The details are in the prompt! AI text to image makes it super easy for anyone to create art. Now that anyone can create art you should focus on making images that really standout and for this all you need is the prompt and your imagination.
AI models are packed with massive amounts of knowledge. In order to get them to do things of higher ground one needs to know the science of prompting.
I’ve been using Midjourney since it launched and have used it to design a book cover, course certificate, my blog, etc. I even played around with Stable Diffusion, another text to image model from it’s early days. When I started I created tons of bad images, but slowly with time and practice I got better at prompting and learned the art of creating beautiful images with AI.
What’s an AI Image
It’s just an image like any other on your computer except for that fact that it was created by an AI using a prompt. A prompt is a text description of what you want the AI to create and there’s a bit of a art in getting the prompt right.
It’s all about the Prompt
If your prompt is one liner like “create a cheetah in african savannah” Then the AI uses it’s imagination to fill in the rest of the image which might not end up looking as good or what you wanted. When you have a specific image in mind describe it in as much detail as you can. Imagine that you have an artist sitting with you and you’re telling him exactly what to draw the color choices, lighting, vibe, concept, style of the art. If you want it to look like it was done in a specific style mention that clearly.
Describe every little detail - the background, color of the sky, the position of the Cheetah, the expression, the scene, time of the day, the photo angle, style etc. AI takes your detailed description into account to create the image. Your output will then look closer to what you want.
Make updates or fixes in seconds. You can change an AI images in real time. Simply change the word and you change how the image looks. And the best part is you can create just about any art when you want in an instant. Some tools like Midjourney let you create variations of the generated images or use them within the prompt as a sort of image prompt. There are other more advanced concepts like in-painting where you select a part of an existing image and use the prompt to change it like putting a crown on a persons head.
AI Image Tools
Midjourney, Dall-E and Flux are amazing text to image AI models. My personal favourite is Midjourney and Flux - the visuals are realistic. Dall-E is inside ChatGPT and Bing. Many creator tools have incorporated AI into their services such as Canva, Substack, Descript etc.
How to Create Epic Image Prompts
These are two of my prompt hacks for better outputs.
Reuse Prompt: Never start with a prompt from scratch, look for inspiration by exploring AI art and see the prompts of this art. I use that prompt as a base to iterate on.
Use Chatbots: I use chatbots to craft image prompts. I have a conversation with Claude and ChatGPT on the concept I have in mind. I go into a long conversation about what I’m looking for and ask the chatbot to help me describe the image prompt. I also seek advice on camera angles and other technical aspects. I use that as my prompt and iterate from there on.
I plan to post more how to guides on AI for Regens.
If you have any questions feel free to reach out here or on Twitter
ICYMI