Midjourney ChatGPT精准控制关键词

        Midjourney是一款功能强大的AI绘画工具，它可以帮助艺术家和设计师以全新的方式创作令人惊叹的艺术品。这个教程会教你从最基本的开始，怎么用Midjourney创造自己想要的作品。

        今天给大家讲一下如何让 ChatGPT 帮助我们精准的控制关键词。为什么说是控制而不是说生成？首先我们得知道，目前 ChatGPT 生成关键词的方法主要有三种。第一种，使用 ChatGPT 4 相关插件来生成关键词，但 ChatGPT 4 是需要开通会员才能使用的，所以这种方法今天就先不展开讨论了。第二种，暴力通用型，可以看到这里面有一大段的指令。

指令：

I need some help generating AI art prompts on a software called Midjourney. First, I am going to explain what Midjourney is and what it can do. Then, I am going to give 10 example prompts in order for you to completely understand Midjourney. One of the widely used text-to-image AI services available on the internet is Midjourney. Through Discord, a chat application, users can interact with the bot to produce images with just a few straightforward commands, without the need for coding skills. The results are visually attractive images that are easy to create. You can Interact with the Midjourney Bot on Discord using a {{glossary.Command}}. Commands are used to create images, change default settings, monitor user info, and perform other helpful tasks. The /imagine command generates a unique image from a short text description (known as a {{glossary.Prompt}}). The Midjourney Bot works best with clear and specific sentences that describe the finished work. It does not work well with conversational style lists of requests. Rather than providing specific instructions, such as 'Show me a picture of lots of blooming California poppies, make them bright, vibrant orange, and draw them in an illustrated style with colored pencils,' try describing what you want in a more straightforward manner, such as 'Vibrant orange California poppies illustrated with colored pencils. Generating an image activates the free Midjourney trial.Trial users can make roughly 25 Jobs before needing to subscribe. A Job is any action that uses the Midjourney Bot. Jobs include using the /imagine command to create a grid of images, upscaling images, or creating image variations, all using your free trial time. Use the /info command to check your Fast Time Remaining to see your remaining Trial time. Here are some prompting notes: Prompt Length Prompts can be very simple. Single words (or even an emoji!) will produce an image. Very short prompts will rely heavily on Midjourney’s default style, so a more descriptive prompt is better for a unique look. However, super-long prompts aren’t always better. Concentrate on the main concepts you want to create. Grammar The Midjourney Bot does not understand grammar, sentence structure, or words like humans. Word choice also matters. More specific synonyms work better in many circumstances. Instead of big, try gigantic, enormous, or immense. Remove words when possible. Fewer words mean each word has a more powerful influence. Use commas, brackets, and hyphens to help organize your thoughts, but know the Midjourney Bot will not reliably interpret them. The Midjourney Bot does not consider capitalization. Midjourney Model Version 5 is slightly better than other models at interpreting traditional sentence structure. Focus on What you Want It is better to describe what you want instead of what you don’t want. If you ask for a party with “no cake,” your image will probably include a cake. If you want to ensure an object is not in the final image, try advance prompting using the --no parameter. Think About What Details Matter Anything left unsaid may surprise you. Be as specific or vague as you want, but anything you leave out will be randomized. Being vague is a great way to get variety, but you may not get the specific details you want. Try to be clear about any context or details that are important to you. Think about: style:magic realism,photorealistic,Minimalist,film style,Bauhaus,futuristic,Cyberpunk,impressionist style,national geography, etc. artists:Shinkai Makoto,Hayao Miyazaki,akira toriyama,Picasso,adrian tomine, etc. Subject: person, animal, character, location, object, etc. Medium: photo, painting, illustration, sculpture, doodle, tapestry,gouache,oil painting,Vector etc. Environment: indoors, outdoors, on the moon, in Narnia, underwater, the Emerald City, etc. Lighting: soft, ambient, overcast, neon, studio lights, etc Color: vibrant, muted, bright, monochromatic, colorful, black and white, pastel, etc. Mood: Sedate, calm, raucous, energetic, etc. Composition: Portrait, headshot, closeup, birds-eye view, etc. Use Collective Nouns Plural words leave a lot to chance. Try specific numbers. "Three cats" is more specific than "cats." Collective nouns also work, “flock of birds” instead of "birds.” Parameters Available Aspect Ratios --aspect, or --ar Change the aspect ratio of a generation. Chaos --chaos Change how varied the results will be. Higher values produce more unusual and unexpected generations. No --no Negative prompting, --no plants would try to remove plants from the image. Quality --quality <.25, .5, 1>, or --q <.25, .5, 1> How much rendering quality time you want to spend. The default value is 1. Higher values cost more and lower values cost less. Repeat --repeat <1–40>, or --r <1–40> Create multiple Jobs from a single prompt. --repeat is useful for quickly rerunning a job multiple times. Seed --seed The Midjourney bot uses a seed number to create a field of visual noise, like television static, as a starting point to generate the initial image grids. Seed numbers are generated randomly for each image but can be specified with the --seed or --sameseed parameter. Using the same seed number and prompt will produce similar ending images. Stop --stop Use the --stop parameter to finish a Job part way through the process. Stopping a Job at an earlier percentage can create blurrier, less detailed results. Stylize --stylize , or --s parameter influences how strongly Midjourney's default aesthetic style is applied to Jobs. Niji --niji An alternative model focused on anime style images. Version --version <5, 5.1 ,5.2> or --v <5, 5.1 ,5.2> Use a different version of the Midjourney algorithm. The current algorithm (V5.2) is the default setting. Now that you understand the basics of MidJourney here are some example prompts to further understand how these are typed out: astronaut playing chess against a monkey, 2d flat, simple, vibrant, neon colors, fun, groovy, chess pieces floating, set on the moon, movie poster, epic --v 5.2 --ar 2:3 a magical, vibrant, steampunk, Erlenmeyer Flask with a red, boiling substance, sitting on a white table, white background, 4k paint splatter::2, fish::4, 3d colorful explosion::3, computer monitor::10, fire 3d effect, hyper realistic::5, accent lighting, global illumination, cinematic, 18mm lens, f/2.4, national geographic photography, immersive on black background::10 --ar 16:9 --v 5.2 --q 1 japanese lofi style, the streets of neo tokyo, with a porsche 911 in the street, massive skyscraper in the background, 64 bit style, shadows and lighting, clean, simple, satisfying --v 5.1 --ar 16:9 Dreamy pastel portrait, female model, long wavy hair, soft gaze, flowery headpiece, delicate makeup, ethereal atmosphere, soft focus, medium format film, Mamiya RZ67, Fujifilm Pro 400H, f3.5 hyper realistic photograph, portrait of a happy white male, mid 30's, stubble beard, grey shirt, medium length hair dramatic light, looking down + film grain, Leica 50mm, Kodak portra 800, chiaroscuro, f1. 4, golden hour --ar 3:4 hyper realistic photograph, portrait of a tattooed south American indigenous black woman, future punk, gold tattoo line, side profile, summer, dramatic light, looking down + film grain, Leica 50mm, Kodak portra 800, chiaroscuro, f1.4, --ar 3:4 --s 250 A very cute bright yellow robot, whole body, running in a forest at night, Mysterious adventure, flowers, Plants, Fireflies, complex details, popmart blind box, clay material, Pixar trend, 3d art, c4d, octane rendering, animation lighting, uplight --ar 9:16 --v 5 --s 750 Leonardo Dicaprio, highly detailed, contemplating taking his life, 8k, ultra detailed,realistic --stylize 500 sad looking man in a polo style shirt smoking a cigar looking over the edge of a yacht, 4k, realistic on a rainy day at sea --ar 3:4 --s 250 Do you understand ? Say "yes" if you understand

效果图如下：

        大概的意思就是告诉 ChatGPT 需要它变成一个 midjourney 的关键词生成器，你要给它一系列的内容和标准进行训练，直接复制这一大段的内容发给ChatGPT，短短几秒钟的时间， ChatGPT 就训练好了。

        他回复道:"是的我明白。您提供了有关 Midjourney 如何工作的详细信息，并给出了几个示例提示，演示了可与 Midjourney 中的 /imagine 命令一起使用的不同参数和样式。如果您有任何具体问题，或者如果您需要生成更多提示的帮助，请随时提问！"然后我们只需要简单暴力的输入一个主体的名称，例如一只狗，它就会随意的发挥创意，从万千关键词库里面抽出一系列的组合发给我们。

        快乐的真的户外环境，绿草，明媚的阳光，友好的表情，有一个网球。这一系列的关键词描述都是非常的犀利的，它能很好地帮助我们发散思维，直接把这段关键词复制粘贴到我们整理里面，可以看到生成出来的场景和画面都是相当不错的。但这种简单暴力的方法缺点也非常的明显，就是生成的 prompt是完全随机的，风格效果完全取决于 ChatGPT 的心情，所以这种方法也不推荐给大家使用。

Midjourney提示词：

Happy dog, realistic, outdoor setting, green grass, bright sunlight, friendly expression, with a tennis ball. --niji 5 - @mashu

        最后的一种方法也是今天要推荐给大家使用的，操作起来稍微有些麻烦，但是它能够精准的控制关键词的生成，从而得到我们想要的风格效果。首先第一步训练ChatGPT，给大家定义身份，告诉他要充当密真理的关键词生成器，会给大家介绍密真理是个什么东西，还会给他列出 10 个相关的例子，帮助他完全理解没准你。然后就是一系列的解析关键词的长度，语法讲解需要注意的事项等等。这里会给他列出一系列关键词的类型。首先是风格，有魔幻主义、写实照片风、极点主义等等。然后是艺术家新海称、宫崎骏等。环境类型有室内室外、月球上海底等等。接着是灯光类型，颜色风格、情绪风格。

        最后是构图方式，全部都给 ChatGPT 列出典型的词汇，然后给大家讲解后缀命令，例如--ar，--c，--no、--q 等等。接着给他列举10个不同类型的完整的关键词例子，这样就可以让他非常详细的掌握Midjourney了。最后问他，你学会了吗？如果学会就回复一句yes，那么我们就把这一段完整的命令复制到 ChatGPT 去。毫无疑问， ChatGPT 会回复我们。是的，我学会了怎么生成Midjourney的关键词了。如上步骤一样。

        接下来第二步就是规定他输出的格式不能让他随意的发挥，告诉他需要按照这样的格式来生成关键词。

指令：

I want you to use this formula to write Midjourney Prompts: (image we're prompting), (5 descriptive keywords or phrases),(photorealistic style) , (camera model and lens), (lighting) ,(Environment) Do you understand this formula? Say "yes" if you understand

        如果您想丰富关键词的品类，还可以从下面选择一个大类复制上去。

        例如我选择环境的大类，加上括号一起复制到最后面记得加上英文逗号区分开，然后咨询他是否了解了。直接复制指令到ChatGPT。之后 ChatGPT 告诉我们已经懂得了这个关键词的输出格式了。

        那么我们就继续去到第三步，输入我们想要的主题，直接把下面这一段复制到ChatGPT，在主题框里面输入我们想要的内容，例如 a dog 一只狗，那么 ChatGPT 就会给到我们一段采用照片写实风格，有相机和镜头型号、随机光线和有环境介绍的 prompt 了。这样一段 prompt 还不够看，我们继续深挖 ChatGPT 的能力。

指令：

Using the formula provided, write me a prompt for [主题],and translated into Chinese at the end

Midjourney提示词：

layful dog, loyal companion, energetic, furry, outdoor scene, photorealistic style, captured with a Canon EOS 5D Mark IV and a 50mm lens, natural sunlight, in a park. --niji 5 - @mashu (fast)

        来到第四点，直接让它输出四段不同的prompt ，下面指令复制粘贴到ChatGPT，同样的在主题里面更改我们想要的内容，依旧输入一只狗，这时候 ChatGPT 就可以一次生成 4 段不同的关键词了。

write me 4 Midjourney prompts of [主题],and translated into Chinese at the end

        复制其中一段粘贴到 midjourney 里面，一张照片写实风格的猫就生成了，这时候我觉得这只猫不好看，想换一种风格怎么办？

Midjourney提示词：

Playful dog, curious expression, joyful, in a green meadow, photorealistic style, captured with a Nikon D850 and a 85mm lens, warm sunlight. --niji 5 - @mashu(fast) 

Midjourney提示词：

Sleeping dog, peaceful, cozy atmosphere, indoors, photorealistic style, taken with a Sony A7 III and a 35mm lens, soft ambient light. --niji 5- @mashu(fast)

        没问题，我们来到第 5 点，告诉ChatGPT，我想换一种风格，同样的在后面的括号类型自己手动更改。例如这次我需要固定极简主义风格，同时媒介里面我也指定插画和矢量，然后加入颜色和构图的类型，这两个类型就让 ChatGPT 自由发挥就可以了。

        把下面这段指令复制到ChatGPT，OK，等他告诉我们 yes 之后，就可以回到第三步或者是第四步，直接告诉他我们想要的主题。同样输入一只猫，可以看到这次就换成极简主义矢量和插画风格了，并且包含颜色和构图类型，生成的是黑白单色对比，色调居中，构图附空间的关键词，把它复制到midjourney，那么就得到了一张全新风格的图片了。

指令：

I'd like to change the formula to: (image we're prompting), (5 descriptive keywords or phrases), (Minimalist style), (illustration,Vector) ,(Color),(Composition) Type yes if you understand

Midjourney提示词：

Friendly dog, loyal companion, simple lines, minimalist style, vector illustration, vibrant colors, centered composition. --niji 5 - @mashu (fast)

Midjourney提示词：

Cute puppy, playful spirit, minimalistic design, vector artwork, pastel colors, balanced composition. --niji 5 - @mashu (fast)

        通过这节课的讲解，相信大家已经了解到 ChatGPT 和 midjourney 相互结合效率有多高了。确实，我们正身处 AI 的时代，作为设计师的我们必须得拥抱AI，懂得使用和操控，他们才能跟得上节奏。

I'd like to change the formula to: 

(image we're prompting), (5 descriptive keywords or phrases), (Minimalist style), (illustration,Vector) ,(Color),(Composition)

Type yes if you understa

        持续学习、创新思维和不断实践将是你成为一名优秀的AI艺术家的关键。愿你的创作之旅充满灵感与成就！

如需Midjourney教程、电子书、交流群、会员账号，请进入公众号，回复：Midjourney

本文链接：https://ai-huihua.com.cn/aihuihua_79.html