Original author：𝚃𝚑𝚎 𝙻𝚊𝚝𝚎𝚜𝚝 𝙽𝚘𝚠 ~ 𝙰𝙸
内容简介：Today I saw this short article explaining the combination of ChatGPT-4 and Midjourney V5, which is a good article，Probably translated it。The main gain is：Don’t expect to generate perfect prompt words all at once，Need to take advantage of ChatGPT’s ability to understand context、Detailed description、Continuous optimization，Finally, make an excellent prompt word。
Two important versions were released just last week，They are GPT-4 and Midjourney V5 respectively。Since the launch of these two versions，We now have the ability to generate extremely realistic images。
This article focuses on understanding the capabilities of advanced artificial intelligence technologies，and how they work together，Produce impressive and realistic images。
- 关于 ChatGPT-4 和 Midjourney V5
- 如何同时使用这两者
ChatGPT 4 is a powerful language model developed by OpenAI，It can understand and generate human-like text based on context。
It can generate prompts for Midjourney V5。 MidjourneyThis is a premium model，Can be described via text，Create high-quality images。
This article discusses how to combine the two，Generate high-quality images。
Boot ChatGPT-4，优化提示
When it comes to generating prompts for desired images，The bootstrapping process is critical to ensuring they are accurate and descriptive。
In order to effectively use ChatGPT-4，It is important to give it the correct information about Midjourney V5，For example, what features should an image have?、What settings to use and sample tips。
This information helps ChatGPT-4 create contextually rich、vivid reminder，帮Midjourney V5理解并转化为高质量图像。
构建上下文意识
ChatGPT-4的独特之处在于其出色的上下文理解能力。
ChatGPT-4可以通过获取Midjourney V5的所有必要信息，理解扩散模型中微妙的细节和复杂性，例如如何将文本描述转化为图像。
ChatGPT-4可以通过上下文意识创建与Midjourney V5完美匹配的提示，确保两种人工智能技术之间的顺畅协作。
生成详细提示
ChatGPT-4可以开发适合不同主题和风格的详细提示，前提是它已经掌握了Midjourney V5。
ChatGPT-4可以根据你的要求，精确地制作出你想要的图像，including color、光线、纹理和情绪等方面。
ChatGPT-4可以处理抽象的东西，比如特定的情感或故事情节。
ChatGPT-4 uses these details to create super descriptive and imaginative prompts，This is exactly what Midjourney V5 needs to create images。
Continuous iterative optimization
Cooperation between ChatGPT-4 and Midjourney V5，Not only can it be done in one go。
This process is more like a continuous improvement process，to ensure we get the best results。
If the resulting image does not meet quality or accuracy standards，我们可以给ChatGPT-4提供更多信息或更好的提示，帮助它为Midjourney V5产生更适合的描述。
不断改进和重新生成图像，可以确保随着时间的推移，结果变得越来越好，从而产生更准确、视觉效果更佳的图像。
ChatGPT-4 + Midjourney V5 创建图案实例
ChatGPT-4和Midjourney V5的配合使用，结果将令人印象深刻。
AI生成图片的潜力无群，Midjourney V5可以通过使用GPT-4提供详细提示，来创建符合不同主题和风格的各种图像。
For example：
历史人物：产生一个美洲原住民女人（Historical figures: Generate A Native American Woman）
ChatGPT-4 的响应：
Midjourney V5 图片输出：
ChatGPT 提示:
Picture a humanoid robot with a sleek and futuristic design, standing around six feet tall. Its head would resemble that of Elon Musk, with his distinctive hairline, sharp jawline, and piercing gaze. The robot’s body would be clad in a form-fitting suit, perhaps similar to the one Mr. Musk wears when presenting at events. The robot’s movements would be precise and controlled, reflecting the CEO’s reputation for attention to detail and perfectionism. （想象一个具有时尚和未来主义设计的人形机器人，站立在大约六英尺高的地方。它的头部类似于埃隆·马斯克（Elon Musk），有着独特的发际线、尖锐的下巴和锐利的目光。机器人的身体将穿着合身的西装，可能与马斯克在活动中展示时穿的西装相似。机器人的动作将是精确和可控的，反映了首席执行官注重细节和完美主义的声誉。）
Midjourney V5 输出：
更多的照片：
直接用Midjourney V5生成的图片——高仿真图片（Realistic Photos）：
穿汉服的女孩 提示词（作者Mage)： In a palace, a girl dressed in a gorgeous Hanfu( traditional Chinese clothing) is being photographed. She is a celebrity named Song Zu’er, a princess with exquisite hair accessories, clean attire, a pure and innocent face, a playful and cute demeanor, and a perfect figure. The background is also very clean, The scene is hyperrealistic, photorealistic, with a clear focus, delicate lighting and shadows, and a studio – like set – up that evokes a cinematic feeling，85mm, f 10, sharp focus, with bright lighting, 4K, ultra photorealistic, ultra detailed, intricate details, Kodak gold 100, 4k –ar 2:3 –stylize 1000 –s 250 –v 5
中国模特 提示词：Beautiful Chinese model, looking at the camera, full body, low-cut short dress, tight –ar 9:16 –q 0.5 –upbeat –v 5 –s 750
illustrate：在测试中发现，如果输入不适当的提示词，Midjourney会自动忽略这些不合适的提示词。
我觉得AI生成模型将继续优先考虑安全，避免生成不适当的内容。
AI生成模型的安全性和内容适宜性正在得到研究人员和科技公司的持续关注。虽然未来无法确定，但可以肯定的是这方面会有不断努力。
结论
GPT-4和Midjourney V5的配合使用，充满无限可能性，令人振奋。
技术和艺术的交汇将会革新创意世界，为艺术家提供探索和表达自我的新颖方法。
期待这个强大组合将会创造出令人惊叹的作品。
