<h3 id="if-you-have-the-means-please-do-support-the-original-artist's-creative-work!">⭐ If you have the means, please do support the original artist's creative work!</h3><h3 id="a-lora-that-reproduces-the-drawing-style-of-the-artist-uutsuki-().">A LoRA that reproduces the drawing style of the artist uutsuki (うー☆月).</h3>Artist page: <a target="_blank" rel="ugc" href="https://www.pixiv.net/users/1709432">https://www.pixiv.net/users/1709432</a>Trigger word: <code>@uumoonstyle</code> — put it at the very beginning of your promptBase model: anima-base-v1.0 | Recommended LoRA weight: 1.0---What this model is tuned forThis artist works mainly in cross-section / x-ray, The training set is overwhelmingly adult material, so the model is tuned for that and will drift that way even on neutral prompts. For clean SFW output, drop to 0.4–0.6 or use a different model.Training data450 images, auto-tagged with WD14 then manually corrected across the whole set (no auto caption left untouched).Two things worth knowing before you use it: By making good use of cross-section, we can achieve good results- The set contains no foot material — see the limitations section, it matters.Recommended settings (verified)- Resolution: 1536 on the long side (trained at 1536, buckets 512–1536)- Sampler: euler + beta, CFG 5, 30–40 steps / CFG 1, 8 step with Turbo LoRa- Weight 1.0. The style is already established at 0.4; 1.0 adds structure coverage rather than more style.---Known limitations⚠Feet — read this before using it for full-body workThe training set contains no foot material.This is structural, not a seed problem. The trained layer applies modulation globally, and there is no in-domain reference for that body part — the texture gets applied to a region with no structural prior behind it.```a) Stack a foot-repair LoRA recommended, verified workingb) Lower this LoRA's weight costs style strengthc) Avoid foot close-ups full-body shots with small feet are fine```<code>x-ray</code> needs context<code>x-ray</code> on its own returns a generic skeletal X-ray — a blue-tinted skeleton through the skin. That is the tag's conventional meaning, not what this artist draws.```x-ray + the relevant scene tags → internal structure visible during the actx-ray alone → generic skeleton, not this artist's stylecross-section → inset frame / abdominal window```If you want a genuine skeletal X-ray, run the base model without this LoRA.### Frame position is not controllableThe inset cutaway may land top-left, on the right edge, or swallow the whole frame instead of staying inset. Which one you get is decided by the seed, not by step count or prompt weight.If you need a fixed position, add composition tags — <code>framed inset</code>, <code>upper left</code>, <code>picture-in-picture</code>. The base model understands them; how strictly the LoRA respects them is untested.---- ⛔ Do not use this model to generate sexual content involving any minor character---Base Model License &amp; AttributionCircleStone Labs Non-Commercial License v1.2:Full license: <a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima/blob/main/LICENSE.md">https://huggingface.co/circlestone-labs/Anima/blob/main/LICENSE.md</a>&gt; This model is not an official CircleStone Labs product and has not been endorsed, approved, or validated by CircleStone Labs.中文：⭐如果各位有能力的话，请务必支持原画师的创作工作！一个旨在还原画师 uutsuki（うー☆月） 绘画风格的 LoRA。画师主页：<a target="_blank" rel="ugc" href="https://www.pixiv.net/users/1709432">https://www.pixiv.net/users/1709432</a>触发词<code>@uumoonstyle</code> —— 放在提示词最前面底模：anima-base-v1.0 ｜ 推荐 LoRA 强度：1.0---这个模型擅长什么该画师的作品的剖面 / 断切构图是特色,配合柔和的皮肤质感与细致线稿。素材中成人内容占绝大多数，因此模型是按这一取向训练的，即使给中性提示词也容易往那个方向漂。需要干净的 SFW 输出时请把强度降到 0.4–0.6，或换用其他模型。训练素材450 张，WD14 自动打标后全量人工清洗（没有一张是纯自动 caption）。用之前有两件事值得知道：- 素材里完全没有足部内容，见下面的限制说明，这一条会影响你的用法。推荐参数（已验证）- 分辨率：长边 1536（训练即 1536，bucket 512–1536）- 采样：euler + beta，CFG 5，30–40 步 / Turbo Lora 下 CFG 1，8步- 强度 1.0。画风在 0.4 就已经成立，1.0 增加的是结构覆盖度而不是更多画风---已知限制⚠足部——画全身之前请先读这段训练集里没有任何足部素材。这是结构性问题，不是抽到坏种子。训练改的是全局调制层，而那个部位在训练集里没有任何参照——质感和光影被强行施加到一个没有结构先验的区域上。```a) 叠加一个修解剖的脚部 LoRA 推荐，实测有效b) 降低本 LoRA 强度 会损失画风c) 避开足部特写 全身图里脚占比小时不受影响```<code>x-ray</code> 需要上下文单独用 <code>x-ray</code> 会得到通用的骨骼透视——蓝色骨架透过皮肤。那是这个词的常规含义，不是这位画师的画法。```x-ray + 相应的场景 tag → 过程中内部结构可见（本画师画法）x-ray 单独 → 通用骨骼透视，不是本画师风格cross-section → 画框内嵌 / 腹部开窗```真的想要骨骼透视，请不要挂这个 LoRA，用底模。画框位置不可控内嵌的画框可能出现在左上、右侧，或者干脆吞掉整个画面而不内嵌。落在哪里由 seed 决定，跟步数和权重无关。需要固定位置就加构图词—<code>framed insetupper leftpicture-in-picture</code>，底模本身认识这些词；LoRA 对它们的服从程度尚未系统测试。---- ⛔ 禁止用于生成任何未成年角色的性内容---底模许可与署名CircleStone Labs Non-Commercial License v1.2许可全文：<a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima/blob/main/LICENSE.md">https://huggingface.co/circlestone-labs/Anima/blob/main/LICENSE.md</a>&gt; 本模型 非 CircleStone Labs 官方产品，也未经其认可、批准或认证。

Files with SHA256: 72b5a7de4804baecc3bbe3e513289c0bb0a4164b9fd1f4cc6d1bae5cb7304ae6