When your physical set is three walls and a ceiling, whatever’s outside the window has to carry a lot of the film. For SNTIENT 2.0, the Fashion Innovation Agency’s film for London College of Fashion, that view was a bleak, abandoned city, and it lived entirely in Unreal Engine.

The story is a dystopia with a twist, set around the idea that resources have run out, humans have retreated into a simulation, and the robots left behind to look after the physical world have started to find beauty in it. “That gave us our visual starting point. A bleak, oppressive sci-fi environment with the potential for transformation,” says Dickon Knowles, director of content at Dandelion + Burdock. The finished film doesn't explore that transformation as far as originally planned, but the city still set the tone for everything else.

  • Read my impressions of being on an LED Volume stage.

Changing your mind on the fly

The big draw of Unreal Engine 5 was speed. Working in a real-time engine meant the team could play with spaces, lighting, and textures, see the result straight away, and change it on the spot, instead of waiting for a render only to find it was wrong. “It made the whole process much more fluid,” Knowles says.

Other routes were tried first. “We tried a couple of generative approaches to creating the city plate for the wall, but resolutions were an issue and the visual fidelity wasn’t there,” Knowles explains. “Which is why we reverted back to our existing Unreal Engine version.”

Even then, the city needed work. The original environment was foggy and slightly desaturated, and wasn't reading through the camera as intended. Knowles reworked the fog, replacing the general wash through the scene with more detailed VDB fog, while reducing its overall presence and boosting the saturation, particularly in the city windows. The LED wall looked better with richer blacks and more contrast, which the fog had been fighting against. It needed a more vivid backdrop to work on camera.

Most of the city was already built before the shoot, with the team making visual refinements, grading and subtle tweaks on set. It wasn't building a digital world from scratch while the cameras were rolling. The advantage was seeing how those changes looked on the wall and through the camera, and making adjustments without waiting for a new render.

Lighting the room from the window

The city also ran on an LED wall as the view from the character’s window, which gave the performer something real to react to rather than a green screen to imagine over.

Cinematographer Ben Saffer, whose company, Supervision VP, handled the virtual production, says the appeal was being able to capture everything together. The virtual and physical lighting could be controlled as one, so the glow on screen matched the light on set. “You’re doing it all in camera and capturing everything together at once, and you’re not having to guess at anything,” he says.

The alternatives were a green screen or a pre-rendered backdrop. Knowles says cost was less of a concern than visual quality, because the city content had already been created for a previous project. “The city lights reflect on the performer’s face and on the surfaces of the props and set,” he says. “There is a level of visual immersion that is hard to capture in a studio without the LED wall to illuminate the space.”

The LED wall wasn't simply providing something pretty to look at outside the window; it was lighting the performer and the physical set at the same time, similar to how Magnopus used Unreal Engine for Fallout. The intention was to keep post-production to a minimum, getting the footage as close to the finished shot as possible during filming.

A shutter made of pixels

Unreal also solved a budget problem because the film needed a shutter that rises to reveal the city, but a moving shutter built into the set wasn't realistic on the budget, and adding one in VFX would have meant losing that in-camera moment.

So the LED wall became the shutter, rising to reveal the digital city behind it. For reverse shots, a second wall on wheels could be brought in to act as a fourth wall. Because the content ran in real time, the team could quickly render new assets for those angles, animate them, and get them moving. “It was nice to get that in camera,” Saffer says.

Not everything went virtual, though, as the pod itself was a physical build, because the team wanted it to feel gritty and lived-in, with the LED wall and Unreal handling the distant environment. The AI-generated dream sequence at the start was a separate part of the production, created using generative tools to show the character floating in water before she wakes up.

  • Read how Unreal Engine and Houdini power the VFX of Surviving Earth

The film has very few beats: the character wakes, the window opens, she looks out. Saffer says the real challenge was getting a sense of the world and her place in it within two minutes. The team also had to make the most of the tiny space, shooting on an URSA Cine 17K 65 with prototype Atlas Lens Co. anamorphic lenses. The larger sensor meant Saffer could capture more of the set without resorting to an extreme wide-angle lens, which would have distorted the perspective.

There was still room for a bit of old-fashioned experimentation. Towards the end of the film, as the character looks towards the window, Saffer held a plastic Polly Pocket toy in front of the lens and shone his phone torch through it from the side to create an optical effect. He also played around with cling film and other materials in front of the lens.

See more about how SNTIENT 2.0 was made on Dandelion + Burdock. Visit Blackmagic Design for cameras.

Like this article? For more stories like this, follow us on MSN by clicking the +Follow button at the top of this page.