Do NOT expect this to be an Anime Style LoRA due to it being from made from doodles and sketches and some incomplete stuff and some bad stuff. It's specialty is it being goofy or assisting checkpoints towards a style like in the thumbnail with Ionsyx V5.8.

So some models might make the images bad looking, sketchy, or goofy. Also catball_/(object/) should be a tag in the dataset, I think.
Note: Realized the tag is actually catball_\(object\) on 12/15/2025 2:40 AM EST/EDT

If the model is mostly illustrious or noobai it should do bad sketch/doodle stuff if it's a merge it shouldn't and should skid all over the place like I said in the above: "So some models might make the images bad looking, sketchy, or goofy." So I cannot guarantee in any compacity this will good all the time.

If you are putting this into a merge checkpoint make sure to not use the trigger word, as I do this myself and strengths between 0.01 to 0.50 unless the model requires a 0.50 to 1.00 which is rare. Also it might turn out overbaked if the LoRA is already inside said model.

BTW Dataset has no people except basically a stick figure doodle lookin' dude, and a oval shaped alien cyclops thing lol.

The exact settings to train V1.5: