AI generated art just made the jump from jumbles of spaghetti to nearly indistinguishable expert artworks.
It’s reminiscent of the position I find myself in today. It is easy to sympathize with the endangered native groups, those peoples that are being shunned, forgotten, and replaced. But, as I hope to explore in this piece, we artists are uniquely positioned to also become the pioneers of a burgeoning era. Instead of six-shooters and dynamite we have Midjourney and Dall-E.
In this blog post, I hope to help artists use this tool to their advantage while providing insight from my own experiences as a manager. How to evolve into an AI artist and embrace this exciting new technology.
Drafting your Concept Art with AI
The drafting process is key to effective art direction, and one of the most exploitable steps by AI artists. This is a time intensive step that lets the art director figure out the “mood” of a world or scene. Some professionals contract artists across the world to bring in a large variety of personal styles. Others may contract a single concept artist to see a range of moods made by a single person.
Adding an AI to the mix is incredibly easy. My latest painting is a city built into a giant ravine. It is a 19th century gilded age version of a dwarven metropolis. But to fully immerse yourself in the idea of the city, it needs to be consistent across the lore and mood.
At first I was inspired by Edward Hopper’s Nighthawks, and the melancholic mood of a lonely twilight. This gave the city a bit of a Bioshock feel. A video-gamey liminal space feel, that didn’t evoke much emotion.
Current Prompt: An underground roaring 20s metropolis, built into a giant cave, skyscrapers stacked on top of each other, bright windows, and incandescent signs, photorealistic epic concept art painting by Edward Hopper
I changed the prompt slightly, indicating " 1800s victorian," and "day." Removing Edward Hopper also brightened the scene considerably. This made the scene a lot brighter, but still felt a little empty without anyone walking around the city.
Current Prompt: An underground 1800s victorian Times Square, built into a giant cave, skyscrapers stacked endlessly on top of each other, bright windows and incandescent signs, photorealistic epic concept art painting, day
As such, I changed the wording to find a more bustling and vibrant scene. After a few iterations, the scene was already looking more like a lived in version of Barrelmouth City. People were crowding around the world, criers stood on crates, and vague cart shapes seemed to materialize. Here are the three finalists I selected from the drafting process.
Final Prompt: A crowded dimly lit magical underground Art Deco victorian Times Square with carts and vendors everywhere, built into a giant cave, skyscrapers stacked endlessly on top of each other, bright windows and incandescent signs, photorealistic concept art painting, day
The first of which gave the impression of a New Orleans inspired market district. Buildings crowd together, while people and carts scatter around the street. Red shapes around the painting indicate the signs of various shops as well as a festive red clothesline. But it doesn’t show off the underground nature of the city at all. This is a problem shared by the second picture. I love the idea of criers standing on boxes while crowds of people wander around—but it feels like a common outdoor area more than anything else.
This leaves us with my selected picture. Crowds of people are scattered throughout the scene, and the varied buildings give a sense of scale to the scene. The main building is dwarfed by the stone arches that loom above. It reads as a gilded age city, and one that is built into the earth. While there isn’t too much in the way of magic, we can use Dall-E’s editing feature (and our own painting) to make some minor adjustments.
Adjusting your AI generation (manually and in Dall-E)
Final Thoughts on AI generated Images
I previously played around with using 3D renders as a base for my cityscapes. Production designers sometimes use tools such as Sketchup to quickly put together scenes in a perspective view. I also tried Maya and Daz Studio to place shadows (at the expense of a much longer creation process). Finishing this picture helped me realize that AI renders might just take that place for now. They have a ton of power for creating small images and drafting ideas. But I’m not too worried -- they aren’t fully replacing a stylized artists or concept artists just yet.
That said, we are still in the wild west of AI art. Maybe one day, you will be able to completely develop a scene by cropping down an image and specifying certain logos, billboards, types of people, and so on. But until then, I’m going to be using it to my advantage as an artist.