This lora was trained on 185 images from the artwork of Musouzuki. If you enjoy their work I encourage you to support them on their social media pages. This lora isn’t a perfect re-creation. It has problems with hands and feet. The backgrounds leave a lot to be desired. I believe this is mostly due to the dataset. Despite this, it mimics the “core concept” of the artist’s style well enough.
This lora was trained on Autism Pony. I haven’t tested it with anything else.
To get the results you see in the images above, these are the settings I used.
Prompt: score_9, [your prompt] <lora:MusouzukiXL:0.8> (‘1’ works as well)
Negatives: mosaic censoring,bar censor,censored,source_cartoon,source_pony,source_furry,
Sample: Euler a
Steps: 33 (Whatever gives you the best results)
CFG Scale: 7
Highres Fix: 4x-UltraSharp + ADetailer