First style LoRA in the books! Still don't know what I'm doing. Maybe it'll suck, maybe it won't. Try it out!<h3 id="what-is-it">What is it?</h3>Just a mix of styles I've adopted over the last month or so. In theory, the style should be a blend of anime, western cartoon, comicbook style and a smidge of old-school Disney animation style.<h3 id="how-does-it-work">How does it work?</h3>Honestly...I have no clue. Civitai, in their infinite wisdom (/s), decided to put private model ability behind a paywall, so until I've worked with it, I can't say for sure. But I can make some assumptions:1) The training images were all made using a DMD2 LoRA with low CFG (1-1.5) and low steps (10 or so), so in theory that would be a good starting point that you can adjust to your taste from there. (I was wrong, images are trash at those settings, 20 steps and CFG 4+ yields better results).2) Almost all training images were generated in PNG format, so you might want to drop the extra 2 buzz on that. 3) Try it with and without quality tags. Some training images used them, some didn't. The ones without quality tags are more of a simplistic style, but with quality tags and the <code>detailed</code> tag in the prompt, you should get some really detailed textures and backgrounds.4) Be descriptive...or don't. Simple prompt = simple style with abstract backgrounds. The more detailed the prompt, the more detailed the image. Check it out, drop some buzz if you like it, leave comments on what you like, what you don't like, tips on how I could improve in the future.

Files with SHA256: 010e732a2e7889bdb7999cf4021a545270046c7d2d125119122a1bb09561cdb9