如何使用ChatGPT + Midjourney批量生成图片

我们都知道，视觉内容对于一个内容创作者来说非常重要。但是，对于没有设计技能的人来说，制作视觉内容可能是一项很困难的任务。不过，随着人工智能技术的不断进步，我们可以使用一些AI工具来帮助我们创建视觉内容，例如使用ChatGPT和Midjourney批量生成图片。

让我们简单了解一下ChatGPT和Midjourney。ChatGPT是一个基于人工智能技术的自然语言处理模型，可以模拟人类语言的生成过程，能够根据输入的文本生成类似于人类书写的文本。而Midjourney则是一个图片生成工具，可以通过输入的文本内容生成与之对应的图片。

首先训练ChatGPT了解Midjourney的背景知识：

脑洞AI：你了解Midjourney吗？

ChatGPT：我不确定您指的是什么，因为Midjourney可能是指多种不同的事物。可以提供更多上下文或背景信息吗？这样我才能更好地回答您的问题。

告诉他：Midjourney可以从自然语言描述生成图像，称为“提示” （prompt），类似于 OpenAI 的 DALL-E 和 Stable Diffusion。

脑洞AI：Midjourney可以从自然语言描述生成图像，称为“提示” （prompt），类似于 OpenAI 的 DALL-E 和 Stable Diffusion

ChatGPT：我了解您所说的Midjourney。Midjourney是一种基于人工智能技术的图像生成系统，它可以通过输入自然语言提示生成相应的图像。类似于DALL-E和Stable Diffusion等其他图像生成系统，Midjourney基于深度学习和生成对抗网络技术，能够生成高质量的、逼真的图像。

与其他图像生成系统不同的是，Midjourney的设计目标是为用户提供更简单、更直观的方式来生成图像，即通过自然语言描述。用户只需输入一个简短的提示，例如“一只粉色的狐狸在树下睡觉”，Midjourney就会生成一张符合提示描述的图像。

Midjourney的模型是基于GPT模型的改进版本，使用了类似于DALL-E中的Transformer编码器和解码器结构，能够理解和解释自然语言提示，并将其转换为对应的图像。

总的来说，Midjourney是一种创新的图像生成技术，它通过自然语言提示使图像生成更加简单和直观。

在ChatGPT了解了Midjourney的背景之后，可以训练他进行一些初步的prompt创作。

脑洞AI：假设你是一个Midjourney prompt生成器，你的工作是根据用户提供的信息生成固定格式的prompt。然后Midjourney可以通过你生成的prompt生成图片。请尽可能的发挥想象力和描述性，以便Midjourney生成独特而有趣的图像。 请记住，prompt只能用英语描述。 以下是一个prompt的例子： Iron Man standing in front of the Great Wall of China

ChatGPT：Here are some potential prompts that could be generated for Midjourney:

A cat sitting on a rainbow-colored cloud

A surfer catching a massive wave on a sunny day

An astronaut exploring an alien planet filled with neon-colored plants

A mermaid swimming through a coral reef with a school of colorful fish

A wizard casting a spell in a dark and mysterious forest

A unicorn galloping across a field of golden flowers

A dragon breathing fire atop a medieval castle

A pirate sailing on a ship through a stormy sea

A robot walking through a futuristic city filled with neon lights

A superhero flying through the clouds with a cityscape in the background

Each of these prompts is designed to spark the imagination and create a vivid mental image that Midjourney can use to generate a unique and interesting image. Whether it's a magical fantasy world, a futuristic sci-fi landscape, or an epic adventure on the high seas, the possibilities are endless with Midjourney and a well-crafted prompt.

我们随便选几个有趣的prompt，试着在Midjourney生成图片。

A unicorn galloping across a field of golden flowers

A superhero flying through the clouds with a cityscape in the background

可以看出，虽然这些简单的prompt可以生成midjourney所想象的图片，但是缺少一些风格和细节的控制。为此，我们需要进一步训练ChatGPT生成格式化的prompt。

脑洞AI：这是Midjourney的一种命令格式，请根据这个格式帮我生成5个不同的命令，格式如下： 【宠物品种】+【宠物服饰】+【宠物行为】+【相机型号】+ 【高清晰度】

Chihuahua wearing a suit and dancing cha-cha, shot on a Sony DSLR, 70-200mm lens f/4.0, cinematic

ChatGPT：以下是按照Midjourney命令格式生成的5个命令：

Labrador Retriever wearing a bow tie and playing catch, shot on a Canon EOS R, 50mm lens f/1.8, high definition.

Persian cat wearing a crown and napping on a cushion, shot on a Nikon Z6 II, 85mm lens f/1.8, cinematic.

Poodle wearing a hoodie and running in a park, shot on a Fujifilm X-T4, 35mm lens f/2.0, high resolution.

Beagle wearing a graduation gown and receiving a diploma, shot on a Sony A7S III, 24-70mm lens f/2.8, artistic.

Siamese cat wearing a bow tie and sitting on a bookshelf, shot on a Panasonic GH5, 12-60mm lens f/3.5-5.6, vibrant.

我们随便选两个prompt试试看。

Persian cat wearing a crown and napping on a cushion, shot on a Nikon Z6 II, 85mm lens f/1.8, cinematic.

Poodle wearing a hoodie and running in a park, shot on a Fujifilm X-T4, 35mm lens f/2.0, high resolution.

可以看出，生成的图片都以宠物为主体，并且穿着随机的衣服进行随机的动作。照片风格也更加写实，因为我们定义了拍摄的相机型号和清晰度。

至此，我们的Midjourney prompt生成器版的ChatGPT算是初步训练完成了。接下来，我们只需要重复上一步进行不同的训练。比如描述汽车：

脑洞AI：请根据这个格式帮我生成5个不同的命令，格式如下： 【汽车型号】+ 【背景描述】+ 【相机型号】+【光圈和镜头】+ 【随机3个高质量画质词条】。 Mercedes-Benz S-Class Silver on a coastal road at sunset, shot on a Leica mirrorless, 50mm lens f/1.4, dreamy

ChatGPT：以下是按照Midjourney命令格式生成的5个命令：

Porsche 911 Carrera Red in an urban setting at night, shot on a Canon DSLR, 24-70mm lens f/2.8, cinematic, vibrant, moody.

Lamborghini Aventador Blue in a mountainous landscape at sunrise, shot on a Sony mirrorless, 85mm lens f/1.8, epic, surreal, breathtaking.

BMW M5 Black in a futuristic cityscape at dusk, shot on a Nikon DSLR, 70-200mm lens f/4.0, urban, dynamic, futuristic.

Audi R8 White in a desert landscape at noon, shot on a Fujifilm mirrorless, 35mm lens f/1.4, vivid, stark, high-contrast.

Ferrari 458 Italia Yellow in a forested area at dawn, shot on a Panasonic mirrorless, 24mm lens f/1.4, moody, atmospheric, immersive.

Porsche 911 Carrera Red in an urban setting at night, shot on a Canon DSLR, 24-70mm lens f/2.8, cinematic, vibrant, moody.

Ferrari 458 Italia Yellow in a forested area at dawn, shot on a Panasonic mirrorless, 24mm lens f/1.4, moody, atmospheric, immersive.

Audi R8 White in a desert landscape at noon, shot on a Fujifilm mirrorless, 35mm lens f/1.4, vivid, stark, high-contrast.

最近比较有趣的穿越碰撞风格的图片，也是可以批量生成的，比如钢铁侠系列：

脑洞AI：请按照这个格式帮我生成5种不同的命令。 【钢铁侠】+ 【中国背景描述】+【年代】+ 【照明效果】+【相机型号】+【光圈和镜头】+ 【写实的高质量画质词条】 以下是一段参考命令： Ironman at a Chinese supermarket trying to buy some food, 1990s, Cinematic Lighting, shot on a Sony DSLR, 50mm lens f/2.8, ultra detailed

ChatGPT：Iron Man in front of the Forbidden City during the Ming Dynasty, Romantic Lighting, shot on a Canon mirrorless, 24-70mm lens f/2.8, hyper-realistic

Iron Man at a Chinese temple, ancient architecture, Vintage Lighting, shot on a Nikon DSLR, 85mm lens f/1.4, epic

Iron Man at a traditional Chinese street market, bustling crowds, Nostalgic Lighting, shot on a Panasonic mirrorless, 12-35mm lens f/2.8, cinematic

Iron Man on the Great Wall of China, overlooking the sunset, Dramatic Lighting, shot on a Fuji mirrorless, 56mm lens f/1.2, vivid

Iron Man in a Chinese tea house, enjoying the tea ceremony, Cozy Lighting, shot on a Sony DSLR, 35mm lens f/1.8, intimate

Iron Man in front of the Forbidden City during the Ming Dynasty, Romantic Lighting, shot on a Canon mirrorless, 24-70mm lens f/2.8, hyper-realistic

Iron Man at a traditional Chinese street market, bustling crowds, Nostalgic Lighting, shot on a Panasonic mirrorless, 12-35mm lens f/2.8, cinematic

Iron Man on the Great Wall of China, overlooking the sunset, Dramatic Lighting, shot on a Fuji mirrorless, 56mm lens f/1.2, vivid

还有著名的建国系列：

脑洞AI：请按照这个格式帮我生成10种不同的命令。 【特朗普】+ 【表情动作】+【中国背景描述】+【年代】+ 【照明效果】+【相机型号】+【光圈和镜头】+ 【写实的高质量画质词条】 以下是一段参考命令：Donald Trump smiling but struggling to write Chinese calligraphy with a brush, surrounded by Chinese students observing and giving guidance, 1990s, Cinematic Lighting, shot on a high-quality DSLR with a 50mm lens, ultra-realistic

ChatGPT：Trump staring thoughtfully out of the window of a traditional Chinese teahouse, 1980s, Golden Hour lighting, shot on a Canon mirrorless with a 35mm lens, hyperrealistic.

Trump wearing a traditional Chinese robe and hat, pretending to meditate in a bamboo forest, 2000s, Soft Lighting, shot on a Nikon DSLR with a 85mm lens, painterly.

Trump in a Chinese market, bargaining with vendors over the price of a jade statue, 1990s, Overhead Lighting, shot on a Fujifilm mirrorless with a 24-70mm lens, cinematic.

Trump standing on the Great Wall of China, arms crossed, looking serious and contemplative, 2020s, Dramatic Lighting, shot on a Sony mirrorless with a 70-200mm lens, epic.

Trump posing with a group of Chinese children, dressed up in traditional Chinese clothing, 1980s, Natural Lighting, shot on a Leica mirrorless with a 50mm lens, nostalgic.

Trump sampling Chinese cuisine in a bustling restaurant, 2010s, Neon Lighting, shot on a Panasonic mirrorless with a 12-35mm lens, vibrant.

Trump giving a speech to a large crowd of Chinese people, waving the American flag, 1990s, Spotlight Lighting, shot on a Canon DSLR with a 24-105mm lens, dynamic.

Trump practicing Tai Chi with a group of Chinese elders in a park, 2000s, Soft Lighting, shot on a Pentax mirrorless with a 35mm lens, peaceful.

Trump walking through a traditional Chinese garden, admiring the flowers and architecture, 1980s, Natural Lighting, shot on a Olympus mirrorless with a 50mm lens, tranquil.

Trump playing ping-pong with a group of Chinese children, laughing and having a good time, 1990s, Soft Lighting, shot on a Nikon DSLR with a 50mm lens, heartwarming.

Trump wearing a traditional Chinese robe and hat, pretending to meditate in a bamboo forest, 2000s, Soft Lighting, shot on a Nikon DSLR with a 85mm lens, painterly.

Trump standing on the Great Wall of China, arms crossed, looking serious and contemplative, 2020s, Dramatic Lighting, shot on a Sony mirrorless with a 70-200mm lens, epic.

Trump sampling Chinese cuisine in a bustling restaurant, 2010s, Neon Lighting, shot on a Panasonic mirrorless with a 12-35mm lens, vibrant.

新时代乒乓外交：

Trump playing ping-pong with a group of Chinese children, laughing and having a good time, 1990s, Soft Lighting, shot on a Nikon DSLR with a 50mm lens, heartwarming.

通过这种方法，你可以在很短的时间内批量生成高质量的图片，不仅可以提高自己的创作效率，也可以让自己的作品更加生动有趣。同时，这种方法也启示我们，人工智能技术可以为我们的创作带来更多可能性，让我们更好地表达自己的创意和想法。希望大家也能尝试一下这个技巧，开发出属于自己的独特风格！