As usual, the model is trained with clip skip 1 on top of <a target="_blank" rel="ugc" href="https://huggingface.co/JosephusCheung/ACertainty">ACertainty</a> and locon is extracted from the trained modelThe locon should work on most models as I show in the xy plotAll the intermediate checkpoints can be found on <a target="_blank" rel="ugc" href="https://huggingface.co/alea31415/shokei-shojo">https://huggingface.co/alea31415/shokei-shojo</a>It includes the following 7 characters from the series<ul><li>Ashuna</li><li>Flare</li><li>ManonLibelle</li><li>Menou</li><li>Momo</li><li>Pandemonium</li><li>TokitoAkari</li></ul>For Menou you have both<ul><li>Menou</li><li>Menou child</li></ul>For Pandemonium you have<ul><li>Pandemonium (the classic one)</li><li>Maya Pandemonium (hair down version, the one after the end of the 5th tome of the novel)</li><li>Centipede Monster (monster version, the one at the 12th episode of the anime)</li></ul>Maya Pandemonium is only trained on 6 images from the light novel: 2 colorful and 4 monochrome, but it should still be usable if you adjust things properlyThe preview of Centipede Monster is only included for the FullCkpt version, because I guess it may make some uncomfortable (and anyway I do not think anyone would ever use it)I split the styles into 3+1 during training<ul><li>aniscreen (3928 images)</li><li>fanart (95 images)</li><li>colorful light novel / light novel (153 images)</li></ul>As usual I use a lot of regularization images during training, in particular I also include most of my dataset for the in container loha (<a target="_blank" rel="ugc" href="https://civarchive.com/models/66021">https://civarchive.com/models/66021</a>) here so I guess these concepts are more or less learned as wellThe example images are either generated by the LoCon on top of <a target="_blank" rel="ugc" href="https://mega.nz/folder/M0twDRRR#yLh3JctQrP8S6J5dnUStbA">VBP2.3</a>Or using a certain merge of the trained modelNone of the images are generated in one shotThey generally include a lot of random draws, img2img, inpainting, and sometimes use of controlnet and regional prompter (for the case of multiple characters)--Just to complain, I really have a hard time figuring out what kind of images violate civitai TOSI wanted to upload an image with Pandemonium on top of ManonLibelle who is lying in the blood (<a target="_blank" rel="ugc" href="https://huggingface.co/alea31415/shokei-shojo/blob/main/example_images/00154-20230604043947.png">this one</a>), but apparently this is detected as violation of TOS and gets removed automaticallyIn the mean time, we can easily find a lot of images that are on civitai that are more violent/horrible/bloody or whatsoeverOtherwise, although probably few people would ever be interested in a model for thisShokei shojo is basically my favorite light novel, so I kind of feel that I must do itBtw I saw an image of Menou from this discussion on outpainting of controlnet <a target="_blank" rel="ugc" href="https://github.com/Mikubill/sd-webui-controlnet/discussions/1464">https://github.com/Mikubill/sd-webui-controlnet/discussions/1464</a> so I told myself it was time to get this model done

The Executioner and Her Way of Life-characters-LoCon/FullCkpt 処刑少女の生きる道 | Menou / TokitoAkari / Momo / Ashuna/ Pandemonium / ManonLibelle

Trained checkpoint + custom merge(for the original trained checkpoint see <a target="_blank" rel="ugc" href="https://huggingface.co/alea31415/shokei-shojo/">https://huggingface.co/alea31415/shokei-shojo/</a>

Files with SHA256: bf4f56da407571b368efeb660ab5d9f72daab4d8a4116fcb3fc1110e5875f5fd