Reminder to users [READ PLEASE!]

You are responsible for what you create | 请做一个有责任感的互联网公民

Please note that this LoRA was trained using some real-person photos, however, it does not resemble any human beings living in the real world. Nonetheless, models trained with such materials may still raise concerns.

As the creator of this LoRA, I will not and can not take any responsibility for how the generated images are used or distributed by others. It is the responsibility of the users to ensure that the images they create and share comply with all relevant laws and ethical standards.

Please use this LoRA responsibly. And bare in mind that what you post online will shape how the masses think of AI technologies.

请注意，这个LoRA是用一些真人照片训练出来的，然而，它并不像生活在现实世界中的任何人类。尽管如此，用这种材料训练出来的模型仍然可能引起人们的关注。

作为这个LoRA的创造者，我不会也不可能对他人如何使用或传播所生成的图像承担任何责任。用户有责任确保他们创建和分享的图像符合所有相关法律和道德标准。

请负责任地使用这个LoRA。请记住，你在网上发布的内容将影响大众对人工智能技术的看法。

License | 许可

No commercial use. Patreon, fanbox, or other equivalent type of pay-walled content is not permitted as well.

不得用于商业用途。也不允许Patreon、fanbox或其他同等类型的付费内容。

Details of this LoRA | 细节介绍

This is my second LoRA, so it is not perfect.

Trigger prompt: <lora:liliz_v2-000014:1>, I recommend setting weight to 0.4-0.7. I use 0.66 for image you see here.

Trained with orangeChillSlushyMix(Mix from Chillout:0.3 and orangeslushymixV2Buff_v10:0.7). Datasets contains 50~ pictures gathered from internet.

While it does not look good on Chillout, but on other models should get great results. Tested with various models, see images for different model output.

Capable of both anime and realistic style. Personally I use DosMix to get 2.5D look. But you're welcome to try different combinations.

这是我的第二个LoRA，所以它并不完美。

触发提示。<lora:liliz_v02:1>，我建议将权重设置为0.4-0.7。我使用0.66来制作你在这里看到的图像。

用orangeChillSlushyMix（来自Chillout的混合：0.3和orangeslushymixV2Buff_v10：0.7）进行训练。数据集包含从互联网上收集的50~张图片。

虽然它在Chillout上看起来不是很好，但在其他模型上应该会得到很好的结果。用各种模型进行了测试，不同模型的输出见图片。

既能做动漫，也能做写实风格。我个人使用DosMix来获得2.5D外观。但也欢迎你尝试不同的组合。

Setup | 设置推荐

Base resolution | 基础分辨率: 512*768 or 768*512

Sampling steps | 采样步数: 20/30, depends on Sampler you choose.

CFG Scale | 提示词相关度: keep it at 7. Higher will results in bad image during my testing.

HiRes fix and face fix is recommended. Parameters for HiRes fix | 建议使用HiRes修复和面部修复。HiRes修复的参数如下:

Upscale method | 超分辨率算法: none,

Hires steps | 步数: 10,

Denoising | 重绘幅度: 0.3,

Upscale by | 放大倍数: 2.

I encourage you to not post lewd or otherwise NSFW pictures in reviews. But for SFW pictures please do. I would love to see how it looks on your setup.

最好不要在评价中发布色情及NSFW(不适宜展示内容)的图片, 对于正常的效果展示很欢迎,我很想知道她在你的机器和设置上看起来是什么样的.