IMPORTANT: This is essentially a lazy port of my Flux lora. I put the training data together centered around Flux and I'm assuming that'll translate into less than ideal performance with sdxl. Still, I noticed there weren't any Satsuki loras on civit so I figured it was worth a shot. So far it seems to working better than I expected. The biggest issue is that the color scheme of her costume screws up every now and then. The size of her clown nose didn't seem to stick very well either so you might want to specify 'small clown nose' in the prompt. The lora also seems to suffer from some slight bleeding between style elements when compared to the Flux version. In short, you'll probably just need to make several tries at an image before getting an error-free Satsuki. Still, I'm fairly happy with the results given that the Flux lora was itself something of an experiment. I didn't really expect the sdxl version of it to be as viable as it turned out to be.Satsuki Iranami - The Ultimate Clown from the fan game Danganronpa Another. This lora should let you create images with Satsuki Iranami in a few different art styles and clothing options.I trained using natural language for the captions rather than tags. However, from limited testing the best option using this lora seems to be a mix of natural language and tags. Natural language for the Satsuki elements and then tags for everything else. However, I suspect that'll be fairly model dependent so it's probably best to just play around with it and see what sticks.I trained on a few specific image types which you can replicate, to varying levels of success. I tried to include a full range of the prompt options in the images. But the following is a brief summary.Danganronpa the Animation: Use 'This is an anime screencap from Danganronpa The Animation' in the prompt to emulate the art style from the anime series Danganronpa the animation.Danganronpa 3 (anime): Use "This is an anime screencap from Danganronpa 3."Sprites from Danganronpa: Mention 'danganronpa sprite style' in the prompt to try emulating the art style of Danganronpa. For example, 'this image is in the danganronpa sprite style'. Despite the name, I trained on full screenshots, with and without the GUI, with this description along with character sprite rips. So in theory it should be able to 'danganronpafy' a full image. Note that this was an attempt to specifically train on the sprite style from the official Danganronpa games rather than Danganronpa Another.Sprites from Danganronpa Another: In theory this is similar but distinct from the prior Danganronpa sprite style. It's the updated sprites used in Danganronpa Another. To use this style, include this within your prompt "The image is in Danganronpa another sprite style."Clothing: You can try different clothing options with prompts like the following:Standard clown outfit: "Satsuki Iranami is wearing a two-toned jester-like outfit split vertically down the middle with blue on the left side and pink on the right, and a matching jester's cap with white polka dots and a yellow band. Her costume features a white ruffled collar, yellow buttons down the front and curled shoes. Satsuki has short blue hair and striking yellow eyes. She wears a red clown nose."Swimsuit: "Satsuki Iranami is wearing a blue and pink bikini top with a yellow trim. The bikini bottom matches the design of the top. The left side of the bikini top and bottom are blue and the right sides are pink. Her short blue hair covers her ears and she has bright yellow eyes."Formal dress: "Satsuki Iranami wearing her dress from Danganronpa Another. She has short blue hair that is tied in a ponytail and yellow eyes. She is wearing a light blue dress with puffy short sleeves and a low neckline. The bottom of the dress features a layered design with ruffled tiers."Locations: I didn't try to train on any specific locations, but some of them appeared enough that you might be able to improve your results by specifying areas in Hope's Peak like "Hope's Peak's pool". Even just mentioning Hope's Peak might be enough to help push the lighting in an image in the right direction if you're generating something set in the school.Model notes: So far I've had the best results with animagineXL v31 and aniverseXL v20. Though animagine really, really, wants Satsuki to have bangs.

Files with SHA256: 8a247d1f27ba8756ffb8f916beac08b2dcf1ec0628226816046e6a47951a9354