Actor Direction and AI Prompting, Seen Side by Side on a Lagos Stage
During last week's rehearsal, while reworking the lighting cues for the act two transition scene, I had my laptop open, writing a prompt for an AI image generation tool to produce prop design drafts. My fingers were on the keyboard, but something I'd said to an actor kept surfacing: "That anger shouldn't come from your jaw. It needs to rise up from the pit of your stomach." That's when it hit me. The way I explain motivation to actors and the way I request output from AI are curiously alike.
Where They Overlap
Whether it's an actor or an AI, instructions thrown out without context produce unexpected results. If you just tell an actor "be sad here," nine times out of ten they'll force a tearful expression. If you just type "stage backdrop with a sad atmosphere" for an AI, out comes a purple, fog-drenched fantasy landscape. In both cases, getting the result you want requires laying down a specific situation.
- To the actor: "You're not at your mother's funeral. You're standing alone in the kitchen she used to cook in. You lift the lid off a pot and there's nothing inside."
- To the AI: "A narrow kitchen in a 1970s Lagos home, an empty aluminium pot sitting on a gas stove, late afternoon sunlight slanting in through the window. Realistic tone."
Given this level of detail, the actor finds a direction for the emotion, and the AI locks in a visual tone. The underlying principle is the same: present a physical situation instead of an abstract emotion.

Where They Completely Differ
This is where the paths split. An actor asks back. "Right before this scene, did I just run in from outside, or have I been standing here for a while?" That single question changes the breathing, the posture, and the tone of the first line entirely. The better the actor, the sharper the questions, and those questions push me to refine my own directorial intent.
AI doesn't ask questions. Whatever I leave out simply stays missing in the output. So when writing an AI prompt, you have to ask yourself the questions an actor would ask in advance. "What time of day is this scene? What's the colour temperature of the lighting? What's the point of view, the equivalent of a camera angle?" It's like simulating the back and forth of a rehearsal conversation all on your own.
And another thing. An actor accumulates with repetition. Run the same scene ten times, and on the eleventh, they bring a nuance you never expected. During rehearsals last month for a production at a small theatre near Iwaya, the lead actor suddenly delivered a line almost in a whisper on the twelfth run, and the entire crew held their breath. That's the product of accumulation. Feed the same prompt to an AI ten times and you get ten variations, but there's no arc of growth between them. Each one comes out as if it were the first attempt.

Using Them Together in Practice
These days I use AI tools more frequently during pre-production. Visualising an initial draft of stage blocking, quickly comparing lighting colour palette combinations, gathering reference images for props. Work that used to take two days now produces a draft in half a day. It saves a fair amount of money, too. Commissioning a design concept from an outside designer usually costs at least ₦50,000, and now I can rough out a direction internally.
But the moment I step into the rehearsal room, whatever AI produced is nothing more than a starting point. Standing in front of an actor, what matters is the words that come out of my mouth, the gestures I show, and the timing I adjust while looking into that person's eyes. That's directing. A prompt extracts an output, but directing passes through a human being.
The rainy season is here, and I've entered pre-production for the next piece. A folder of AI-generated stage designs and my actor audition notes sit side by side on the same desktop. Glancing between the two windows, the thought that comes to mind is simple. No matter how many tools you gain, the centre of gravity in directing is still between people.
Comments