[SDXL]PureSerenity Style
详情
下载文件
关于此版本
模型描述
本模型提供中文、英语、日文三种语言的介绍。
===========================
基于不同风格的样例进行的训练,后续进行了融合,以提供一种可爱唯美又简单的画风,以及富有细节的眼眸。
模型采用了Locon的Lycoris网络算法来提升了模型稳定性以及鲁棒性,便于与其他lora模型进行融合,而不会造成过多的tags污染。
参数推荐:权重可以在0.1-1.5以内随意调控(推荐权重在0.8-1.2),具体效果依据于您希望的效果以及您所选用的checkpoint对于Lora模型的兼容度。
模型参数:network_dim 64
network_alpha 32
conv_dim 16
conv_alpha 8
注:非常推荐使用高分辨率修复 (Hires. fix),可以显著提升画面细节,修复可能存在的模糊问题。
———————————————————
Training was conducted based on examples of different styles, followed by fusion to provide a cute, beautiful, and simple art style, as well as detailed eyes.
The model adopts Locon's Lycoris network algorithm to improve its stability and robustness, making it easy to integrate with other Lora models without causing excessive tag pollution.
Parameter recommendation: The weight can be adjusted freely within 0.1-1.5 (recommended weight is 0.8-1.2), and the specific effect depends on the desired effect and the compatibility of the checkpoint you choose with the Lora model.
model parameter:network_dim 64
network_alpha 32
conv_dim 16
conv_alpha 8
Note: It is highly recommended to use high-resolution fixes (Hire. fix), which can significantly improve the details of the image and fix possible blurring issues.
———————————————————




