处理素材实在太花费时间了，每一个素材都需要经过十几次混合，而且这还不包括测试。Processing the materials is indeed very time-consuming, as each material requires more than ten rounds of mixing, not to mention the testing involved回归主题，接下来分享一下近期发现Returning to the main topic, I would like to share some recent discoveries1：使用细分画风模型进行画风混合效果很不错。这类模型通常是画风模型再次分解的产物。1: The use of a subdivided style model for style blending effects is quite impressive. Such models are typically the result of further decomposition of style models.2：对模型进行细分时会看到模型出现了回溯现象。When the model is subdivided, a phenomenon of backtracking can be observed in the model3：出现回溯的模型在混合时与原模型有本质的区别。建议你们使用0.1~0.3这个数值进行对比测试The model that exhibits backtracking is fundamentally different from the original model when mixed. It is recommended that you use the values ranging from 0.1 to 0.3 for comparative testing

KM-MIX-2.5DRealism-Linn

画风 style <a target="_blank" rel="ugc" href="https://civitai.com/models/245294?modelVersionId=276717">https://civitai.com/models/245294</a>人物脸模 Facial model ：<a target="_blank" rel="ugc" href="https://civitai.com/models/50794">https://civitai.com/models/50794</a>我以这个模型的初始版本为平台进行了一些混合尝试，但是素材提取过程非常繁琐，所以实验进展十分缓慢。I conducted some mixed attempts using the initial version of this model as a platform, but the material extraction process is very cumbersome, resulting in slow experimental progress实验性的使用了LORA强效补剂（内含15个LORA进行混合）用于对模型进行升级优化。当然这并没有影响本次混合的主要LORAExperimental use of Lora high-efficiency supplements (containing 15 Lora for mixing) was conducted for the purpose of upgrading and optimizing the model. Of course, this did not affect the primary Lora of this mixture接下来同步近期发现，应该有人跟上我的进度了吧Next, I will synchronize the recent findings; there should be someone who has caught up with my progress

Files with SHA256: 7c4f19e53ea9fb64317fb3a4d1c02e9c150e8fe4880962c6d2518a01f9ccf9f1