South African Director Donovan Marsh Bets on AI Filmmaking—And Says It’s Already Changing Jobs
South African filmmaker and director Donovan Marsh says he first knew the traditional movie-making era was ending when he watched pre-release AI clips generated by OpenAI’s Sora on his phone. What he saw didn’t feel like a novelty—it felt like a final verdict on conventional workflows, from production planning to visual effects and performance creation.
Speaking in an interview shared by TechCentral, Marsh recalled standing up in shock, convinced that conventional filmmaking would struggle to survive against tools that could generate moving scenes long before a camera ever rolls. “That’s it for my industry,” he said at the time, describing the moment as a genuine turning point rather than a passing trend.
In the years since, Marsh has leaned fully into AI as both an artistic approach and a business strategy. He co-founded Dragon Studios AI, alongside South African internet pioneer and indie film producer Ronnie Apteker, and London-based writer and video producer Steve Cholerton. The studio has already released award-winning AI shorts, and Marsh believes blockbuster-quality AI feature films are closer than many people think.
Donovan Marsh AI filmmaking: Why he believes the future is here
Marsh’s argument is built around speed, cost, and scale—three areas where AI tools are moving aggressively fast. He pointed to his experience on major productions, including the Hollywood submarine thriller Hunter Killer, starring Gerard Butler and Gary Oldman. That film, he noted, carried a US$40-million budget and required 55 days of shooting, plus a long preparation cycle before production even started.
After filming, the work shifted into post-production—especially visual effects. Marsh reminded listeners that around a thousand VFX shots had to be created, and each one could take two to six months depending on complexity. For him, that model represents precisely what AI is now able to disrupt.
“If you look at that same kind of project today, I could make it using AI for 100 times less money,” Marsh said. He argued that the big bottleneck—visual effects—can be handled far more efficiently with modern generation tools, meaning fewer months of manual cleanup and compositing.
His remaining concern is performance. AI can create images and sequences, but delivering consistent acting across a full 90 minutes—with the right emotional timing and believable nuance—is still a hurdle. Marsh doesn’t deny that challenge. Instead, he frames it as the final technical wall, adding that he can’t see why it wouldn’t eventually fall too.
Marsh also said that the tools that first caught his attention—OpenAI’s Sora and Midjourney—have since been overtaken. OpenAI, he noted, recently announced it was shutting down Sora, and while he didn’t treat that as a major surprise, his reasoning was blunt: the product wasn’t competitive enough and, according to him, it wasn’t widely used.
Instead of leaning on those early platforms, Marsh said he now relies heavily on newer generation systems coming from China, including Kling and Seedance. He dismissed Google’s Veo 3.1 as “a distant cousin,” at least compared to what he is using today. In practice, Marsh described working with eight to 11 different tools in a single project, selecting the best option for each part of the creative pipeline.
He also pushed back on the idea that AI filmmaking is somehow “less magical” because there’s no room full of humans improvising on set. Marsh compared AI collaboration to working with a foreign cinematographer: you reduce your own language to the essentials—what you want, what the scene needs, the intended style—and then you let another creative system interpret that vision.
In his view, simpler prompts often outperform overly detailed instructions. “The AI is the more creative one,” he said, arguing that filmmakers still provide direction, but the machine supplies much of the expressive output.
Job losses, music generation, and the disruption Marsh sees coming
Marsh was equally direct about employment consequences. He told TechCentral that AI filmmaking is already decimating jobs, and he believes the displacement is not limited to one role or one department.
He suggested there’s “no need any more” for graphic artists, storyboard artists, and composers once AI generation capabilities become mainstream. That statement is jarring, but Marsh insisted that music is already past the quality threshold many critics feared. He said AI-generated tracks can sound like work you would normally associate with a top composer, and importantly, music creation and licensing are among the most expensive parts of building a film.
Beyond creative teams, Marsh predicted that other roles will be pulled in next: location scouts, on-set crews, actors, producers, and eventually even directors. His most striking scenario is that AI agents could produce personalised feature films on demand, tailored to individual viewers’ preferences.
That future, he said, would make “people like me” unnecessary—not because filmmaking stops existing, but because the process becomes automated enough that fewer human roles are required to deliver a finished feature.
Still, Marsh sees a different angle for South Africa. He believes local disruption could be beneficial, not only destructive. He argues that the South African film industry has struggled for years, and a major reason is simple: local filmmakers often can’t compete with international studios on budget, equipment, and production experience. That creates a cycle where fewer big opportunities come in, making it harder to grow.
From Marsh’s perspective, AI breaks that cycle by letting smaller teams create productions with a visual quality that previously required Hollywood-scale resources. “The gatekeepers are gone,” he said, describing the moment as an industry reset where new entrants can compete more easily—and where talent becomes the real differentiator.
Film schools urged to pivot—or risk producing unemployable graduates
Marsh said that one of the most concerning parts of this transition is not only whether AI will grow, but whether education will adapt quickly enough. When he lectured third-year film students recently, he asked how many were using AI tools for projects. He claimed not a single hand went up, which he described as shocking.
His message to film schools was uncompromising: pivot toward AI or risk graduating students who will struggle to find work. In his framing, “my industry is growing” while another industry is shrinking, and students need the skills that match where production is moving next.
At Dragon Studios AI, Marsh described the operation as still in a proof-of-concept stage, with early projects focused on short-form content while the team prepares for AI-generated features. Apteker, he said, brings experience from a range of film projects, including executive producing works such as Jerusalema and Material. Apteker’s personal story also reflects the realities of disruption in wider life—he lived in Kyiv during Russia’s invasion and later relocated to London for family safety. Cholerton’s portfolio includes documentary work for organisations such as Reuters, Unesco, and the World Economic Forum.
One of the studio’s first major projects, Dragon Hunter, follows a township boy who discovers a rare fossil worth millions. Marsh noted that the project has its first ten minutes online and has earned awards at AI-focused film festivals. Another upcoming work close to his heart is an AI adaptation of Shakespeare’s Julius Caesar, set in Jacob Zuma-era South Africa, which he described as full of political parallels that are hard to replicate in any traditional setting.
Marsh also offered a bold quality prediction. When asked how long it will take before AI-produced films sit alongside Oscar winners, he said he expects it could happen this year or next year, and even suggested audiences might start preferring them over some award-winning human-made films.
He knows it’s a bet—maybe an aggressive one—but Marsh is making it anyway. For him, the question isn’t whether AI will change filmmaking; it’s whether the industry will adjust fast enough to remain competitive, creative, and employable in a world where tools can generate entire film worlds at speeds and costs that once felt impossible.