art style model from akisora.
After brief testing, the results with a weight of 1 looked decent.
The tag used is aki-sora.
I categorized parts of the training set and attempted to derive additional tags for aki-lesser and aki-greater, though this wasn’t done thoroughly due to time constraints. Feel free to give it a try.
Suggested negative tags: comic, mosaic censoring, censor, censored.
-----
aki-sora, 0.5.
n: mosaic censoring, censor, censored,
I will improve this model later. It has significant issues with parameter settings and training set processing……