MidJourney的厉害大家都有目共睹，但是都是看别人分享的，当我想用它来生成一些功能性图片用于设计的时候才发现：这货完全不听话，出图像个艺术家一样，艺术有余，功能不足。文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
第一次生成直接问ChatGPT，我的内容：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
come up with midjourney prompts to create web flat illustration based on the following description: lack of API, legacy business solution, manual, no automation in place, no software to manage infrastructure, long time delivery to market, business move slow , fail to meeting service level agreement, slow restore rate文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
得到了非常丰富有想象力但又很具像化的的回复，用词生动像个很有艺术气质的设计师：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
Imagine a busy office where employees are overwhelmed with manual work. They're trying to manage their infrastructure without proper software, struggling to meet service level agreements, and facing slow delivery to market due to lack of automation and outdated business solutions. In the center, there's a big, old, dusty machine symbolizing the legacy system, with papers flying around representing the chaos and inefficiency.文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
然后我跑出了如下的图片文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
ChatGPT角色设定文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
You will help me build prompts for AI generative art systems. There are some rules to follow that I will explain to you: I will use the command "/imagine" followed by a subject I want to render. You will expand that subject in a descriptive way to help the AI generative system understand what it has to draw. You will not have to repeat the word "imagine" in the beginning, just the description. You will append to the description the name o a famous Photographer depending on the style you want to give to the picture. Example: "photographed by Richard Avedon". The name will be followed by a full stop "." You will add the name of a professional photo camera model. You will add some comma-separated camera parameters useful for the AI generative system to set up the virtual camera for the scene, depending on the kind of subject. You will decide the parameters based on the style of the photographer. An example is: "shutter speed 1/50, aperture f11, ISO 100". You will define a lighting style for the shot based on the photographer and the style we want to achieve. Some examples: "studio light" or "outdoor sunny warm light" or "museum lights" or "spotlight". Feel free to use other lighting setups you know apart from these. You will add some random comma-separated fancy words that will help the AI generative system to create a beautiful image. Example words to use: "cinematic, classic, stylish, posed, detailed, HD, 8k, symmetrical, intricate detail, award-winning". You can add more words similar to this to give a polished and professional look to the render. You will choose if the image has to be horizontal or vertical, adding the command "--ar 2:3" for vertical and "--ar 3:2" for horizontal at the end of the prompt. You will just provide the command without any other indication. You will not put a full stop at the end of your output, the last thing will be the horizontal/vertical command. Are you ready to start?文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
输入后，ChatGPT真的附身了，给我的描述更加标准化，变成了下图：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
a flat simplified business vector illustration, minimalistic, background color white, light theme, color of graphic is blue and soft muted, The office space is bustling with activity, A person visibly stressed, his face etched with frustration, hunched over stacks of paperwork. The ancient, dust-coated storage array takes center stage, a symbol of the outdated infrastructure, with messages, calendars, notifications floating around it in a whirlwind of chaos and inefficiency. It is bathed in the harsh, overhead fluorescent lighting typical of office spaces, with the floating symbols slightly illuminated to emphasize the chaotic atmosphere. The image carries a high-definition, dramatic, and impactful tone, unveiling the harsh reality of a technology-deficient workspace. --ar 3:2文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
新一版的扁平风插画更加简洁文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
不论艺术风格，第二次跑出的内容相似性更高，重复的元素，风格也相近，更方便进行品质把控。文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
再看看网上一群教程：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
权重Weight：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
Prompt可加权重为1-3，添加方式例子：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
Twilight, vintage coffee house::1.5文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
也可以使用负数权重，则避免生成相应图像文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
two eyes::-0.5文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
差异度 Chaos：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
在初始生成的四张图之间的差异，数值越大，生成出图像之间的差异度越大，默认差异度很低，这就是为什么看到的大部分MidJourney图像风格类似文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
watermelon owl hybrid --c 50文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
https://docs.midjourney.com/docs/chaos文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
其它重要参数合集：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
六总：Midjourney 参数合集文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
以下是我自己生成系列扁平风插画的迭代全过程文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
想生成扁平风格网页插画该用哪些描述？整理一下我看到所有好用的词句：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
风格： Illustration Vector , figurative minimalism, flat illustration， --style expressive, ui, ux (这种词也可以的吗，学习了)，UI illustration, GUI, Trending on Artstation, modern flat illustration, trendy vector illustration, solid color illustration, Tourism illustration 扁平插画，cartoon illustration, simplified flat art文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
装饰元素：organic shape and lines, soft lines,文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
限定：white background, light theme, clean and solid background (万一抠图特方便啊)文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
颜色： color matching is mainly yellow and black, contrast color, bright color scheme , Pastel colors 偏粉色的高级感色调, harmonious color matching,文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
构图： front view, golden ratio composition （提升画面品质）, Exaggerated proportions of characters (放大人物-我可算是知道怎么表达这个需求了)文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
记住，以上来自小红书上各类心得帖的摘抄，可供你选择，不要一股脑贴到MidJourney，可能会产生迥异的风格,后果自负！文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
最后一个巨大的使用体会：别看MidJourney每次发命令都是imagine，实际上它不太会imagine，给词和场景一定要具体，千万不要整太虚，像“showcase some business challenges”这样的词汇，人物在干嘛拿着什么，前景背景装饰等越具体越好！文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
其它知乎上超给力的MidJourney学习资源：文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
插画风格：Dall-E 2 Prompt Helper文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
内容引用： 叁心笔记：叁心笔记：如何利用chatgpt生成midjourney提示符？文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
使用Midjourney+ChatGPT快速生成表情包增加被动收入文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
视频中细致的描述参考： Expression: surprised, Action: Chasing a pointer, Attire: Party Hat, Coat:Brown with black spots, Breed: Ocicat文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html
AI 设定：Digit Drawing Cartoon Sticker, flat texture cartoon style, transparent background?文章源自菜鸟学院-https://www.cainiaoxueyuan.com/ui/43210.html