依然是用一个全身图片训练的，更换姿势还是略有一定的难度，使劲加权重。人物提示词看情况和权重配合使用能尽量改变动作，Marie Therese+0.2都行。发饰原因，AI概率画成双马尾。Google Translate:It is still training with only one full-body picture, and it is still a little difficult to change the posture, so increase the weight.Character prompts can be used in conjunction with weights to change actions depending on the situation, Marie Therese+0.2 is also possible.<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/36d556bb-f918-4c2f-0c28-f231a8062300/width=525/36d556bb-f918-4c2f-0c28-f231a8062300" />

Banner of the Maid - Marie Therese / 圣女战旗 - 玛丽·特蕾兹

我发现这角色脑袋下面都是对称的，就镜像了几张图尝试抵消那一成不变的动作。Google Translate:I found that the character below the head is symmetrical, so I mirrored a few pictures to try to offset the unchanging movements.

Files with SHA256: 1d50109e3cfd4845c59fb96145aefe063f3793e32053f7e508f741e4f30099a4