<ul><li>汉服/唐装写真LoRA，</li><li>权重0.4-0.8，</li><li>底模选择真实系就可以，示例图用的底模是majicMIX realistic 麦橘写实_v7，</li><li>AD修复开不开都可以，模型本身可以控住人脸，但是手部最好开修复，训练素材里手部模型较少，手部可能会崩坏。</li></ul><ul><li>LoRA of Hanfu&amp;Tang style Photography/chinese clothes,</li><li> weight ：0.4-0.8, </li><li>model：realistic model. The base model used in the example image is majicMIX realistic majicMX-v7, </li><li>AD: turn on or turn off is not significant,</li><li>The model itself can control outline of facial features, but it is best to use ad-hand. There are few close up of hand in the training materials, so the generation may be wrong.</li></ul>

Files with SHA256: 682ea4d0c13b42b546c5a4e5e3367c06d0afe94dfc88206ee713b4ed5fa533c2