Recently, I submitted a print to the Saskatoon Camera Club for a year-end competition. The print was in the Altered Reality category, which had the following rules:
“Manipulate your own photographs to create surreal, abstract, impressionistic, or unrealistic images. Experiment with either in-camera or post-processing software using special effects, montages, collages, composites, heavy HDR, or any other technique. There is no limit to what you can do! Be creative, but at the same time, try to make your image look less like it might belong in the Traditional category. If you think in terms of “artistic” – that’s the image we’re looking for. Think “digital manipulation – creating art” for this category.”
I especially noted the sentence, “There is no limit to what you can do! Be creative.” This suggested to me that I had Carte Blanche to go all the way. Since several people asked me how I did this work, I thought explaining the process and my various choices would be fun.
After playing around with the image for a while in Photoshop and not getting the result I was after, I decided to try and use Midjourney. In case you don’t know, Midjourney is an artificial intelligence tool for generating images from textual descriptions. This tool allows users to create detailed and varied visual content simply by inputting text prompts, leveraging the power of advanced AI algorithms to interpret and visualize ideas.
My starting point was based on an early morning image I made during a road trip to the Alberta Badlands in Drumheller. As you can probably see, It is not hard to imagine dinosaurs roaming around here doing dinosaur things.
After post-processing the image in Lightroom, I created a Midjourney account through Discord and created a server that I could work with on my own. Once that was done, I uploaded the image as a starting point. After uploading the image, I provided the following description to Midjourney.
“/imagine [Image URL] Badlands landscape with Volcanic activity in the background and various dinosaurs roaming the badlands and a flock of pterodactyls flying overhead looking for prey. The light in the image must be that of an early morning sunrise with an orange hue.”
Midjourney generated these four images based on my first description
Midjourney provides four images, allowing you to select, further work on, and refine an image closer to your vision. I did not particularly care for those images, as I found them too far removed from my original work.
To make a long story short, I continued to refine my textual description and found out that I could specify how much “Image Weight" Midjourney should assign to my original photograph. My final instructions, after sixteen different versions of my textual description, ended up like this.
“/imagine [Image URL] Badlands landscape with an active volcanic mountain range in the background. Overhead flies a large flock of pterodactyls, and in the middle ground is a large group of Trex moving from east to west. In the foreground, we have a large group grazing Sauropods; the light in the image must be that of an early morning sunrise with an orange hue. --iw 1.9.”
The final instructions –iw 1.9 stands for Image Weight, which indicates to Midjourney the amount of weight to assign to the original image. This value can go from 0, no weight at all, to 2.0, which is as much weight as possible. In other words, the --iw parameter allows you to control how much liberty Midjourney can take with the original image.
A value of 1.9 is pretty high, and I am asking Midjourney to preserve a large portion of the original image.
After running the last generation, Midjourney provided me with the following four images.
Based on the four images provided, I selected the bottom left and then ran an upscale process, which generated the final image in high resolution. This final image was re-imported into Lightroom for minor enhancements and printing, resulting in this final image.
And now for the real question: Is this cheating? If not, what do we call it?
Based on the competition rules stating I am allowed to do anything – it’s not cheating. To me, a competition with altered reality images is a harmless spot of fun, allowing us to experiment with something new. However, the image coming out of this is good enough that if you did this with something more believable than Pterodactyls and grazing Sauropods, it would start having more ethical consequences.
In this article, I will not get drawn into a discussion about AI and altered images, as that is fodder for a much more in-depth look at some complicated topics. All I want to do is start the discussion in the Saskatoon Camera Club and perhaps others about what we should do with technology like this.