Playing with Midjourney - art for non-artists


I hated art class at school. I could see so clearly in my head what I wanted my drawing to look like - but my hands just wouldn't obey me. Despite endless tutoring from sympathetic teachers, I left school with an unhealthy distaste for creating my own art. I simply didn't have the physical skills, knowledge of technique, or the temperament to continue.

These days, I enjoy playing with algorithmic art where I am more comfortable controlling the input and output.

A few weeks ago, I went to a "PromptJam" at London's Newspeak House. Think of it like an informal hackathon with people playing with various AI tools to create... something.

After chatting to lots of lovely people, I decided to try my hand at "Midjourney". The concept is very simple - give it a prompt like "Oil painting, woman, dark skin, red dress, eating an apple, ornate frame" and, after a few minutes, it will spit out something like:

Four images. Each show a dark skinned woman in a red dress. She is eating an apple. The frame around the painting is ornate.

Umm... What? Excuse me... What is this witchcraft?

One of the marvellous thing is - the more you look, the more oddities you notice. Why does the subject's hair extend beyond the frame? Do those apples look right to you? What on earth is going on with their fingers? Are those eyes pointing in the right direction? Why are they all in the same pose? I didn't say anything about the model's age - why are they all young women?

That's kinda what prompt-jamming is. How do you convince the computer to draw what you want it to draw? How do you get your imagination onto the screen?

(Aside - the Midjourney user interface is dog-shit. It's all Discord-based. So you have a barrage of other people's images scrolling by. There's no instructions on what the buttons do. GAH! Just build a proper website!)

Midjourney is great for a first draft. If, like me, you struggle to give shape to your ideas then it is nothing short of magic. It gets you through the first 90% of the hard work. It's then up to you to refine things.

It's fascinating to see where artistic AIs tend to fail. Here's "William Shakespeare programming a commodore 64, 1980s, magazine cover, 4k, realistic, photograph". Four generated images.

Amazing! But take a look at the hands! If you squint, those could be old-school computers. Why has it put Shakespeare inside the computer? Why is the text just random squiggles?

Hands and text. Why are they so difficult? You know when you're in a dream and you try to read - and all you see is illegible blobs of ink? That's how Modjourney sees text. Here is a "photo of text, high resolution, Shakespearean, sonnet, poetry":

Four drawings of scrolls. The text is just squiggles.

It's also interesting to see how well it recognises common media properties. Here's "propaganda poster, mickey mouse crushes the workers, museum piece, jagged edges"

Mikey Mouse is distorted on a series of posters.

It gets the cliché of what a Soviet propaganda poster looks like! And then, when you look closer, it breaks down. In fact, you could argue that AI art is merely a cliché generation engine. Give it a prompt like "video game cover art" or "romantic novel cover" and it will conform entirely to your expectations.

As an example, here's a simple prompt "John Lennon staring as an action hero, film poster". Even before the generation started, I had a fair idea of what it would produce.

Four photos of John Lennon staring at the camera. It looks like an action movie poster.

Curiously, Midjourney bans certain words. You can't order it to draw a "nude" artwork. Presumably to stop people creating deepfakes of real people. But throw it a prompt with euphemisms like "rubens, renaissance painting, Aphrodite, full length, diaphanous, oil on canvas, ornate frame" and it'll happily spit out something a bit risqué.

Four semi-nude artworks.

Again, look at the feet! And the hands! Yes, Rubens painted women of roughly that shape - but if he'd done them that badly he'd've been laughed out of the gallery.

There's another curio - take a look at the top left painting. Notice anything in the bottom right corner? Midjourney has seen that lots of photos of oil paintings have a copyright notice in the corner, so has placed some text there.

I like building silly artistic things like "A Collection of Imaginary Software" but I don't have the skill to create illustrations. At the moment things like Midjourney are a starting off point. Or a way to generate placeholders. Or a prompt for a human artist to create something better.

Midjourney lets me quickly be wrong in an interesting direction.


Share this post on…

  • Mastodon
  • Facebook
  • LinkedIn
  • BlueSky
  • Threads
  • Reddit
  • HackerNews
  • Lobsters
  • WhatsApp
  • Telegram

One thought on “Playing with Midjourney - art for non-artists”

What are your reckons?

All comments are moderated and may not be published immediately. Your email address will not be published.

Allowed HTML: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong> <p> <pre> <br> <img src="" alt="" title="" srcset="">