Buy me a coffee ☕

Flux.1 D version: zyd232's Chinese Girl LoRA - Flux

SDXL version: zyd232's Chinese Girl LoRA - SDXL

This LORA is created with more than 100 pictures of some cute Chinese girls. The resolution of these pics is 768*768. I recommend you use HI-RES to generate pics with this LoRA.

Theoretically, you can use it with any models based on SD 1.5. In order to get good results, I suggest you use this LoRA with those mixed models containing Asian female faces, such as shampooMix, ChilloutMix, RealDosMix, schoolmax 2.5d and so on.

For version 1.6 and 3.9:

In my tests, the perfect weight of this LoRA is around 0.4~0.7. If you find your other tags don't change the result, just try increasing the weight of your tags (1~1.5).

For version 4.0 and later:

You can use this LoRA with the weight of 1.0 or lower.

The negative prompt in example images has something wrong: the second half of the negative tags is duplicated with the prompt tags.

Just ignore it. (The problem is caused by an extension.)

You are welcome to share your pics generated with this LORA. But please keep those extremely NSFW ones to yourself.

-

这个LORA由一百多张可爱女生的照片炼成。分辨率为768*768。推荐使用高清修复（HI-RES）功能，以得到更多细节。

理论上，任何基于SD 1.5而炼成的模型都可以与这个LORA兼容。为了获得更好的结果，建议使用那些带有亚洲女性面孔的模型，比如 shampooMix、ChilloutMix、RealDosMix、schoolmax 2.5d 等等。

SDXL 版本: zyd232's Chinese Girl LoRA - SDXL

对于 version 1.6 及 3.9:

在我的测试中，此LORA的权重在0.4~0.7这个区间表现最好。如果你发现你的其他tag对结果产生的影响太小，可以尝试增加这些tag的权重（我的经验是增加到1~1.5）。

对于 version 4.0 以及更高版本：

这个 LoRA 的权重可以设为 1.0 或再低一些。

示例图片中的负面提示词有些问题：负面提示词后半部分的内容和提示词重复了。

请无视。（这个问题是由于我用的某个插件造成的。）

欢迎在评论区分享你用这个LORA生成的图片。但是请不要分享过于暴露的内容。