worklife Real-world business for pro filmmakers
Hybrid power Jack Cotlevski’s workflow combines carefully captured footage with AI-constructed imagery – and here’s how he does it Words Jack Cotlevski, director of Chroma Films
ybrid cinema isn’t about replacing filmmaking with a computer. It’s about using conventional filmmaking for the
problem of the perfect light disappearing – the sky turns white or it rains on the one day the location and talent are available. Computer construction of the scene gives another option. And then there is something I call presentation movement. Imagine an orbit around a static subject whose only purpose is to reveal the scene. Traditionally, that could require a drone and its operator, securing permissions and potentially another shooting day. I distinguish that from movement which contains information. If movement is part of the performance or story, I capture it. If movement simply presents something to the viewer, I consider whether it can be constructed. That’s where hybrid production can be economically interesting. The savings can fall on some of the most expensive elements while reality remains precisely where the audience needs it. The importance of a plan The difference between convincing hybrid cinema and a collection of disconnected elements is usually decided before the shoot. I create a plan in three columns. The first is what stays real, the second is what will be constructed, but the third is the crucial one: how they will work together. If I intend to construct movement between two camera positions, I need to capture both at the same focal length and camera height. If I’m putting a real performer into a constructed environment, I need to shoot them in a way that allows a clean extraction. That influences the background, lighting, framing and lens choice. Ignore these considerations until post and you’re immediately fighting your own footage. The light comes from the wrong direction. You haven’t left enough room in the composition. The second
things it does brilliantly, then using image construction for the things it does better. I still arrive on-set with essentially the same equipment I would have brought ten years ago: cameras, lenses, tripod, lights, diffusion, flags, cables, colour chart, slate and monitor. There is still a crew, we still spend time setting up and I’m still just as tired when we wrap. Yet when you watch the finished film, you may see things none of those tools ever recorded. Perhaps it’s a sunset that didn’t happen on the day. Maybe the city continues beyond the buildings actually at location. Sometimes an entire moving sequence grows from a couple of photographs. It’s not camera vs computer Discussion around generative imagery often becomes unnecessarily binary: either you shoot something or generate it. Make cinema or make something else, for example. Reject the technology completely and you potentially lose a valuable production tool. But believe they can replace everything that happens on- set and you risk discovering that your beautifully generated film has no truth or emotional connection. The camera and image construction aren’t competing to do the same job, as they solve different problems. The useful question is, which tool solves each particular shot best? There are things I nearly always want to capture for real. The first is human performance. A face, glance, gesture, breath or the fraction of a second somebody takes before delivering a line contains an extraordinary amount of information.
Humans are incredibly sensitive to other humans. You can construct something that appears superficially correct but get the rhythm wrong. The audience may not consciously identify what’s wrong, but they simply won’t feel it. Then there is the product itself, particularly in commercial work. A client knows their product’s shape, materials, colour and finish because they see it every day. Change something subtly and they may not be able to explain precisely what is incorrect, but they’ll know it doesn’t look right. The same applies to recognisable architecture and locations. Reconstruct a company’s headquarters and make it objectively more beautiful, but if the owner doesn’t recognise the building, it’s a fail. None of that is about taking a moral position on what’s ‘real’. It’s really about understanding what an audience and a client notices. Those are the things I protect. Build the impossible Some problems are much better suited to AI-constructed images. Scale is an obvious one. Perhaps the shot requires an enormous set, expensive travel or a location that simply doesn’t exist. Then there’s the light and weather conditions. Every filmmaker knows the
08
September/October 2026
Powered by FlippingBook