当下,AI作画可以说相当火爆,已经有很多人利用stable diffusion之类的AI绘画工具创作出惊艳的作品。
我们知道,Stable diffusion是一个基于Latent Diffusion Models(潜在扩散模型,LDMs)的文图生成(text-to-image)模型。 Stable Diffusion 作为一款高性能模型,它生成的图像质量更高、运行速度更快、消耗的资源以及内存占用更小,也是备受很多创作者喜爱的原因。
不过也有很多人却在stable diffusion关键词tag怎么写犯难了。首先大家需要先明白Stable diffusion关键词的基本写法套路,才能根据自的想法写出合适的关键词,这也是创作出比较理想绘画作品的关键。
stable diffusion支持使用自然语言,关键词tag和表情emoji等方式来写关键词,下面一起来看看几种关键词常见的写法:
1.关键词分隔:使用英文逗号 , 可以分隔不同的关键词tag。除此之外,空格和换行等不影响tag分隔;
案例:1girl,loli,long hair,low twintails(1个女孩,loli,长发,低双马尾)
2.混合:WebUi 使用 | 分隔多个关键词,实现混合多个要素,注意混合是同等比例混合,同时混。
ex:1girl,red|blue hair, long hair(1个女孩,红色与蓝色头发混合,长发)
3.增强/减弱:有两种写法
第一种 (提示词:权重数值):数值从0.1~100,默认状态是1,低于1就是减弱,大于1就是加强
ex: ,(loli:1.21),(one girl:1.21),(cat ears:1.1),(flower hairpin:0.9)
第二种 (((提示词))),每套一层()括号增强1.1倍,每套一层[]减弱1.1倍。
ex: ((loli)),((one girl)),(cat ears),[flower hairpin]
上面两种写法是一样的效果,但是第一种写法更清晰。所以在关键词较多时,用第一种写法更清晰明了,方便阅读。
4.分步描写: 工具生成图片时会根据关键词进行步数渲染,可以使用 [关键词1:关键词2:数字]这种方式来选择我们的关键词从第几步开始渲染。可以理解为先按某种关键词生成,然后再此基础上向某个方向变化。
数字大于1理解为第X步前为关键词1,第X步后变成关键词2,数字小于1理解为总步数的百分之X前为关键词1,之后变成关键词2
ex:a girl with very long [white:yellow:16] hair
以上就是stable diffusion关键词的基础写法。除此之外,大家还有注意关键词书写要点:
- 不是越长的提示词生成的图片越好,尽量将关键词控制在75个(100个)以内。
- 越关键的词,越往前放
- 相似的同类,放在一起。
- 只写必要的关键词。
以下是关于stable diffusion一些美女主题先关关键词案例:
选用的都是Lexica Aperture v2模型:
Facial portrait. Nurgle, looking at the camera, disgusted. fear inspiring mood, intimidating, extremely detailed painting. by Greg Rutkowski and by Henry Justice Ford and by Steve Henderson.
Es bell�sima, de mirada penetrante, sus ojos amarillos misteriosos hipnotizan, cejas gruesas, labios gruesos, cabello corto, piel muy blanca
body portrait of a gorgeous ghost princess, wearing torn clothes, by terry o'neill intricate, elegant, highly detailed, digital painting, glistening skin, artstation, concept art, smooth, sharp focus, contrasting lighting, bright colors, dark background, illustration, art by artgerm and greg rutkowski and alphonse mucha,8k
Mulher, cute, eyes, full body, biqu�ni
a beautiful hot girl
portrait of a female pinup model in love in pakistan forest lake wearing a golden Swimsuit, year1977with hallucinatory experience, lush magic plants, armani style : : by martine johanna and simon stalenhag and chie yoshii and casey weldon and wlop : : ornate, dynamic, particulate, rich colors, intricate, elegant, highly detailed, the vogue, harper's bazaar art, fashion magazine, smooth, sharp focus,8k, octane render, style of sports illustrated
analog style, analog style, delicate artist's studio model, matte digital painting. partially with red linen, delicate body, soft skin, soft and delicate lighting, ethereal, .smooth, natural skin, soft glow, sharp focus, illustration, art by Stanley Lau by Artgerm and WLOP
""Artgerm, WLOP, Greg Rutkowski; ugly hideous features lady,
cyborg women with computer, a character side portrait, realistic, full body shot, sharp focus, elegant
Neekolul, highly detailed, digital painting, artstation, concept art, sharp focus, illustration, cinematic lighting, art by artgerm and greg rutkowski and alphonse mucha
high angle photo of a gorgeous young woman in the style of stefan kostic, realistic skin texture, black sweatshirt,1/2body crop,85mm art lens, f1.2, sharp focus,8k high definition, insanely detailed, intricate, elegant, art by stanley lau and artgerm
ultra realistic illustration, a stunningly beautiful greek goddess of chaos played by Adriana Lima and and Anushka Shankar and Sara Sampaio , intricate, elegant, highly detailed, digital painting, artstation, concept art, smooth, sharp focus, illustration, art by artgerm and greg rutkowski and alphonse mucha
portrait of a sexy model, super realistic, showing herself
beautiful robot woman , gray skin , cyan glowing eyes , big cyan lips
回复