Notes for Illustrious:

Much more compatible. Highly recommend using uncensored in the prompt, because the tagging system struggled to understand his style of censoring.

Useful information:

No activation word is necessary. This model works best at higher strengths (due to a lack of good noise free images in the dataset). I would suggest not going below 0.8 and even seems to work decently well, modestly above strengths of 1.

Not particularly adept at multiple girls.

Background:

This is the second attempt, but first published lora I've made based off of Oosawara Sadao's art. This artist has a special talent for curvy (non-fat/chubby) all the way to petite women. They have a very distinctive way of detailing hair, too. This artist often has very expressive characters. They also have draw chibis, but I haven't made the effort to really sub categorize the artstyle.

As always, I do not inpaint generated images that I create with Loras and try to keep the prompt as simple as possible.

This artist is very talented, but the majority of their works are in doujins. As such, I needed to do a lot of editing to try and remove speech bubbles and text. I didn't want to take ages to do it the more better way, so it introduced noise to the full data sample. These appear as smudges. This artist has some serious health conditions (related to their eyesight) that have reduced this output over time.

I may take time in the future to refine this lora.