This LoRA was trained to capture the style of the artist si_wooupp.It is particularly suitable for creating illustrations that blend 2D/anime-style characters with realistic or photographic backgrounds.Recommended trigger word:<code>@si_wooupp</code>Recommended weight:A weight of <code>1.0</code> is recommended.PS:<ul><li>This LoRA was trained with space-separated tags, not underscore-style tags. For example, use <code>1girl</code>, <code>upper body</code>, etc.</li></ul><ul><li><s>I‘m currently experimenting with using generated images plus manual edits to build a more diverse training set and achieve a stronger dreamcore/weirdcore aesthetic, which is planned for a future v1.0 release.</s> (already achieved in v1.0)</li></ul><ul><li><s>The current version can't generate charater with eyes closed, that because of dataset and will be fixed in the future version.</s></li></ul><hr />这个 LoRA 的训练目标是为了还原画师 si_wooupp 的独特艺术风格。它尤其适合生成二次元人物与真实感背景相融合的插画。推荐触发词：<code>@si_wooupp</code>推荐权重：推荐使用 <code>1.0</code>PS:<ul><li>本 LoRA 训练时使用的是空格格式的标签（tag），而不是下划线格式。例如，请使用 <code>1 girl</code>, <code>upper body</code> 等。</li><li><s>我目前正在尝试通过“</s><s>生成图片 + 手动修改</s><s>”的方式来获得更多样的训练集与</s><s>更强的梦核感</s><s>，并计划在 </s><s>v1.0</s><s> 版本中发布。</s> （已经在v1.0版本中实现了）</li><li><s>目前的LoRA在绘制闭眼的角色时会出现一些问题，这是数据集导致的。这些问题都会在未来的版本中修复。</s></li></ul>

artist：Si_/wooupp Style - Anima Base-v1.0

<ul><li>refine the generation of backgrounds</li><li>refine the generation of characters with eyes closed</li></ul>

Files with SHA256: 2dedb1ae9e6704ef36413e394d0fd7bb254ad8f1afce05afb4779b189c19bbe8