<h3>使用前必读</h3>重要：一定要填写触发词语：Shiranui底模为 Anything3.0 ，经过测试，个人感觉在 anything3.0中有比较好的状态。<ul><li>在 Anything 3.0 的模型下，较好的参数为 0.6 - 0.8 之间</li><li>在 counterfeit2.5 的模型下也能获得比较好的效果，但整体颜色会略淡，尝试加高参数，或者使用 vae 改善部分效果</li><li>该模型对类 chilloutmix 模型（其他真人模型一样）适配并不是特别好，整体表现为：颜色过淡、形态不够</li></ul>需要注意的是：因为阿离在 CG 以及立绘中包含太多蝴蝶、樱花的元素，这类元素在 ai 学习的过程中很难去除，所以你会发现用了此 lora 的图会带上很多这类内容。（想要降低这部分可以通过添加反向 Prompt ，或者调低参数）<h3>ReadMe</h3>Important: Always fill in the trigger word: ShiranuiThe bottom mold is anything3.0. After testing, I personally feel that it is in good condition in anything3.0.Under the model of anything 3.0, the better parameters are between 0.6 and 0.8Better results can also be obtained under the counterfeit2.5 model, but the overall color will be slightly lighter. Try to increase the parameters or use vae to improve some effectsThe model is not particularly suitable for the model like chilloutmix (like other human models). The overall performance is that the color is too light and the shape is not enoughIt should be noted that because Ali contains too many elements of butterflies and cherry blossoms in CG and vertical painting, it is difficult to remove these elements in the process of AI learning, so you will find that using this Lora will bring a lot of these elements. (To reduce this part, you can add a reverse Prompt or lower the parameter)

Files with SHA256: d48b914a13b73d498428a5de644dc1e992454fb162cb07b1067a9617ac89ddb0