Stable diffusion Lora person arrangement #7
I summarized the feel of the model with a Lora file that can be downloaded from https://civitai.com/ .
The prompts are commonly PureErosFace_v1:0.1, ulzzang-6500:0.1, cute, (20 years old beautiful Korean woman:1.1), and (kpop idol, Korean beauty, Korean mixed) values, so I applied how they change by giving them Lora 1:1 Among them, Lora, which came out a little strange, gave a value of 1:0.5 to 1:0.3.
Steps: 40, Sampler: DPM++ SDE Karras, CFG scale: 7, Face restoration: CodeFormer, Size: 512x512,
Dynamic thresholding enabled: True, Mimic scale: 10, Threshold percentile: 100