[English / 中文]

This is a LoRA of Aris Tendou by which you could easily generate her images in different clothes. As I have been doing, I tried my best to reflect her personality in the model.

Prompts below are trigger words trained for generating particular elements:

head: halo, (very) long hair, ponytail, one side up, hair between eyes
clothes: aris tendou \(millennium unifrom\), aris tendou \(maid\), gym unifrom, idol clothes, nude

You could learn from example images how to use these tags properly.

The recommended range of weight is from 0.55 to 0.75. Stabiliy or similarity, it's up to you.

Suggestions: Under some circumstances results tend to be in 3D model style. This is because your prompts are similar to the training datasets. Adjusting prompts can solve the problem. From example, you could add more descriptions. Besides, adding the trigger word into negative prompts is effectibe when results are polluted by particular clothes.

Thanks for your support!

这是一个天童爱丽丝的模型，你可以用它轻易地生成不同服饰的爱丽丝。正如我一直在做的那样，我尽量让模型能够反映爱丽丝的独特气质。

下面是经过特别训练的关键元素触发词：

头部: halo, (very) long hair, ponytail, one side up, hair between eyes
服饰: aris tendou \(millennium unifrom\), aris tendou \(maid\), gym unifrom, idol clothes, nude

你可以参考例图来正确使用这些触发词。

推荐权重范围是0.55至0.75，代表着稳定性与相似度两种选择，你可以根据需要决定具体权重。

提示：一些情况下出图会有比较明显的3D质感，这是因为此时提示词过于接近数据集的训练提示词。修改提示词即可有效解决该问题，比如添加更详细的描述词。此外，将触发词加到负面提示词中可以有效避免服装污染的情况。

最后感谢每一位的支持！