<h3 id="highly-recommended:-v2.0-!!!!">Highly recommended: v2.0 !!!!</h3><h2 id="v1.0-exp-is-for-experiment-if-you-want-to-have-a-more-detailed-and-controllable-output-plz-use-normal-v2.0!">v1.0 exp is for experiment, if you want to have a more detailed and controllable output plz use normal v2.0!</h2>This LoRA was trained to capture and replicate the art style of <code>@Antifreez3_ </code>.Recommended trigger word:<code>@tldhfdy_style</code> for v2.0<code>@tldhfdy</code> for exp v1.0 and beforeRecommended weight:A weight of <code>1.0</code> is recommended, but you can adjust it slightly higher for a stronger style effect.PS:This LoRA was trained with space-separated tags, not underscore-style tags. For example, please use <code>simple background</code> and <code>upper body</code>, not <code>simple_background</code> or <code>upper_body</code>.When I trained this LoRA, I did not know the artist had an English name, so I used the pinyin of the artist’s name as the trigger word. If enough people like this model, I may consider adding more training images and training a new version with the correct trigger word. (ps: sorry guys, I forgot QWQ) (ps2: sorrrrrrrrry , I forgot again QWQ QWQ)<hr /><h3></h3>这个 LoRA 的训练目标是为了还原画师 天凉多喝防冻液（<code>@Antifreez3_</code>）的个人艺术风格。推荐触发词：<code>@tldhfdy_style</code> （v2.0）<code>@tldhfdy</code> （exp v1.0 及 以前）推荐权重：推荐使用 <code>1.0</code> 的权重，如果需要更强烈的风格表现，可以适当调高权重。PS: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>simple background</code> 和 <code>upper body</code>，而不是 <code>simple_background</code> 或 <code>upper_body</code>。当我训练这个 LoRA 时，我还不知道画师有英文名，所以用了他名字的拼音当作触发词。如果有足够人喜欢这个模型的话，我会考虑补充一下训练集，然后训练一个有着正确触发词的版本。(对不起各位，我把这茬忘了 QWQ) （对不起，我又又忘了QWQ QWQ）

artist：天凉多喝防冻液 (tldhfdy) / Antifreez3_  Style - Anima Base-v1.0

This is a MAJOR update! The LoRA has been improved across the board, especially in terms of material quality and texture. The overfitting issues found in previous versions have also been significantly reduced. However, this version may occasionally generate distorted toes. I’ll try to address this issue in future updates.

Files with SHA256: eba01194aaed8f9bb5de18fb2b59d74c52ed4df12d97b97e8b8b42b124dccd6d