V5看上去还原度更高也更清晰一点，但是也看上去更像AI生成了（困扰）

Version V5 seems to have higher fidelity and more clarity, but it also looks more like AI-generated

=====================

十张图让男人炼了一百炉 孩子们5090的电费好贵

recommend settings:

Euler a / SGM Uniform / step 25-30 / cfg 4-5 / RTX upscale x2 or latent upscale

v2 更少过拟合，更闪亮的裤袜，更还原风格，代价是提示词必须写得更详细

V2: Less overfitting, the pantyhose look is more shiny. The cost is that the prompts must be written in more detail.

触发词还是@dksnstyle

trigger word: @dksnstyle

一些你可能需要的提示词：shiny/see-through pantyhose/bodystocking/legwear/clothes

=====================

v1 trigger word: @dksnstyle

It's slightly overfitted; you could try lowering the LoRA weight or not using the trigger word.

v1 有点过拟合，可以试试降低Lora权重或者不使用触发词

呃啊 关注美少女(?)画师谢谢喵 https://www.pixiv.net/users/2306287