사용자 LoRA
XL
HCQ Style,
This model was trained on a dataset that leaned towards western style cartoons with a little bit of anime mixed in too. The Anime seems to be a bit more powerful than expected but I'm still getting enjoyable results from it. I was rushing out the door when I ran this model and didn't realize the trigger was "HCQ style" instead of "HQC style" but for version 1 it will do. Hopefully it doesn't really matter.I highly recommend using quality tags when prompting in illustrious, I think "masterpiece(quality)" in the positive and "bad quality" in the negative helps. This model is meant for fun portraits of science fiction and fantasy but should also do okay with many modern settings too. It does work with base Illustrious model but I've found better results can be achieved with a few other checkpoints like, Hassaku.This model runs best between 0.5-1 strength depending on the desired outcome and checkpoint you are using. I do notice some significant changes between 0.6 and 1.0 strength, even noticeable color changes at times. All of the training data was SFW but it should still be capable of making NSFW content. This is my 2nd attempt at making an Illustrious Lora and I think I will continue as its giving interesting outputs and seems to have a lot of room to grow and become even better.