V1: LoRA trained on NeoArtCorE, a famous NSFW artist. I should have input 'abs' in the prompts because that's definitely a signature of his art x) (not my favorite part though, to each their own but in the end I got used to it lol) I'm using DPM++ 2M Karras with 40 steps and CFG scale 7 I had great results with Abyss-based models like AbyssOrangeMix or KotoAbyssProto ..but it's up to you, really. Enjoy ! V3: This is a new version for SDXL. Check out "About this version" on the right for more information. Recommended resolution is 832x1216, with Pony Diffusion V6 XL model or other Pony-based models ONLY Note: Although it's called "v3C_PonyXL" here, in webUI it will be called as "PonyXL01C" which is the first name I gave it for the training then I realized my mistake later. It's confusing, sorry about that... Hopefully I won't do that error again...

NeoArtCorE-Lora

New NeoArtCorE LoRA based on the awesome PonyDiffusionV6XL model :) Almost every image in the dataset were in 832x1216 so that's the resolution I would recommend. A very few are 1216x832 but that's only 4% of the dataset... so that might work too but it's probably not going to be very creative. I also recommend using Pony as it might not work with other models; I haven' fully tested this LoRA yet, this description will be updated when I can find time to experiment more.

Files with SHA256: fa226063d5ed4ecf3a3b9dcc12468bb89768ee726519f7c5d5bf87819795da5a