Recent experiments with Midjourney

This post originated on Substack, which I have recently joined and found inspirational to pick up writing again. I've written quite a few of these over there, and decided to join them together going forward. I used to write a lot, and even considered applying to Creative Writing courses in the past, but alas chose Television Production instead. I enjoyed writing and experimenting with the language, with the form, not for anything other than to do it for myself.

First - disclaimer - this post is not written, or checked, by AI in any way. I wanted to write “from my heart”, just for the sake of writing something. Plus, I think we’re all getting tired of concise, bullet-point-ed, clever and perfectly shaped pieces of texts plastered everywhere. Substack has attracted me to the “freshness” of human writing. Exploring. Sharing thoughts and opinions. Sharing information. I say “freshness” because we used to write blogs and articles on the internet like this. WE did, humans - not computers. We seemed to have left it behind for a bit, to pursue super-quick video reels. And getting back into reading insightful, opinionated, but also science / proof - backed texts has been a real joy and a wholesome experience. And it eventually led me to wanting to explore the same for myself.

Having slightly “dissed” the LLM’s, the ChatGPT’s and Claude’s, I want to dive a little into the synthography and AI generated images nonetheless. Being a photographer, the topic interests me greatly, and I don’t want to be left behind when things inevitably shift. Therefore, I did two “takes” of Midjourney for a month each, one last year, and one - right now. For a $10 dollar fee, one month of experimentation is sustainable for my pocket, though not so sure about the planet’s sustainability.

What have I found…. A couple of major things that are very interesting in their polarity, so to speak. For one, I personally don’t think photography is going to “die”, cease to exist, purely because the computer can generate an image. My first attempt at using Midjourney and synthography (artificially generated image, as I have come to name it through Rankin’s own experiments with AI) were not impressive, in all honesty. It struggled to comprehend fairly basic, but layered prompts, it hallucinated a lot (three legs on a model? a bottle of Coke suspended in the air? no, thank you), and it felt 2D - not 3D, like a photograph of an actual human being would feel. Having said that, the quality was still fairly impressive, and some ideas it came up with did look arresting and stunning. It had some interesting takes on my prompts, giving me new perspectives and ideas. But nonetheless, the three-legged monsters were not my cup of tea.

Exactly one year later, I come back to see if things have improved and whether I can create more accurate results (rather strange way of describing a creative process and art as “getting results” but it may also indicate how it makes us feel about this system?). The quality does look better, most definitively, but I am still not impressed - yet again, it all just looks like digitally designed / produced images. It looks fake, albeit very, very close to the real thing. Is fake a bad thing, one might ask? Not necessarily. Midjourney can create genuinely stunning, captivating ideas; it can offer unique ways of interpreting what you have in mind, allowing you to find something new and fresh about your image. It can rise above the logical mind, too constrained by rules and techniques. It can merge different artistic movements, tools, elements of the past and lead it into the future by creating refreshing new “art”. And, of course, it can create generally good work for almost next to no money. If you’re in need of that, then this solution is perfect.

But - it’s not real. It’s a computer generated imitation of “real”, based on all that is on the internet. It’s flooded the internet so quickly and vastly, we’re all starting to be very aware of it, and look out for it. What’s real? What’s fake? What can I trust? I see people say they’re getting tired of the influx of artificially generated content of any kind. If I go to Pinterest, and look for specific image or make-up reference, chances are most of what I see is AI. Yes, it looks lovely, perfect, but I’m looking for human generated “real” thing. And it’s getting harder to find it. Brands are starting to experiment with it as well. Mango released two campaigns now featuring AI generated models and images. H&M is asking for models to give away their AI imported scans, to be used ad infinitum, with only one-off payment. I feel we’re slowly walking into a place and time in our evolution where internet is becoming a complete parody of itself, almost rendering itself useless. I veer off the tracks here, but I guess I’m just not sold on AI replacing real photographs of real people. Ever:).

Getting back to the images feeling fake, I was talking to a friend of mine recently and she was asking what is an AI generated image, what does it look like, how do you make it. I sent her an example of what I conjured up in Midjourney, and she said the perfectly fitting thing - it just looks like a digital product, not an image. Yes, it looks lovely, but not real at all. And frankly, no matter how much I told it to make it “4k realistic image”, with “—iw 3” (for those not in the know, “—iw” is a command you enter into this DOS-like system to reflect the realistic “image weight”), it just does not feel real. See below - a fantastic (as in dream-like, not necessarily “amazing”) take on a fashion image, with all the abstract elements of it, but despite me saying “4K realistic image”, I just can’t see it - can you? I guess I can’t see the life in it.