<h2 id="introduction">Introduction</h2>This is a concept/style LoRA for generating animegao kigurumi photography: performers wearing full-head anime character masks and costumes. It focuses on the distinctive combination of a rigid character head, a fixed anime-style expression, a wig, and a human-proportioned costumed body.Anima is not particularly strong at rendering photorealistic images, so the results may retain a noticeable cartoon-like appearance. Combining this with a realism-enhancing LoRA might help, but I have not tested that yet.<h2 id="usage">Usage</h2>Recommended LoRA strength: 1.0Trigger word:<pre><code>animekig</code></pre>Describe the character, outfit, pose, and setting after the trigger word. For example:<pre><code>animekig, photorealistic, solo, pink wig, blue eyes, maid dress, white thighhighs, standing in a studio</code></pre><pre><code>animekig, photorealistic, two performers, colorful idol costumes, posing together on a stage</code></pre><h2 id="training-details">Training Details</h2><ul><li>Base model: Anima Base 1.0</li><li>Training tool: <code>sorryhyun/anima_lora v1.17.1</code></li><li>Dataset: 162 curated and captioned photographs</li><li>Captioning model: JoyCaption Beta One Q8</li><li>LoRA rank / alpha: 32 / 128</li><li>Learning rate: <code>2e-5</code></li><li>Optimizer: AdamW</li><li>Learning rate schedule: Cosine decay with 5% warmup</li><li>Training duration: 8 epochs / 1,296 steps</li><li>Released checkpoint: Epoch 8</li><li>Precision: BF16</li><li>Resolution: Aspect-ratio buckets, approximately 1 megapixel</li><li>GPU: NVIDIA GeForce RTX 3090</li></ul><h2>简介</h2>这是一个用于生成 Animegao Kigurumi 摄影图像的风格/概念 LoRA，也就是由真人穿戴动漫角色头壳及角色服装进行扮演的形式。它主要学习硬质角色头壳、固定的动漫面部表情、假发，以及真人比例服装身体之间的视觉组合。Anima 对真人图片的表现并不特别好，因此生成结果仍可能带有比较明显的卡通感和涂抹感。搭配一些强化写实效果的 LoRA 也许会有所改善，不过我还没有实际测试过。<h2>使用</h2>推荐权重：1.0激活词：<pre><code>animekig</code></pre>在激活词后描述角色外观、服装、姿势和场景。例如：<pre><code>animekig, photorealistic, solo, pink wig, blue eyes, maid dress, white thighhighs, standing in a studio</code></pre><pre><code>animekig, photorealistic, two performers, colorful idol costumes, posing together on a stage</code></pre><h2>训练参数</h2><ul><li>基础模型：Anima Base 1.0</li><li>训练工具：<code>sorryhyun/anima_lora v1.17.1</code></li><li>训练数据：162 张经过筛选和标注的照片</li><li>标注模型：JoyCaption Beta One Q8</li><li>LoRA rank / alpha：32 / 128</li><li>学习率：<code>2e-5</code></li><li>优化器：AdamW</li><li>学习率调度：余弦衰减，预热比例 5%</li><li>训练量：8 epochs / 1,296 steps</li><li>发布版本：第 8 epoch</li><li>训练精度：BF16</li><li>训练分辨率：按长宽比分桶，约 1 M 像素</li><li>训练显卡：NVIDIA GeForce RTX 3090</li></ul>

Files with SHA256: ab8d4b89d6886d00f10a5cf00a5ebc1fbe99a46e3f9146b37230cc91eec854d8