style of the visual nove kindred spirits on the roof the lora doesn't complete capture the style of game game, but i still like the pictures it produces.triggerword: kd_vnstyleweight: 0.7 - 1works with anime and realistic modelstipp: use (Detail Tweakter -0.5) to -1.0 to get the style from the preview imagesSDXL Version Available here:<a target="_blank" rel="ugc" href="https://civarchive.com/models/303397/sdxl-kindred-spirits-on-the-roof-or-okujo-no-yurirei-san-or">https://civarchive.com/models/303397/sdxl-kindred-spirits-on-the-roof-or-okujo-no-yurirei-san-or</a><ul><li>no-keywords: just as said, no keywords where used for training</li><li>v2: first posted version on civitai</li><li>v3: yuri edtion: first version which includes some yuri pictures as training images</li><li>v4: works somewhat with anime models, doesnt capture the style as good as v5 but creates sometimes better images with different concepts</li><li>v5:was trained with an semi realistic model now anime models also work correctly, best results are still with kawaii realistic models and detail tweaker follows the style better then previous version, is not so good with creating completely different pictures from kindred spirits ( lik in space, military uniform, etc)</li></ul>

kindred spirits on the roof | Okujō no Yurirei-san | 屋上の百合

kindred_spirits_v5.safetensors

v5.0

v4.0

no-keyword.experimental

v3 - yuri edition

Kindred Spirits on the Roof.v2

style of the visual nove kindred spirits on the roof the lora doesn't complete capture the style of game game, but i still like the pictures it produces.triggerword: kd_vnstyleweight: 0.7 - 1works with anime and realistic modelstipp: use (Detail Tweakter -0.5) to -1.0 to get the style from the preview imagesSDXL Version Available here:<a href="https://civitai.com/models/303397/sdxl-kindred-spirits-on-the-roof-or-okujo-no-yurirei-san-or" rel="ugc noopener noreferrer nofollow" target="_blank">https://civitai.com/models/303397/sdxl-kindred-spirits-on-the-roof-or-okujo-no-yurirei-san-or</a><ul><li>no-keywords: just as said, no keywords where used for training</li><li>v2: first posted version on civitai</li><li>v3: yuri edtion: first version which includes some yuri pictures as training images</li><li>v4: works somewhat with anime models, doesnt capture the style as good as v5 but creates sometimes better images with different concepts</li><li>v5:was trained with an semi realistic model now anime models also work correctly, best results are still with kawaii realistic models and detail tweaker follows the style better then previous version, is not so good with creating completely different pictures from kindred spirits ( lik in space, military uniform, etc)</li></ul>

Files with SHA256: 538983155b109ba38f5739b0a43aa7babd8470c0b758ff7fdfc10d8a943d6d24