This thing is hard to find dataset for........it's not too popular of a concept.so again i went the MMD-&gt;Cinema4D-&gt;marvelous designer-&gt;blender-&gt;first lora-&gt;inpaint first lora outputs-&gt;2nd lora route.The reflections of the first raw dataset looks like this, so when overfitted or cranked up too high, lora is gonna spit out some stereotypical PBR reflections.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c9332adc-0264-47c6-a50f-f4bb50c22ea7/width=525/c9332adc-0264-47c6-a50f-f4bb50c22ea7.jpeg" />I recommend the 40 epoch version, but like when i tried to use it with the other lora i just trained, it only started working with &gt;100 epoch versions, so im uploading several versions so people can pick their poison.(why is there a 300epoch version when it's done in 40epochs? cuz i left my PC on and went to sleep

Files with SHA256: ad70813c010df402e1eea16e9ee116e051cabde030485d5de53084d75ecf4257