AI scares the hell out of me. So I learned how to art direct it.

Our Associate Creative Director Chris Graver shares his take on the new wave of image generators and how he learned to make them work for him.

As a creative director, I think of pitching in two distinct eras: the time before the proliferation of AI, and the time after this 2023 tidal wave of technology.

Pre-AI image generators, I was reliant on leaps of imagination—helping the client picture in their minds what I had already envisioned in mine, using pencil sketches or stock photography.

The process of creating with AI tools has been a revolution. I can create things in so much detail that my clients can see exactly what is in my brain, on the page or screen, in perfect clarity. The weight on the creative to persuade, which can feel akin to a filmmaker fast talking a movie pitch to skeptical executives, is much lighter with a tool like Midjourney, DALL-E or Stable Diffusion, the three leading AI art generators. That magic “click” between agency and client in the planning stages is happening more easily thanks to AI.

I art directed AI to generate my own self portrait for this piece.

My approach to AI is one I’ve advised my peers to take (and it’s an attitude cited by two wise cultural figures —Ted Lasso and Walt Whitman): “be curious, not judgmental.” This tech scared me, so I became hellbent on understanding it. Now it’s a tool in my arsenal, but to make the most of it, I had to learn how to art direct it effectively.

The key is learning how to manipulate the way AI borrows things from the world of human creation—the more specific and dialed in your instructions, the more in line your results will be with what you’re seeking. Here are the four essentials that designers and creatives should know to get the most out of AI’s image-generating capabilities.

1. Know when and where to use it

AI is an effective tool when it’s applied to the right stage of the creative process, like early ideation, when it can make you faster and better at sketching out an idea you want to help others understand. But when it comes to creating actual work for clients, it’s not dependable—you technically don’t own it under the current copyright rules, given that you didn't make it, the computer did.

2. Know how to prompt

So you know at what stage you want to use AI; now you have to become skilled at prompting, or art directing, it. Simple prompts don’t work, but neither do prompts that demand too much of the AI. It’s a balance, learning how to build incredibly detailed directions that produce the results you want. The prompts I write take enough time and energy that I save them for later use.

3. Know the language

You must have a strong grasp on the language of photography and art, understanding terms around lighting, composition, camera and rendering so you can speak in exquisite detail and pull from the vast body of human work the AI is using to create the thing you’re looking for. Think of yourself as a composer of the most epic list of search terms you’ve ever written. In my prompts, I’m focusing on giving lighting direction, specifying right down to the shutter speed; I’m using expressive adjectives to describe what I’m aiming for—words like “cinematic, shimmering, elegant, ambient.” In fact, those types of terms dominate, while the description of the actual elements of the photo can be as simple as “portrait of a woman in a car taking a photo, looking out the window and smiling.”

4. Practice

It takes time to get the hang of it. I’ve created some frightening Frankensteins—images more fit for an R.L. Stein book cover than a creative pitch for a building materials company. But I’ve learned what the AI can and can’t handle through repeated experimentation.

Most recently, I’ve created a storyboard for a video that thrilled not only the client, but the director of photography, who now had a sharper understanding of our vision for the project. We’re 90 percent of the way there, because we’re looking at the same visual (rather than depending on what our minds conjure up separately).

During my initial attempts at generating a warm-looking image of a man in front of his suburban home with a lawnmower, I got some freakish results; it took tinkering with my prompt and simplifying what I requested to eventually get the image I was aiming for. (These didn't pass muster, of course.)

key takeaway:

While I wouldn’t use an AI-generated image for a client’s public-facing project because of concerns around copyright, I do believe it’s an incredible tool to speed up my work and create clearer understanding between me, my clients and my colleagues. It allows me to be more visually expressive with my ideas faster, without relying on imagination only to fill in the gaps. Like any new technology, it’s intimidating—but you can’t dismiss something you don’t understand. My ultimate advice is to take the time to learn how to make AI work for you.

More Good Reads

7 tips for successfully navigating a merger or acquisition

So you’re managing a new acquisition. How do you prepare to create one happy family? In our experience guiding c-suite leaders as they navigate the process, we've noticed there is no one-size-fits-all solution. But there are seven critical questions that make the difference between a graceful implementation or a floundering one. Learn how to answer them with confidence and ensure your newly expanded organization can thrive.

AI is here to stay. Are you ready to leverage it?

The building materials industry is at a unique crossroads: customers want more personalized service than ever before. And AI may lend itself best to the needs of the building materials industry to deliver that. Executive Creative Director and Brand Strategist Anita Holman explores how this game-changing tool could enhance every brand’s ability to develop personalized advice in ways that were never humanly possible before.