当ChatGPT可生成Prompt提示词，文生图进入“图能达意”的新阶段
当ChatGPT超强的文本理解能力和文本生成能力，与Dall·E 3的文本生成图片能力相结合，文生图才真正可以按照你的想法生成图片。
国庆假期余额不足时，打开ChatGPT发现新增DALL·E 3可以用了，由ChatGPT超强的文本能力生成Prompt提示词，亲测，简直不要太爽！
选择DALL·E 3（beta）之后，直接输入“一只在树林里奔跑的狗，漫画风格”，ChatGPT会自动生成四个Prompt提示词和对应的图片。
Prompt：cartoon style image of a dog joyfully running through a dense forest, leaves fluttering around and the sun peeking through the branches.
用户输入自然语言，ChatGPT可以按照你的描述生成Prompt提示词。
即使你说的是废话，它也能智能地过滤掉无效信息，非常准确地理解你的意图，这种语义理解能力是ChatGPT最擅长的！
对话框里输入：“满满是服务器和服务器机架的数据中心里，一个工作人员把在检查一个服务器的硬件。”
Prompt：photo inside a modern data center, where rows of server racks can be seen. A technician of European descent is working diligently, adjusting some cables connected to a server.
Prompt：picture showcasing a vast data center with humming servers. A technician of South American descent, holding a tool, is intently checking a server's hardware components.
这里出现了一点超自然的现象，单手端起了一台2U服务器，我随后输入：“刚才这个图片，单手托着服务器不太合理，这个东西太沉了，最好是展示从机架里抽出服务器的画面，机架是有滑轨的，像抽屉一样，这样操作更轻松一些”。
我随便用口语反馈了修改意见，于是就生成了下图。
Prompt：photo inside a state-of-the-art data center where rows of server racks are visible. A technician of European descent is smoothly extracting a server from its slot using the rail system, preparing for a check-up.
Prompt：picture showcasing the interior of a vast data center. A technician of South American descent, with both hands, is gently drawing out a server from the rack, taking advantage of the sliding mechanism.
Prompt：photo of a modern data center filled with server racks. A male technician of Asian descent is carefully sliding out a server from its rack using the built-in rails, resembling a drawer mechanism.
这几张图片最大的问题是运维工程师穿的太帅了，但并不是不合理，所以，还是比较完美的。
当我输入：“给我一张文章配图，展示的是一家公司在拓展全球版图的抽象配图”。
Prompt：photo concept of a company's global expansion: a 3D globe with highlighted regions and connecting lines, symbolizing the company's reach and influence.
Prompt：illustration depicting a company's worldwide growth: a stylized map of the world with highlighted countries and growth arrows pointing outwards.
当我输入：“帮我设计微信公众号的logo，风格参考多啦A梦，但是不要侵权。我的微信公众号叫云体验师，意思是帮助大家解读云计算的技术创新和落地案例创新故事，有时候会以亲自动手体验的方式来介绍，这样显得有声有色，有血有肉。”（同时也有提醒ChatGPT注意版权问题，不要侵权。）
Prompt：digital render for the '云体验师' WeChat public account logo: a charming robotic blue cat, drawing inspiration from Doraemon but with a different design. The cat is sitting on a cloud, with a virtual reality headset, emphasizing immersive experiences in the realm of cloud technology.
Prompt：digital render for the '云体验师' WeChat public account logo: a charming robotic blue cat, drawing inspiration from Doraemon but with a different design. The cat is sitting on a cloud, with a virtual reality headset, emphasizing immersive experiences in the realm of cloud technology.
Prompt：graphic representation of a logo for the WeChat public account '云体验师': a cartoonish blue cat, echoing the essence of Doraemon but distinctively original. The cat stands next to a server rack cloud, holding a manual, signifying the hands-on approach to understanding and interpreting cloud innovations.
Prompt：illustration of a logo for the WeChat public account '云体验师': a robotic cat inspired by the style of Doraemon, but distinct enough to avoid copyright infringement. The cat wears a cloud-shaped badge, holding a digital tablet showcasing cloud computing symbols, representing hands-on experience in the cloud tech world.
经过多轮对话，最后生成了这样一张图，正在考虑把它替换成微信号的Logo。
Prompt：graphic representation of a logo for the WeChat public account '云体验师': a version of the animated blue cat, distinct from Doraemon and without the bell or background circle, with a visible small mouth, sitting against a plain pure white background. The cat holds a magnifying glass with a cloud visible through it, emphasizing exploration of cloud innovations.
当然，作为Word工作者，无数次幻想加入插图来给文章添点色彩，又要注意不要侵犯别人的版权。
恰好最近写了一些“闪存存储对磁盘存储发起冲击”为主题的文章。
于是，就让ChatGPT帮我生成了这样一张图。
Prompt：graphic representation of a technology face-off in a wider resolution: a fast, shining character epitomizing flash storage, poised to launch an attack. Confronting it, a hefty, machine-like character with the letters 'HARD' clearly visible, representing the mechanical hard disk, setting up barriers, preparing to defend, in a tech-inspired environment.
这张图的效果还是挺惊艳的，现在正在考虑以后每一篇文章都让ChatGPT帮我生成一张配图。
作为文生图技术爱好者，ChatGPT生成Prompt提示词给我的震撼还是挺大的，算是文生图领域一次很大的进步。
Stable Diffusion那边，SD 1.5的时候还需要一大长串Prompt提示词，而SDXL用较短的提示词即可生成不错的图片，但还是要手写Prompt提示词。
而现在，让ChatGPT帮你写Prompt提示词才算真正实现了Prompt提示词自由了吧。
所以，Prompt魔法师还没有上岗，就已经要失业了吗？
作为Stable Diffusion用户的笔者，今天在看到ChatGPT加持的DALL·E 3之后，终于有了始乱终弃的想法了。
目前，ChatGPT上的DALL·E 3修改图片功能还有所欠缺，比如，后续可以加入上传图片的功能。
此外，在生成图片的可控性方面也有些不足，如果能支持ContronNet就会好很多。