墨卣的SDXL lora版本,训练素材来自被誉为”东方的米开朗基罗"的名家,刘继卣大师的作品
“刘继卣画的人物仕女、花鸟走兽,多为工笔与写意相结合,在其绚丽多彩和准确造型的的基础上,显得格外凝重、奔放、潇洒、传神。他的作品把西洋画的情调渗透进中国画的意境中,却丝毫不露痕迹,形成自己独特风格。”
水墨笔触风格明显,同时也融合了部分西方油画美学,显现出一种独特的油墨质感。
没有绝对推荐的生成参数,大部分情况都可以呈现水墨质感:
底模:SDXL1.0
起手关键词:Chinese ink painting, traditional media, liujiyou (在SDXL底模下可以很好的调用水墨画的原始素材)
人物内容:1 girl 服装:hanfu/cheongsam/chinese armor/strapless dress
CFG:7 sampling step:40
分辨率:1024*1536
这个版本的lora得益于SDXL1.0版本强大的泛化能力,可以将大师的水墨笔触特点表现在各种题材上,甚至超出了我训练集的内容,SDXL真的未来可期!希望大家尽可能尝试更多的国风题材,绘制更多展现东方艺术之美的作品~
擅长人物,但是得益于SDXL优秀的泛化性,场景和动物也不在话下
English:
The SDXL lora version of Moyou, the training material comes from the work of Master Liu Jiyou, who is known as the "Michelangelo of the East".
Most of the characters, flowers, birds and animals in Liu Jiyou's paintings are a combination of meticulous brushwork and freehand brushwork. On the basis of their colorful and accurate modeling, they appear to be particularly dignified, unrestrained, unrestrained and vivid.
His works infiltrate the sentiment of Western painting into the artistic conception of Chinese painting, but show no trace, forming their own unique style. "
Ink stroke style is obvious, but also the integration of some Western oil painting aesthetics, showing a unique ink texture.
There are no absolutely recommended generation parameters, and ink texture can be presented in most cases:
Bottom die: SDXL1.0.
Starting keywords: Chinese ink painting, traditional media, liujiyou (the original material of ink painting can be used well under the SDXL bottom mold).
Character content: 1 girl clothing: hanfu/cheongsam/chinese armor/strapless dress.
CFG:7 sampling step:40.
Resolution: 102401536.
This version of lora benefits from the strong generalization ability of the SDXL1.0 version, which can show the characteristics of the master's ink strokes on a variety of topics, even beyond the content of my training set, SDXL is really promising in the future!
I hope you can try as many national themes as possible and draw more works that show the beauty of oriental art.
Good at characters, but thanks to SDXL's excellent generalization, scenes and animals are no problem.