version was trained on, well, Anima. I haven't tested it with anything else.
To get the results you see in the images above, these are the settings I used.
Description
Trained on 25 images at 1792 resolution. I went back and re-selected and edited the images to make sure I captured the style I originally intended. After thorough testing and retraining a few times, I am now 100% confident you can faithfully recreate the intended style with this version of the lora. Funny enough, I couldn't find half of the images I used in the old dataset as I was searching online. (I wonder why...? lol) Luckily, I still have all of the original data, with the most important ones being in good quality.
You might get some wrong hands and feet every once in a while. Just add 5 fingers or 5 toes to the prompt and it should fix itself.