You can create original work!
Creating original work with AI art apps is a challenge, but it’s not impossible. Of course, it’s much easier to just type in a few words and let the AI produce a finished picture. But for me at least, that’s not satisfying. As an artist I wan’t to be part of the creative process.
I want to work with the AI (or have the AI work with me)!
So, how do I do that?
Here’s one way!
I used the Procreate app and drew this abstract form on my iPad. I just scribbled some lines and added various colors. Later I tweaked the color balance a bit. Then I uploaded this into my AI app (Midjourney) and used the following prompt:
/imagine <https://s.mj.run/oNlDcQalbz8>, a stained glass window
Midjourney took my abstract image and basic prompt and returned four “interpretations.”
Now it was time to play what I call the “cloud” game. Do you remember looking at clouds and trying to identify what kind of animal they looked like? This is pretty much the same. I looked at each of the four images and tried to see an identifiable shape. Sometimes it’s easy; sometimes it isn’t.
(Note: Grid images are numbered 1, 2, 3, and 4 beginning at the top left and going clockwise.)
To me, #1 looked a bit like a sailboat. #2 looked like a Puffin. For #3 I couldn’t come up with anything. #4 looked like a parrot. I could have proceeded with any of them, but I decided to do a reroll, which will give me 4 fresh interpretations of my original image.
This is what the reroll gave me:
On the reroll images #1, 2, and 4 were too abstract but #3 had an identifiable bird. So I decided to go with that. The next step was to upscale the image (increase its resolution). Here’s what the upscale looked like:
Now it was time to begin refining the image. Since the bird was predominatly red, I decided to add the word “cardinal” to my prompt. So I uploaded the upscaled image and then submitted the following prompt:
/Imagine [prompt:]<https://s.mj.run/oNlDcQalbz8>, a stained glass window with a cardinal
Again, I got four interpretations of the prompt, but these were more focused:
Now I had four birds that are what I’d call “cardinal-esque.” Not perfect by any means but something I can work with. Although each image had something I liked, I ultimately decided to upscale #1.
Now that I was getting closer to what I wanted, I downloaded the image again and then uploaded it to Midjourney with a slightly refined prompt:
/Imagine [prompt:] https://s.mj.run/baePPESMjS8, stained glass window:: cardinal on a branch:: –ar 2:3 –seed 100
I added double colons after elements to (::) serve as separators to help Midjourney focus on the specific details I was looking for. Also I added “on a branch” to the cardinal prompt. –ar 2:3 tells the app that I want a portrait orientation, and –seed 100 is just a reference number so I can go back and tweak the specific image without Midjourney changing everything.
As usual, Midjourney gave me four interpretations:
Of all four versions, I thought #2 looked the most like a cardinal, although each had elements that I liked. So I upscaled that one:
It’s still not perfect. But at this point, I’ll take it into Photoshop and/or Procreate and tweak it, adjust the colors, refine the details, etc. until I’m happy with the final product.
So, that’s how I wnt from an abstract form to a recognizable design using an AI app.
With AI there are as many possible approaches as there are people, and I’ve hardly scratched the surface. But I’ve found it to be an amazing creative tool!