Connie Maheswaran in DoompyPomp Style trained with 73 images. Epochs 12Main/Base Trigger words to put in positive prompts: conniemaheswaranDP, 1girl, dark-skinned female, long jacaranda hair, round glassesGym Uniform: gym uniform, amaranth short shorts, white and blue shirt, socks with stripes, pink sneakersOther words for more style to put in positive prompts: wet, sweat, cum, excessive cumWords to add for negative prompt (Not absolutely necessary from what I've tested but in case you start seeing speech bubbles, sound effects, and text in the images for any reason; just put the following words in negative) : english text, speech bubbleOTHER NOTES: Not a perfect training on the character but good enough for now. A few mistakes I made in training this LoRA was that 1) I didn't train in tags of backgrounds because I didn't factor in using the same backgrounds over and over again would overcook them into the LoRA for each gen so you'll probably see those same backgrounds constantly if you don't add them to negative prompts (well at least from what I've tested with the Illustrious-XL checkpoint. If you use them in another checkpoint though it should work a bit though OR just repeat what I did in my positive and negative prompts for the previews). 2) Not factoring or simplifying the tagging of where the character was actually looking made a bad mixture where sometimes she doesn't actually look at the viewer if put in positive prompts. Next version I have these mistakes corrected though but generally everything else works the way I like it to be generated.Alright as usual, The checkpoints I'll suggest with using this are Gehenna, Hassaku, WAI-NSFW, and Illustrious-XL. Surprisingly, You should be able to get good results even from Illustrious-XLYOU DO NOT NEED TO PUT WATERMARK/SIGNATURE IN NEGATIVE TAGS SINCE I ONLY USED IMAGES THAT DIDN'T HAVE WATERMARKS/SIGNATURES IN THE FIRST PLACE FOR TRAINING!!Also, I've been seeing people try to use SDXL and/or SD 1.5 with Illustrious Models (specifically the ones I've trained too). I heavily suggest against this unless you're going to turn the strength down on them by at least 0.8 or lower as Illustrious doesn't mix well with those models and can give you some really bad artifacts and meshing when generating. Illustrious suprisingly does better blending with Pony than anything despite not fully getting the original concepts on the things the Pony models were trained on BUT even then I wouldn't suggest unless you're REALLY curious and know your way with other LoRAs and models. If you seen me do it in the past it was only because I was curious about how they worked NOT because it helps with generations.

Files with SHA256: 152e08f76807eed680aca0408dd264676e7540c1deffd2869d9b3a07ea413428