S:AI能取代设计师么?
I :至少在设计行业,目前AI扮演的主要角色还是超级工具,要顶替?除非甲方对设计效果无所畏惧~~
预先学习:
- 安装及其问题解决参考:《Windows安装Stable Diffusion WebUI及问题解决记录》;
- 运行使用时问题《Windows使用Stable Diffusion时遇到的各种问题整理》;
- 模型运用及参数《Stable Diffusion 个人推荐的各种模型及设置参数、扩展应用等合集》;
- 提示词生图咒语《AI绘图提示词/咒语/词缀/关键词使用指南(Stable Diffusion Prompt 设计师操作手册)》;
- 不同类的模型Models说明《解析不同种类的StableDiffusion模型Models》;
- 绘制人物动作及手脚细节《Stable Diffusion 准确绘制人物动作及手脚细节(需ControlNet扩展)》;
- 各种风格对比及实际运用《AI绘图风格对照表/画风样稿详细研究记录及经验总结》;
stable diffusion就只能小姐姐么?不,今天我们用stable diffusion对三种建筑方面的图片进行修改。
一、渲染室内设计线稿图
成图目标
网上弄来一张线稿,要求贴色贴材质,尽量追求与线稿中物体接近。
物品一目了然,就不多说了。
线稿导入controlnet
边缘检测权重1
线条检测权重0.9
深度检测权重0.8,可以不用
这张图里如果没有边缘检测,左下角的电视机柜往往出现一堆书
提示词及渲染出图
- 尽量将每个位置的内容都提到,这样才可能尽可能相似;
- 注意提示词相关性设置得可以高一点,比如这里是
22
((photorealistic)), 8k, clean room, sunlight, white ceiling, blue curtains, red sofa, yellow cabinet, wooden bed, Persian carpet under bed, bed end bench, wooden floor,
lower left corner is low cabinet, upper left corner is wall, upper right corner is windows and tatami, the wall above the head of the bed is made of wood, desk lamp on the nightstand to the right of the bed
Negative prompt: low res,text,error,cropped,worst quality,low quality,normal quality,jpeg artifacts,signature,watermark,username,blurry,text font ui,futa,yaoi,
Steps: 20, Sampler: Euler a, CFG scale: 22,
出图基本符合设计稿结构,小修改用图生图里的局部重绘,这里不再赘述,可参考下面景观调整操作。
二、调整景观设计手绘稿部分元素
成图目标
根据要求修改效果图中部分元素,加入人物和水中活物
修改元素
用局部重绘方式增加成年人元素,注意:
- 水彩风格前先使用物体边缘线条,否则可能与原图不符
- 同时
重绘区域
选择仅蒙版
,这样可以让人物不会画坏了
((object edge line)),((watercolor painting)),1male walking on flagstone path, green grass, blue pond, full body,Negative prompt: ((nsfw)),((ugly))),(((duplicate))),((morbid)),((mutilated)),(((tranny))),(((trans))),((trannsexual))),(((mutation))),(((deformed)))(((long neck))),((bad anatomy))(((bad proportions))),(((extra arms))),(((extra legs))), (((disfigured))),((more than 2 nipples))),malformed,mutated,(hermaphrodite),((extra limbs)).((missing arms)).((missing legs)).((poorly drawn hands))((poorty drawn face)),(mutation,poorly drawn :1.2),(long body :1.3).multiple breasts,cloned face,gross proportions, mutated hands,bad hands,bad feet,long neck,missing limb,(malformed limbs),malformed hands,(fused fingers).(too many fingers),extra fingers,missing fingers,extra digit,fewer digits,(mutated hands and fingers :1.5),low
其它提示词不变,修改为小孩,注意这个蒙版不能太小,否则小孩动作施展不开容易缺胳膊少腿。
((object edge line)),((watercolor painting)),1child walking on flagstone path, green grass, blue pond, full body,
增加鱼
((object edge line)),((watercolor painting)),kois in blue pond,
出图修补
基本算定型了,小修改和上面操作一样,不赘述。
大家有没有发现,类似这种风格的稿子,如果是在绘图软件里做的,其实让原作者直接修改来得更快更准确。
三、改建筑照片为线稿
这是用PS修改的线稿图,可以参考一下:《用PS将一张照片调成简笔画风格》
选择照片
把线条风格先提示了。
((a line drawing, line art, line work)), simple line, black AND white, buildings, street, trees, cars, street light
negative prompt: colorful
图生图
先将一张空白图片放入图生图
将照片放入controlnet中,使用相关模式生成线条,注意优先选择controlnet is more important
测试。
模型可用lineart
、canny
、mlsd
、depth
等,但具体用哪个需要自己试验,不同的图片最终效果不同,比如此图用lineart_realistic
找建筑线条就比用canny
描绘更准确和精炼,canny
描绘出细节太多,按照线稿要求看就太杂了。
线稿对比
对比一下不同重绘幅度的区别,这种模式下,0.82~0.90
最合适。
下面各张对比图的重绘幅度、种子、模型、提示词等都有略微调整,画面按照由浓到淡排列。
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Negative prompt: colorful
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 199447306, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.9, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: canny, ControlNet-0 Model: control_v11p_sd15_canny [d14c016b], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black line, white background, high contrast, buildings, street, trees, cars, street light
Negative prompt: colorful, gray
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 847333786, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.9, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Negative prompt: colorful
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 1879644544, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.85, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black line, sketch, white background, high contrast, buildings, street, trees, cars, street light
Negative prompt: color
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 1607122200, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.84, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black line, white background, high contrast, buildings, street, trees, cars, street light
Negative prompt: colorful, gray
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 2330278839, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.84, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black line, white background, high contrast, buildings, street, trees, cars, (white ground),
Negative prompt: colorful, gray,
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 3402894627, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.84, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black line, white background, high contrast, buildings, street, trees, cars, street light
Negative prompt: colorful, gray
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 167399156, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.84, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Steps: 64, Sampler: Euler a, CFG scale: 22, Seed: 1994739401, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.8, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_lineart [43d4be0d], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Negative prompt: colorful, gray
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 4145571130, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.85, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: mlsd, ControlNet-0 Model: control_v11p_sd15_mlsd [aca30ff0], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Negative prompt: colorful, gray
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 3008838972, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.85, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: depth_zoe, ControlNet-0 Model: control_v11f1p_sd15_depth [cfd03158], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
下面两张由ai自己发挥,仅提示词配了线条画词缀,出图相比用了controlnet要更线稿。
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Negative prompt: colorful
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 178003217, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.85, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: depth_zoe, ControlNet-0 Model: coadapter-depth-sd15v1 [93aff3ab], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
a line drawing, line art, line work, black and white, buildings, street, trees, cars, street light
Negative prompt: colorful
Steps: 64, Sampler: Euler a, CFG scale: 7, Seed: 3853790227, Size: 1024x512, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae, Denoising strength: 0.9, Clip skip: 2, ENSD: 31337, ControlNet-0 Enabled: True, ControlNet-0 Module: lineart_realistic, ControlNet-0 Model: control_v11p_sd15_canny [d14c016b], ControlNet-0 Weight: 1, ControlNet-0 Guidance Start: 0, ControlNet-0 Guidance End: 1
最后,大家学习过这篇内容了么?AI绘图实战(六):制作一张庆祝五一劳动节的海报 | Stable Diffusion成为设计师生产力工具