2024/12/25 update:use highfix or ADetailer if character's face get blurry or twisted.Recommand Weight: 1.1 ~ 1.5Trigger Word:<pre><code>asanagi-figure,</code></pre>Some Illustrious-XL based models have stong tendency to draw in anime style rather than in realistic style (for example, <a target="_blank" rel="ugc" href="https://civarchive.com/models/820208?modelVersionId=917170">Obsession</a>), for those models, increase this lora's weight accordingly, and you may also need to put trigger words like:<pre><code>(((asanagi-figure, photorealistic, photo \(medium\), realistic))), </code></pre>Fixed the problem that v1 will get blurry on recent illustrious model (for example, Noob 1.1)Add a new feature "cum on figure": use tag like:<pre><code>((cum on figure)), </code></pre>can generate realistic semen flooding on figure, but need cum related tags and will need to do some gachaOther Useful tags for cum on figure : cum,cum overflow, bukkake, cum on body, facial, etc.建议使用高清修复或者ADetailer来修复人物脸部模糊或者扭曲的问题推荐权重：1.1~1.5修复了v1无法兼容最近较新的illustrious模型的问题，此外增加了射手办上的新功能，能生成某些变态的返图一样的效果。。。不过需添加精液相关tag并多roll几次。<hr />Trained specifically for figure works originally Illustrated by Asanagi, trained on Illustrious-XL 0.1Also works on several Illustrious-XL based model (for example, <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294?modelVersionId=968495">NoobAiXL, </a><a target="_blank" rel="ugc" href="https://civarchive.com/models/818750?modelVersionId=915538">Lucereon</a>), NoobAiXL seems to have best result so far in my test.Recommand Weight: 1.0 ~ 1.2Trigger Word:<pre><code>asanagi-figure,</code></pre>Some Illustrious-XL based models have stong tendency to draw in anime style rather than in realistic style (for example, <a target="_blank" rel="ugc" href="https://civarchive.com/models/820208?modelVersionId=917170">Obsession</a>), for those models, increase this lora's weight accordingly, and you may also need to put trigger words like:<pre><code>(((asanagi-figure, photorealistic, photo \(medium\), realistic))), </code></pre>Most data was collected on Twitter.Old version for Animagine V3 &amp; V3.1: <a target="_blank" rel="ugc" href="https://civarchive.com/models/362689/pvc-figure-illustrated-by-asanagi-for-animaginexlv3-and-animaginexlv31-or">PVC Figure Illustrated by Asanagi (for AnimagineXLV3 and AnimagineXLV3.1) | 朝凪手办风 - v2.0 for ani3.1 | Stable Diffusion XL LyCORIS | Civitai</a>（Because the usage is different so I open this new post）<hr />专门为朝凪的手办化作品训练的画风，底模为Illustrious-XL 0.1, 其他基于Illustrious-XL 0.1的微调模型理伦上也能生效(比如<a target="_blank" rel="ugc" href="https://civarchive.com/models/833294?modelVersionId=968495">NoobAiXL, </a><a target="_blank" rel="ugc" href="https://civarchive.com/models/818750?modelVersionId=915538">Lucereon</a>)，NoobAiXL在我的测试里表现是最好的。推荐权重：1.0~1.2触发词：<pre><code>asanagi-figure,</code></pre>部分基于Illustrious-XL 0.1的微调模型有很强的动漫画风倾向（比如 <a target="_blank" rel="ugc" href="https://civarchive.com/models/820208?modelVersionId=917170">Obsession</a>），在这类底模上需要提高这个lora的权重，同时你可能需要这么写触发词：<pre><code>(((asanagi-figure, photorealistic, photo \(medium\), realistic))), </code></pre>绝大部分数据收集自推特的买家返图。针对Animagine V3 &amp; V3.1训练的旧版模型：<a target="_blank" rel="ugc" href="https://civarchive.com/models/362689/pvc-figure-illustrated-by-asanagi-for-animaginexlv3-and-animaginexlv31-or">PVC Figure Illustrated by Asanagi (for AnimagineXLV3 and AnimagineXLV3.1) | 朝凪手办风 - v2.0 for ani3.1 | Stable Diffusion XL LyCORIS | Civitai</a>（主要是因为用法不同所以我重新开了这个新的post）

Files with SHA256: 375c0ade6fc217fd24779bf979ebe608afcea18a6d2fac7638007f49aa534792