AFTER ADDITIONAL TESTING, I REALIZED THAT THE CFG SCALE MUST BE AT A LOW NUMBER. I HIGHLY RECCOMEND KEEPING IT AROUND 5 AND LOWER. AT LEAST BETWEEN 3.5 - 5 CFG. FOR STEPS IT DOESN'T MATTER MUCH. also lil edit as of 2/6/24: so uh, lil thing i should clarify... grimm's not going to exactly look the best in civitai.he's either going to look like an artifact or a vague appearance of himself. that i'm hoping to solve someway somehow, but if you're wanting to get some general grimm pics, then i definitely recommend using stable diffusion or comfy ui for this lora. if not, that's okay! if you don't mind the little goofs then that's perfectly fine!Be aware that CivitAI will not give you the same results as the showcases provided. CivitAI generates things differently from Stable Diffusion and ComfyUI. Put your effort in and work him to your own style and image!<hr />The trigger words for this lora; grimm, red eyes, featureless legs, black cloak.Additional non-trigger words to help with the prompting; mouthless, featureless arms (if you want him generated without fingers), hollow knight, slim body (and/or legs), red and black cloak, black hands, black cloak.<hr />Trained on 16 images, I'm honestly very, VERY happy to be publicizing my first EVER Lora both here and in general! I've always wanted to make Loras for a good few months now, but I never knew how to and I felt like I was NOT ready to do the patience-inducing methods of Lora training. So, instead, I relied on CivitAI's model training to help me instead. Of course I end up being a lot more limited, especially knowing how grand in capabilities you have if you did the manual labor way of making a Lora. Am I being lazy? Probably. You're not caring what I'm saying and you don't have to. Some glaring issues you'd probably encounter with the model is the fact you're mainly going to see a red background, whether you add it in or not. It's best to put "red background" in the negatives since most of the images I used in the dataset were red backgrounds because... duh, Grimm.Also there's sometimes a tail involved, so put that in the negatives, too. There's a chance the AI may generate multiple lines on his face (compared to his design where he only has one on each eye).Model's really good with EasyFluff (maybe cause I keep relying on it so much-) and struggles a bit with the Redwater checkpoint for anyone who uses it. But I think it's capable for general models; furry or not. If not, please do let me know which model the lora seems to do the worst (or even the best!) in. I'd really appreciate any sort of feedback as I'm both excited and incredibly nervous. ;_; A lil bit of spewing below. ----- People need to know there's more than just Hornet that exist in Hollow Knight. I noticed no other Lora for a Hollow Knight character has been made (especially surprised there isn't one for the Knight, who's equally as popular as Hornet lol), so I thought, "Y'know what? Be the change you want to be.", and thus began my process of making a Hollow Knight Lora. It was honestly a hit or miss, as many attempts to make a Lora out of lesser drawn/known characters was practically difficult as heck. I've tried numerous attempts and all I got was half-baked attempts or attempts that the AI didn't seem to try with at all. But honestly, that was on my end since it was hard to find "the right images". But after deciding on Grimm, I said "heck it" and just carefully spent like an hour or two finding the right images to put into my database. And lo and behold, the Grimm lora became a shocking success. There're still issues to be dealt with, but I hope to handle it eventually (when I get enough buzz to retrain lol).

Grimm (Hollow Knight)

First version. Not perfect by a longshot, but hopefully still passable to publicize as a lora. Will try and work through a better database for even better Grimm pics. Hoping to add more variety to him.

Files with SHA256: 415ef9d45ac26cc5bd9d6fd9fafd9b3cdea4706164a2e1ec6fe6983ff5b57b5f