<h1 id="konosuba-collection-n7dux01yk">Konosuba - この素晴らしい世界に爆焔を！ - Collection</h1>Collection of character LoRA's for the kono subarashii sekai ni shufuku wo! (この素晴らしい世界に爆焔を！) series, intended to generate non-official / fan art style images.<h2 id="character-loras:-egtqam813">Character LoRAs:</h2>See versions for each character LoRA and recommended settings.<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/398623?modelVersionId=444547">Aqua - アクア</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/398623?modelVersionId=445565">Darkness - ダクネス</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/398623?modelVersionId=445632">Megumin - めぐみん</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/398623?modelVersionId=445716">Yunyun - ゆんゆん</a></li></ul><h2 id="multiple-characters-lora:-adfmrmgvk">Multiple Characters LoRA:</h2><a target="_blank" rel="ugc" href="https://civarchive.com/models/398623?modelVersionId=445826">Konosuba Merge Collection</a>Extra LoRA trained on images that included multiple characters, then merged with individual character LoRA's above.This one works best with promps for multiple subjects, i.e. <code>2girls</code>, or <code>1girl, 1boy</code>.Includes additional characters:<ul><li>Kazuma Satou - 佐藤 和真</li><li>Wiz - ウィズ</li><li>Iris - アイリス</li><li>Eris - エリス</li><li>Shinjin Succubus - 新人サキュバス</li></ul>Advanced tip (A1111 or Forge users)I recommend trying the merge collection LoRA with the <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-regional-prompter">sd-webui-regional-prompter</a> or <a target="_blank" rel="ugc" href="https://github.com/Haoming02/sd-forge-couple">sd-forge-couple</a> extensions.I used <a target="_blank" rel="ugc" href="https://github.com/Haoming02/sd-forge-couple">sd-forge-couple</a> and adetailer with [SEP] tags for the group shot generations in the previews.Great article by <a target="_blank" rel="ugc" href="https://civarchive.com/user/FallenIncursio">FallenIncursio </a>on how that works:<a target="_blank" rel="ugc" href="https://civarchive.com/models/339604?modelVersionId=403892">How to Generate Multiple Different Characters, Mix Characters, and/or Minimize Color Contamination</a><h2 id="combined-v2-ujo60gwjm">Combined v2</h2>This version combines all datasets, train together with additional revised tagging of quality and nsfw:Quality modifiers:<code>masterpiece, best quality, low quality, aesthetic, very aesthetic</code> NSFW tags:<code>nsfw, explicit</code>Recommended prompt structure:I recommend using around a ~0.8 weight.Positive prompt:<pre><code>{{tags}}
score_9, score_8_up, score_7_up, score_6_up, source anime, masterpiece, best quality, very aesthetic</code></pre>Negative prompt:<pre><code>score_4, score_5, score_6, monochrome, (censor), 3d, realistic, render, CGI, text, watermark, cartoon, sketch, greyscale</code></pre><hr /><h3 id="license-elijr4rvb">License</h3><a target="_blank" rel="ugc" href="https://freedevproject.org/faipl-1.0-sd/">Fair AI Public License 1.0-SD</a>

Konosuba - この素晴らしい世界に爆焔を！- Collection

Trained on <a target="_blank" rel="ugc" href="https://civitai.com/models/288584/autismmix-sdxl">AutismMix SDXL</a>At 50 Epochs, it may be a little over-trained. Even low weights (~0.4) should have a clear effect.

Files with SHA256: 3f7a013636c91efcbf6bc3e1a5dd5b2825e155c40e955fe645723901f7db0d0e