I’ve recently been creating AI images with Midjourney, and I’ve had some people push back on my work – so I thought maybe it’d be helpful to see the full process. My hope is that, after reading this article, you can see that working with AI to create images is not just about plugging in a few words and watching as the program spits out great photos.
I’m sure not everyone will use Midjourney this way, but for what it’s worth, I’ve included step-by-step instructions for producing AI images, and I’ve also included example photos to show the approach in action. Here is one of the images that I created using this process:
I should note that I don’t mind people pushing back on using AI. I think it’s a debate that needs to be had, and I, too, have some concerns about AI’s use and how it will impact artists. We need to continue to talk about it, but I would love for artists to see how AI can be used as part of their creative process.
I do hope this post helps with that! Also, I’d love to hear about other people’s processes and am happy to answer any questions in the comments below. And if you want to connect or see more of my work, I’m on Instagram. I don’t use this process for all of my shots, though.
1. Idea generation
For me, coming up with an idea is part of the creative process. The idea for the series featured in this article came from a brainstorming process as well as my recent time spent in Greece, where I visited numerous ancient ruins and wondered who the people who built them were.
I see the writing of words as a creative process, though some images take more creativity than others.
Over the last handful of weeks, I’ve learned how to order my words effectively and what words produce what kinds of images. I draw on my 40 years of experience behind the camera to use words related to photographic lighting, camera settings, etc., to tell the AI what I want. And I draw on my knowledge of composition, color, and what makes an image good.
I also try to bring some words into the prompts that will create an atmosphere or mood in my images. I then run these initial prompts to see what I get.
3. Refining and iterating with images and words
Once I see what the AI creates, I then refine the words, experiment with different ordering of the words, add new words, etc. This process takes many iterations to get a series like this (I’d estimate up to 100).
Sometimes I will add into the mix a photo that I myself have taken to see how that impacts the images that the AI is producing. In this case, I used a picture I took of one of my kids under studio lighting conditions to see if it might help create specific lighting effects:
The idea was to create a series of images taken under modern-day studio lighting yet depicting people from an ancient civilization.
4. Further refining
At that point, I take the best images that the AI produces and run some of them back through the program to teach it what I like (iterating it further).
Sometimes I take two, three, or more images and run them together with different prompts to see if I can improve the results.
In this case, some of the images that the AI gave me based on the image of my son were decent, but the subjects were a little too clean and perhaps younger than I wanted:
So I added words that made the subject look older, dirtier, and scruffier:
After a few more iterations, the program produced this set of photos:
But I wanted to take it even further:
My approach is to continue running the best images, together with various prompts, to evolve the shots the way I want them to go (developing the characters further). Here is my near-final result:
I then take the images I’m most happy with and sometimes tweak them in Photoshop. I rarely add much, but I do sometimes heal blemishes or retouch the images (as I would adjust a photo that I captured with my camera).
6. Upsizing and additional adjusting
Before I’m ready to share my photos, I’ll run them through Topaz’s AI software. I’ll do upsizing, sharpening, denoising, etc., until I’m happy with the result.
I then publish the images to my Instagram, which sometimes involves creativity, as I’ll describe the images or what prompted me to produce them.
Creating AI images: final words
So yes; I see this as a creative process. It usually takes hours and is something I will often do over days.
The series of images that I ultimately produced took hundreds and hundreds of iterations and many hours. Not all of them evolved from the image of my son that you see, but most did draw on it or other images I have taken.
I hope this process helps those who are using the tool to develop their images and offers some insight for those who doubt that AI photography can be a creative process.
I’m sure other people are also using AI creatively, and I hope this might stimulate some interesting discussion. So be sure to share your thoughts (and your AI images) in the comments below!