重新打了关键词，概念也重新做了分类，这次可以生成更接近实物造型且好看的凤冠。同时服装和面部妆造也更为稳定。
Redo the tag of all images and made new concept. improve the looking of the headdress. also costume and face makeup will be more stable then old one.
背后旗子的问题还是没有太大的提高，大家可以试试使用不同的提示词
the backing flag is still not good, but you can still try it with prompt
在负面词里加入flowers 避免头上的绒球变成花的样子
add “flowers“ to your negative prompt to avoid balls on headdress become flowers.
迭代步数控制在30左右，但是偶尔也会出现色块
please keep Sampling steps around 30, it may generate some small color blocks on face
测试了一些2.5d和2d模型 效果也还可以
tested on some 2.5d and 2d style model, emmm.....it works, but only works, you can have try
重新打了关键词，概念也重新做了分类，这次可以生成更接近实物造型且好看的凤冠。同时服装和面部妆造也更为稳定。
Redo the tag of all images and made new concept. improve the looking of the headdress. also costume and face makeup will be more stable then old one.
背后旗子的问题还是没有太大的提高，大家可以试试使用不同的提示词
the backing flag is still not good, but you can still try it with prompt
在负面词里加入flowers 避免头上的绒球变成花的样子
add “flowers“ to your negative prompt to avoid balls on headdress become flowers.
迭代步数控制在30左右，但是偶尔也会出现色块
please keep Sampling steps around 30, it may generate some small color blocks on face
测试了一些2.5d和2d模型 效果也还可以
tested on some 2.5d and 2d style model, emmm.....it works, but only works, you can have try