How I Created “Moby Fist 3.1”, a mixed-media AI piece

AI has a lot to offer the art world, and it may sound crass, but I’ve been making art AND working in IT for long enough to feel that I can say the following:

Get your heads out of your collective asses.

I’ve been an accomplished (read: Have exhibited, have won awards) artist for over 20 years, and a programmer for over 30 now (holy shit). This gives me a unique perspective, I think, in why AI is relevant, and I’d like to showcase that point by pulling back the curtain just a bit on how I created “Moby Fist 3.1”, and how AI art is more than typing some words into a prompt.

Let’s begin with the goal.

My goal was to turn this watercolor by yours truly:

Please excuse the oversaturation and rough edges; this is a poor quality scan

Into a photo-realistic digital painting (which is the image you see in the header of this post).

Here’s the broad strokes of how I did it.

First I decided to test how the above image would run through Stable Diffusion’s IMG2IMG generator.

It didn’t go well. I adjusted my prompt (the string of text you feed into the AI to describe what you want essentially) and strength (basically a % value of how much the AI should follow the image as a guide), and couldn’t get anything solid to come out.

I later switched to a recolored digital illustration of the above as a seed image, but it took a lot of testing on the prompt to get the right one: The AI doesn’t think it makes much sense for a fist to be INSIDE a white whale, so it’d keep doing.. Weird.. Things (Ok, well, the piece itself is weird, but I’m talking WEIRD).

I also had to give the AI a lot of freedom in order to ensure I get the photo-realism I craved. I believe I went with a 85% strength, which leans heavily on the AI to come up with it’s own idea.

After I finally got the prompt and strength settled I generated 10 images to use for the whale & arm combo. I ended up using 2 of them to craft the arm & whale portion.

Next, I needed the sea.

This was easy: I just used the AI as, essentially, a stock photo source, and told it to give me a very specific view of a stormy sea and let it run overnight.

I picked out a few out of 200 images that I liked the best, and got to work. I photoshopped in Moby Fist into the sea, crafted the “underwater” view, blended in some extra stormy waves, and also took the skyline from another to complete Moby Fist 3.1 (plus a little digital painting touch-up work).

You may be wondering what the “.1” is in that title; well, I actually created two pieces because I had two different ideas. 3.2 will be released later.

Overall, this was not a quick “typing in some words and getting art” process.

In fact, most of my AI art is an iterative process.. Just like my regular art. The only difference is I’m utilizing an alternative technical process to create it.

Artists have been experimenting with mediums and techniques for a long, long time, and AI is just a continuing evolution of that.

Traditional art still has it’s place in the world. AI is just another tool in the artists’ toolbox.. But in the case of AI art, it’s making creativity even more accessible than it was before.