On creating AI art
I think one of the reasons why many people are very negative about AI art is because they think it requires little to no effort. You just type a few words and voila: you have an exact depiction of what you had in mind.
It’s an understandable viewpoint and lots of AI art is made with little effort and imagination. However, I think it’s important that we show how some people do try to be a bit more inventive both in subject and process.
I hope that by showing how it’s made people start realising there’s a lot of creative choices that need to be made. To me creating with AI is a lot more like creating abstract art, instead of photography. What colours will I use? What composition am I after? Which effects?
To give people insight into this process I’ll discuss how I created one of my favorite works. This one ↴
If you like my work, please give me a follow on https://www.instagram.com/oh_x_d/ or https://twitter.com/oh_x_d/
I won’t be showing prompts and there’s a few techniques I won’t go into detail in: I think it’s important to keep some of the magic to myself. Also; it’s not the focus of this post, I just want to give an idea of the whole process. MY process.
I tend to use MidJourney these days, so that’s what I used for this image too.
Let’s talk about the initial idea: I knew I wanted to do something with a balding man wearing a long dress in a field, being followed by a bear (Yeah, don’t ask me why). I imagined him being Sam Louwyck, a Belgian actor and dancer. He’s got a very characterful face and the camera just loves him.
At first I always start out very basic, just to get a feel of what I’m after and what’s the best way to get it. I like to place the subject into a completely different setting at first, because it helps me to focus on the subject 100%. Here’s some example generations:
Next I start placing the subject in the desired setting. For me it was clear it needed to be a field in Finland. I have Finnish roots and I just love the sober beauty of the country side there. Also, I’m testing out mood and lighting. Again some sample generations:
Two of these caught my eye: this Rasputin-like figure staring straight into the camera. I like his expression on the left, and I like the pale light at the back in the right image. Let’s see what we can do with this.
Just a quick pan to the right, and let’s introduce a bear.
Yeah, no, that’s not it.
Some more generations, trying to create the subject in the setting:
Again a pan to the right and … well, I’m not sure what I was thinking here…
Let’s continue rolling.
Finally, I’m getting Sam-like results. It wasn’t my intention to get him exactly, but I knew I wanted a man, balding with long black hair.
Some panning downwards to get the dress to reveal.
Unfortunately MidJourney is adamant about men never wearing dresses. I had to choose: either he wore a dress, or he’s balding, but not both.
Now, let’s introduce some dynamic. I like that he’s facing away from the camera, so I definitely wanted to keep that in. I’m absolutely in love with blurring things, so here we go, let’s start to blur him up.
And if you compare to the final image, you’ll see this is the one I went with. I like that his posture is somewhere in between slouching and walking, he’s perfectly blurred, he reminds me of Sam, but without being a clone.
Let’s pan a little to the right, I didn’t want the bear-person to be too close, but rather farther in the distance. So here’ Im just broadening the field, pulling back the camera.
Let’s create this bear-person that’s chasing him, by panning to the right. This literally was a lucky shot, I just needed one generation.
I loooove the mask-like quality of the bear head. It’s clearly not a real bear, but a person wearing a (failed) bearmask.
Also I knew I wasn’t there yet. It’s a good image, but something is still missing, not just compositionally, but also with regards to the setting. It conjures up stories in my head, but doesn’t make me question things enough.
Let’s pan even more to the right. What’s more Finnish than a wooden house or barn? I like that they’re very simple buildings.
The one in the left upper corner is it. It has a very tiny window and even though the door is open, there’s just another door.
I like it. I like the separate lighting of the barn, maybe it’s just a bit too light though. But this is a good image; it makes me wonder what is going on: Who is this man? Who is this bear-person? Does he really exist? Are they both coming from the house? Whose house is it anyway? Is the bear chasing the man away? Maybe the man’s an intruder? Or was the bear lurking in the grass or behind the trees to follow the man when he leaves the house? etc.
However, we’re not quite there yet. I like everything about the image, but it’s too wide. The ratio does not fit the mood. Wide images have this cinematic quality, since we’re used to seeing such a ratio in films, but to me, that’s too glamorous for this image. I wanted it to have a pedestrian quality to it, in contrast to the bizarre nature of what’s depicted. So let’s hit the square-button.
After some regenerating I got a good one.
This is it. I just need to finalise it in Photoshop. First I cropped the image and corrected the color differences between the various parts. Then I adjusted the lighting of the barn.
Compositionally it’s good. Automatically you “read” the image from left to right, your eye is drawn to the man since it’s the biggest object, his shadow leads you to the bear, which in turn leads you to the barn.
There’s just one thing: there’s a lot of open space in the left upper corner. That’s ok, since it gives breathing space to the composition, and nicely contrasts with the dark woods on the right. It tells you the man is walking away from a dangerous situation, out into the open. But even then, there’s still something missing.
It needs something… A pale, half transparant moon is perfect. (Photoshop generative fill is really amazing.) It further balances the composition and also completes the mood. Now just needs a little bit of film grain.
So, that was it. Now, in all honesty, if you still think this was NOT a creative process, please tell me how it’s different from someone creating abstract art in e.g. illustrator? Or a photographer messing about with developing times in the lab? And tell me this: if I had used stock photos and created a digital composite out of those, would it then suddenly be art? If so, why? When synthesizers were invented some people were dismissive because they didn’t think of them as “real” instruments. And it’s true: a synth trumpet sounded nothing like a real trumpet. But the moment artists started embracing synths for the unique sounds you can create with them, it triggered a creative revolution. I honestly think that’s what we’re facing here too. AI image generators are synthesizers for images and will revolutionize the artistic field.
Please, don’t forget to click the clap-button if you liked this story!
