AI, Abomination of Desolation or Artist’s Little Helper

A confession. I, an artist, have tried my hand at making AI art on Midjourney. If your done clutching your pearls and gnashing your teeth, I'll tell you about it. I never thought I would be interested in using artificial intelligence to create art. But when I stumbled on Midjourney, an AI art generator, my curiosity got the best of me. I wanted to see what AI could create, and I have to say, the results were both fascinating and frustrating.

The AI can create pretty cool images with very simple prompts (think "tabby cat wearing a crown" or “beautiful woman”) but trying to get it to do anything specific is unimaginably difficult. Hours of reading online guides and meticulously constructing prompts would meet with only limited success (and bizarre looking hands if not multiple arms). Even the most current and advanced AI pictures tend to have that noticeable AI look to them. Personally, I never see what the AI makes as a finished product, only as a starting point. I always take the images to Photoshop where the real work begins. I am fully aware of the controversy surrounding the use of AI, and so in fairness I think it’s best to be up front when it's being used.

I think people would be interested in seeing the before and after photos so they could see (and judge) for themselves what the AI is capable of vs the artist. Like many using Midjourney I have a background in art. I’ve taken numerous college and independent drawing, pastel, illustration, watercolor, oil, acrylic, and photoshop courses. I say this because I don’t think AI will replace artists, but it is another medium for artists to use. Recently, I posted some illustrations I made for my poem. Originally the images began as AI generated art but I put many hours of work into creating the finished images.

So, like most AI images, it looks really nice on the surface. The woman looks beautiful, there are stars, mountains, mist, clouds, kind of a walking path below, all the things I asked for are present. I won’t get into the prompt, other than to say the AI sometimes gets obsessed with an association between two things. For me, if something is happening “at night” or “night-time” the AI will sneak in a piano somewhere in the image. You start going a little mad, adding things to your prompt like, “her hands are at her side, no piano, there was no piano anywhere near that mountain, and no one played the piano, --no piano –no keyboard.” I’m exaggerating but this is why good prompts start looking more like deranged creative writing assignments. Preventing what you don’t want is just as important as prompting to get what you do want.

Back to the image, while initially it looks good there are numerous problems. The first thing I noticed was that the light source is wrong. Everything is clearly lit up from the front but the moon is behind the woman and the mountains. The moon also has an ominous vibe; it’s a bright white moon but somehow also has a blood moon glow. I only wanted clouds anyway, so I got rid of the moon entirely. The mountains also have problems. On both sides of the women, the stars descend too far down where only mountains should be.

I also wanted the mountains to have an atmospheric perspective, so, I added in additional mountains which helped shape a natural path far below. Another strange problem is that one entire mountain range is somehow in front of part of her billowing cloak. The AI seems to have used the fabric to outline the mountain top. There were multiple other small problems with her dress that I fixed and I removed the AI star theme from her outfit. After I made these changes I moved on to fixing the woman.

I entirely changed her face and hands. I am not great at hands still but at least mine aren’t the monstrosities the AI created. She was also missing a lot of an arm as well. I worked on the color last. The color of the original image was very washed out and muted. It made it look like a winter scene rather than a foggy springtime night. I also thought her skin color was very pale and lacked any warmth or life to it. She wouldn’t be a very good adventurer if she had an iron deficiency. So, I fixed that. Looking at the scene I realized she could never travel in winter wearing that dress so I wanted the greens in the background a little more pronounced to imply springtime. Overall, I felt that everything just lacked those deep darks and highlights that make images look vibrant. So, I spend a lot of time working to fix these issues as well. Is my image perfect? No. But I think it’s a nice fantasy illustration that captures the mood of my poem. When comparing my art to the AI image I do think mine is better. I hope people can see how much work goes into making an AI image into artwork.

Now a little note about copyright. Midjourney says I own the copyright to what my prompts create but the US Copyright Office laughed that idea right out of the building. Yet even the US Copyright Office acknowledged that sufficient human authorship added to an AI image would be enough to create a copyright. I agree that the original work isn’t protected. If someone else had my exact prompt they could come up with a very similar image. But, I do feel that the second image was painstakingly altered by my own hand and I would argue that one is mine and has copyright protection.

Whether you agree or disagree with AI art, one thing is for certain, you cannot close Pandora's box. AI is here to stay and I'll be interested to see how other artists make use of it.