Three Thousand
Years of Longing

The bathhouse is an octagonal space, and production built half of it. For coverage, the crew would rotate props and redress the set rather than construct the other side. It was a practical decision that kept the shoot moving but left us with the job of completing the room.

We were given high-resolution on-set photography taken from multiple angles across the partial set. From that reference, we extended the space through each archway so that the full bathhouse read as a complete, continuous environment. No scan on this one, just careful matching work using the photography to carry the texture, colour and detail of the real set into the extended portions.

The other significant element in the bathhouse was light. Production had rigged practical spotlights to cast pools of light onto the floor, suggesting sunlight from above. Our job was to complete that idea. We added light rays connecting those floor pools back up to the windows, with shadow casting that moved correctly with the camera. Smoke and dust were layered through the rays to give them body and atmosphere. Tea cups and urns in the scene needed steam, so we added that as well. Small details that hold the warmth and intimacy the scene required.

The tunnel sequence was our largest environment build on the show. Production constructed a partial set: a few archways and a short section of tunnel. We extended it into something that feels genuinely ancient and seemingly endless.

We were given a LiDAR scan of the set alongside bracketed, high-detail texture photography shot on set. We retopologised the scan and retextured it using those photographs, extracting the surface detail of the real stonework. That asset was then repeated and extended through the tunnel ends, with lighting built to suggest depth. Pools of light in the distance, archways receding back into darkness.

As the camera moved through the space and peaked through archways, we extended each one as needed. Wherever the edge of the set was visible (ladders, crew, the physical limits of what was built), we patched it with our CG build. The goal was a tunnel that felt as though it had existed for centuries and went on considerably further than any camera could follow.

The Djinn, when he moves as a spirit, has no body. George Miller's term for the effect that captures his perspective was the Chrome Donut: the idea that as he passes through solid matter, the matter doesn't part or dissolve. It flows and warps around his point of view, as though reality is a surface being pressed through from the inside.

Realising this effect meant working with long, complex takes full of retimes, re-racks and stabilisations, and stitching them together into what reads as a single continuous move. The camera transitions from inside the tunnel to outside it, passes through walls and into other spaces, and the edit has to feel seamless throughout. Grading work in post pushed the look further for the final film.

Before we locked the approach, we did some practical R&D. We bought a borescope camera, rigged it into a test tube housing, and spent time pushing it through physical objects: buckets of sand, foam padding, cardboard. The idea was to understand what passage through matter actually looks like at close range, and whether we could capture elements that might feed into the effect. Some of it was useful. Some of it wasn't. Hitting the specific quality that George had in mind was genuinely difficult, and the experimentation was part of finding it.

One of the Chrome Donut shots involves a character in a red mask, kicking up red dust. The camera pushes in close to the mask (we had a scan of it to maintain detail at that proximity) then passes all the way through and comes out the other side. The character, now seen from behind, is rotoscoped and surrounded by a volume of dust that rises toward camera. That moment becomes the transition into the next shot.

Later in the tunnel sequence, the Djinn's perspective arrives at a rusty iron door covered in spider webs. The camera passes through it, door, webs, everything, as the Djinn would. Then the boy who has been following arrives, and he has to do the same thing physically: push the door open, break through the webs.

The door was a partial practical set piece, built just to the extent the boy needed to interact with it. We built a full CG door from a scan of the practical, matching it precisely. For the spider webs, production had shot physical web elements with fishing line threaded through them to allow controlled breaking. We took those elements and, using the LiDAR scan of the tunnel environment, layered them up into a volumetric web that the camera had already passed through invisibly. The challenge was timing the boy's hand movements against the web elements so that the interaction read as real, that he was genuinely pushing through the same webs the audience had just watched the camera pass through without resistance. It is, in a straightforward sense, an impossible shot. The Djinn can pass through anything. The boy cannot. VFX is what makes both things true in the same frame.

Beyond the door, the sequence stitches through another Chrome Donut transition into a new set: a space that needed to feel ancient, sealed, and undisturbed. We added extensive spider webs throughout. A dead animal in the scene required CG flies, built as a Houdini particle crowd simulation. Light beams and dust completed the atmosphere, pushing the set back in time and expanding it beyond its real dimensions.

There is also a shot where the boy's head becomes caught in the webs. For that we built larger, thicker web elements with debris worked into them, leaves and particulate, and used Nuke to warp and attach them to his head, working from the practical web elements production had provided.

Sugar Lump is a very large woman, and she needed to be larger still. The scale-up was 125 percent. She was shot on set with the rest of the cast, so the first task was isolating her through rotoscoping, a considerable undertaking given that she interacts physically with other characters throughout.

The most demanding interaction is with the king, who is seated on his throne wearing a fur costume. Sugar Lump leans over and strokes his shoulder as the camera pulls back from close. The scale-up had to hold through all of that movement without the contact point breaking. Her costume was semi-transparent, which complicated the extraction significantly. When scaling, the mosaic set piece behind her was visible through the fabric, and we could not allow that to distort. Pulling a clean extraction of a semi-transparent costume against a high-detail background, at scale, across a moving camera shot, was the central technical problem of this sequence.

In a second shot, Sugar Lump walks down the tunnel. We kept the ground near her feet at something close to its original scale to preserve her contact with the floor, then rotoscoped her from approximately the knees up, working around the semi-transparent costume. She was partially on green screen, which helped, but the transparency challenges were consistent throughout.

Alithea is a creature of the real world: a narratologist, a scholar, a woman who boards planes and moves through airports. The airport sequence is where the film's two registers meet, and our work there reflects that: precise, functional, invisible.

Production built a partial body scanner set piece, dressed with tracking markers and a monitor. The scanner itself was non-functional, with no moving parts. We built a full CG scanner, matching the geometry of the practical set, and animated the rotating glass and metal components so that the machine moves correctly as Alithea walks through it and back out. Screen replacement on the monitor completed the picture. Set extensions to the back wall of the airport, shot on blue screen, pushed the environment out to a believable depth.

A dialogue sequence in a van was shot with an LED volume behind it. Production shot three 6K plates of the environment at different angles to play on the screen. We stabilised, undistorted and stitched them into a single 18K file for LED playback.