The final performance of Autoplay, Africa's first AI opera, runs on June 29 from 7pm until 8.15pm at the Drill Hall in Cape Town. Created by Darkroom Contemporary founder Louise Coetzer, the production uses generative music software and real-time AI to shape both sound and movement. The cast must navigate two possible endings chosen by the audience, and the music shifts nightly - a test of how far live performance can push human-machine collaboration.
Coetzer started with research into AI's worldview. "Initially I started out with a lot of research and trying to understand AI's worldview," she said. "And in doing that it became clear that AI systems are not neutral. They have biases depending on the country or company that made them. None of them have neutrality." That discovery drove the piece's central question: what happens to autonomy when algorithms shape creative choices?
Training AI on past work
Rather than using AI to generate ideas from scratch, Coetzer fed it her previous projects. The system would produce variations - some predictable, some surprising. "The outputs were often predictable and simple, but sometimes surprising. It was helpful in the creative process and would lead my thoughts in different directions," she said. This critical engagement became the backbone of the opera's development, turning the machine into a collaborator that could be critiqued and steered.
The process fits within a broader AI for Creatives approach: using tools to amplify a concept rather than replace human intent. Coetzer said she wasn't drawn to AI because she lacked ideas. "For me AI wasn't used because I don't have ideas, it was used to amplify the concept. How could we engage this technology in ways it's not meant for?"
Choreography inspired by early generative art
The physical language of Autoplay draws from the strange, often deformed figures that early Generative Art models produced. "We took inspiration from early generative AI art, where you'd have this beautiful figure generated but a hand or foot would be deformed or out of place," Coetzer said. That aesthetic translates into movement that is erratic, bizarre, and deliberately inhuman - a response to the uncanny experience of interacting with non-human systems.
The cast, she said, had to be deeply committed. "The performers are really pushing themselves to the limit and bring a high level of emotion and intensity to it." The choreography is tightly blocked, but the soundscape is mostly fluid. Sections are generated in real time using generative music programmes, and audiences hear the transition from organic sound to machine-manipulated output.
Rehearsing for real-time uncertainty
The production has a detailed script with solid blocking and choreography, but the music changes constantly. "We kind of have a script of what each scene is about and all the blocking and choreography is solid," Coetzer said. "We then bring the music in, which is mostly always changing. There are set pieces, but sections are generated in real time using generative music programmes and software." The performers only learn which of the two endings will play out once the audience decides, so they must stay ready for either.
Why this matters for creatives
Coetzer's approach doesn't offer answers about AI's role in art. "The production isn't meant to give any answers. I don't have the answers, it's about what's actually happening," she said. But the work demonstrates a practical model: use AI for something it wasn't designed to do, and the result can be a live performance that no human would have made alone. For creatives weighing whether to adopt generative tools, Autoplay shows that the technology can extend a concept rather than dilute it - if you're willing to treat the machine as a partner with biases, not a neutral assistant.
Your membership also unlocks:
