This lora is trained on a selected group of Chinese tradional portraits, mostly royal portraits and ancestral portraits from Ming and Qing dynasty. I made it only to test AI, and to commemorate traditional portrait artisans in China history who disappeared a century ago after photography and hyperrealistic charcoal painting went popular, they remain quite nameless today.

本lora使用中国传统肖像画中的特定一类——帝皇帝后像/祖先像——进行训练。“祖先画像是中国人物画中的一个重要分支，也是一个具有社会学意义的特殊画种，包括由宫廷画家或专业画师为皇室贵族祖先绘制、用于皇家祭祀并具有政权象征意义的先皇、先后肖像绘画，由民间画工绘画、以百姓已故先人为对象的人物画像。”主要用于“庆贺福寿”和“宗族祭祀”。所选画像大部分属于明清时期的“大影”（全身个人像，又称“喜神像”“神祇”“福样”等）。

我训练这个lora仅仅是为了测试AI的学习能力，以及纪念那些籍籍无名的民间写真画匠们，由于社会的变化、碳精画和摄影术等新技术的传入，他们在上世纪初离开了历史舞台。

What to expect: front view portrait, full body or upper body, sitting, very formal, clothes in Chinese traditional style(details inaccurate though).

一般效果：生成正面坐像，半身和全身皆可，风格端庄严肃，中式传统服装（但不准确）。

Known problem: bad hand(I guess it's because original portraits seldom paint whole hand, only fingertips). results unstable when used with anime-style model.

已知问题：手很糟糕（可能因为这类肖像极少画完整的手……经常只画几根手指尖）。跟二次元模型搭配效果不稳定。

tags you can try （一些tag）:

xiaoxiang

mingstyle ----(Ming dynasty style 明朝风格)

qingstyle ----(Qing dynasty style 清朝风格)

crown/elaborate jewelry crown ----(a kind of female headwear 大概能出凤冠)

qing dynasty official hat ----(清朝帽子)

black hat/headwear/brimless hat ----(tend to get some Ming style hat 比较容易出明朝的帽子)

Chinese clothes ----(中式服装)

Chinese dragon pattern ----(龙纹)

References参考文献:

[1] 吴卫鸣. 《明清祖先像图式研究》，社会科学文献出版社，2020.

[2] Jan Stuart,Evelyn Sakakida Rawski. Worshiping the Ancestors.Chinese Commemorative Paintings. Stanford University Press, 2001.