<h1 id="hakuro">白狼 / Hakuro</h1><h2>中文</h2><p>白狼是游戏《阴阳师》中的角色 LoRA，基于 <strong>Illustrious XL</strong> 训练。</p><h3 id="v2.0">V2.0 触发词</h3><p><code>hakurou*</code></p><h3>角色服装</h3><p><code>japanese_clothes, asymmetrical_clothes, hakama, muneate, single_glove, thigh_strap, bangle, socks, hair_ornament</code></p><h3>角色武器</h3><p><code>weapon, bow_\(weapon\), quiver, arrow_\(projectile\)</code></p><h3>使用提示</h3><ol><li><p>使用 <code>facepaint</code> 可以稳定触发角色的面部花纹。</p></li><li><p>使用 <code>gradient_hair</code> 可以稳定触发角色的渐变发梢。</p></li></ol><h3 id="v2.0">V2.0 更新内容</h3><ol><li><p>调整了 LoRA 触发词，减少底模词条污染。</p></li><li><p>修正了角色发色，并增加了对 <code>gradient_hair</code> 的稳定支持。</p></li><li><p>优化了面部花纹，现在可以使用 <code>facepaint</code> 稳定触发。</p></li><li><p>精简了 LoRA 模型文件大小。</p></li><li><p>调整并优化了部分训练集标签。</p></li></ol><hr /><h2 id="english">English</h2><p>This is a character LoRA for <strong>Hakuro</strong> from the game <em>Onmyoji</em>, trained on <strong>Illustrious XL</strong>.</p><h3 id="v2.0-trigger-word">V2.0 Trigger Word</h3><p><code>hakurou*</code></p><h3 id="character-outfit">Character Outfit</h3><p><code>japanese_clothes, asymmetrical_clothes, hakama, muneate, single_glove, thigh_strap, bangle, socks, hair_ornament</code></p><h3 id="character-weapons">Character Weapons</h3><p><code>weapon, bow_\(weapon\), quiver, arrow_\(projectile\)</code></p><h3 id="usage-tips">Usage Tips</h3><ol><li><p>Use <code>facepaint</code> to consistently reproduce the character’s facial markings.</p></li><li><p>Use <code>gradient_hair</code> to consistently reproduce the gradient at the ends of the character’s hair.</p></li></ol><h3 id="what's-new-in-v2.0">What’s New in V2.0</h3><ol><li><p>Adjusted the LoRA trigger word to reduce tag contamination from the base model.</p></li><li><p>Corrected the character’s hair color and added more consistent support for <code>gradient_hair</code>.</p></li><li><p>Improved the facial markings, which can now be reliably triggered with <code>facepaint</code>.</p></li><li><p>Reduced the LoRA model’s file size.</p></li><li><p>Revised and optimized tags in parts of the training dataset.</p></li></ol><hr /><h2 id="v1.0-v1.0-overview">V1.0 简介 / V1.0 Overview</h2><p>V1.0 是早期版本，使用不同的触发词</p><p>V1.0 is the original release and uses a different trigger word</p><h3 id="trigger-word">触发词 / Trigger Word</h3><p><code>hakurou_\(onmyoji\)</code></p><h3 id="character-outfit">角色服装 / Character Outfit</h3><p><code>japanese_clothes, hakama, muneate, gloves, thigh_strap, rope, bangle, socks</code></p><h3 id="character-weapons">角色武器 / Character Weapons</h3><p><code>weapon, bow, quiver, arrow_\(projectile\)</code></p>

白狼-阴阳师/hakurou_(onmyoji)-LoRA-光辉illustrious

<p>Update Notes:</p><ol><li><p>Adjusted the LoRA trigger words to prevent tag contamination from the base model.</p></li><li><p>Corrected the character’s hair color. To consistently reproduce the original hair color, use <code>gradient_hair</code>.</p></li><li><p>Improved the facial markings. You can now use <code>facepaint</code> for consistent results.</p></li><li><p>Reduced the LoRA model’s file size.</p></li><li><p>Revised the tags for parts of the dataset.</p></li></ol>

Files with SHA256: 86c62d4fa40b60f07e566f2dc8349c2ed064c81637400cd607d565f6f2491ad2