I got curious if sdxl could replicate his unique and sexy artstyle, so I made a lora.

Trained on this https://artichoke3.newgrounds.com/art

V1: All images tagged with smilingwolf tagger and trained at a very slow rate

V2: Same dataset as V2, but each image had toon (style) and artichoke2 (dickpeacock11) appended to the tags. I also fixed a small portion of incorrectly tagged images.

The outputs are still very noisy though. Ideally, generate your base image with a model with very weak default style and hires fix it with your model of choice.

You probably NEED to use "toon \(style\), outline" to get close to the original artstyle. Even then, expect some seeds to produce weak results.