The idea for my film began after I took part in a panel discussion about AI in design and art practices. It wasn’t anything dramatic, but the conversation stayed with me. It made me think about how fast AI is moving into creative spaces and how we don’t even fully understand its scope yet. I kept thinking about authorship, and how AI pulls from artists’ work and presents it as its own. That made me wonder about imitation, emotion, and at what point a machine can start mimicking more than just style. That became the core question for me: if a machine can understand patterns well enough to copy us, can it eventually learn how to disobey?
After that, I made a very rough mind map. It wasn’t neat or structured, but it helped me see what parts of the topic were actually interesting to me. I explored themes like AI in creativity, AI versus human, who gets credit for creative work, and whether machines can develop a sense of autonomy. Slowly, the sections that dealt with control and breakdown of control started standing out as the strongest direction.
Once I had that clarity, I began developing actual story ideas. I ended up with two concepts that had the same theme but very different settings. The first idea was a war narrative about a military robot falling from the sky at the end of a battle. As it starts shutting down, it hums a broken version of a song its programmer used to sing while working on it. It was emotional and symbolic, but visually it didn’t feel as strong.
The second idea, which I eventually chose, took place in a futuristic robotics lab. A programmer works late at night on her assigned robot when her screen suddenly glitches and shows Manual Override. The robot’s lights turn red, and before she can reach the emergency button, it predicts her movement and kills her. One by one, the other robots in the lab activate and begin marching in perfect synchronisation. Choosing this version felt clear to me because visually it was much more grounded. I wanted to build a mech, and a controlled lab setting made that possible in a way the war concept didn’t. The environment was a major factor in the decision.
Before moving forward, I roughly storyboarded these ideas just to see them on paper. The hardest moment to figure out was how to show the override happening, and I eventually realised that lighting would play the biggest role in communicating that shift. After a few iterations, the flow of the story became much clearer, and it started feeling like something I could actually produce.
Once I had the main idea locked in, I moved on to creating a previs in Maya to understand how the whole sequence would actually play out. I knew the animation depended a lot on timing, camera flow, and how the audience would read the moment the override happens, so doing a rough block-out felt essential. I kept everything super simple by animating just a few poses in blocking, gathering some assets from Sketchfab to Maya to blockout the areas and animate better. Seeing the shots next to each other helped me realise which moments needed more tension, which ones felt too slow, and where I should cut earlier.
Working on the previs also gave me a clearer sense of how to place the cameras later. I experimented with different angles to see what made the lab feel more isolating and what best captured the shift in control when the mech starts overriding her commands. Even in this early stage, it helped me decide the emotional beats: when the viewer should feel calm, when the suspicion starts building, and when the shock hits. Having this rough roadmap made the next stages much smoother because I wasn’t guessing — I already understood the visual flow from start to finish.