Here's How I Made This Image With Midjourney

A walk through a very peculiar process of choose your own adventure.

There is nothing particularly spectacular about this image which I generated using the AI program Midjourney (v6) but I do think it’s an interesting process. People often ask me to share my prompts with them when I post these images, but I’m always hesitant because it’s never just one prompt and even when it is (which is the case for this image) the result is not based on just one prompt, but rather a kind of “choose your own adventure” that involves dozens and dozens of variations. This image began life with the prompt:

Dungeons and Dragons Saturday morning cartoon, vintage 1980s animation style --ar 32:15

Stylization was set at 750 and the results this gave me at first were these. I toyed around with a couple of them and then had the program run the prompt again. And then again.

The third batch of images it came back with were these:

If you look closely at the image at the top of this post, you’ll recognize our friend in the bottom right corner image. A young man with brown shaggy hair standing near a green dragon-alien-dude, looking like they’ve just come off the set of Star Wars or some early 80’s fantasy movie. Here:

From here I used the pan down feature and created a new batch of images:

I liked the third and fourth the best and started toying with them both, but it’s the third image that ended up in our final product. I began panning it up and down until I came up with a much taller image showing his full frame, the alien-looking-dude still standing behind him (this is probably my favorite of all these images, better than the final product because it’s more focused and simple):

Each time I panned, I’d have choices to make:

The fourth image in this block became the image above. It has the best hands.

At this point, I panned to the right, resulting in a rather shocking reveal:

Huh, I thought the dude was the main subject of this piece, but now a beautiful woman has appeared even closer to the camera. They’re dressed similarly, however, so I assume they’re traveling together. Siblings or lovers (or maybe both, you never know).

Of course, using this woman was also a choice. The pan left me with four images to pick from:

What if we pan down again?

What if we pan left instead?

What if we pan down and then pan left and then pan right?

I don’t want her sitting down!

Now what if we zoom the entire scene out?

One more zoom out and you get the image at the top of this post. As you can see, there’s a lot more to it than a simple prompt. It’s not hard mind you. There’s no artistry here, really, but there is a game involved. And it helps to have an eye for it. (And you’ll likely accidentally get some nudes when you do this stuff—I didn’t share a couple where the woman just mysteriously had no bottoms on at all!)

And that, dearest readers, is how I do AI art. Obviously I left out a ton of variations that were discarded along the way, and simply choosing a different path would have led me to a totally different image. Infinite possibilities from a single prompt. Pretty wild.