Old Glass, New Light

Why the right premium vintage lenses and ray tracing saved us more in post than they cost on set

The wood-slats shot from June July's proof-of-concept shoot on the LED volume stage.

We recently ran a proof-of-concept shoot for our upcoming feature June July, directed by Daniel Thron and produced by Ron Cohen, Janelle Croshaw Ralla, AJ Kim, and myself. The point was to test something on an actual production rather than in theory: whether old technology and new technology, chosen deliberately, work better together than either does on its own.

One shot covers most of what we learned. Our lead stares through a gap in weathered wood slats at a woman on the other side of the wall. Neither one speaks. Everything in the scene happens in the eyes.

Conventionally, that shot is closer to three or four separate shoots than one. You put your lead in front of a blue screen and get her side. You do the same for the other actor. Anything between them, in front of them, or behind them gets shot separately or rendered as CG elements. All of it then goes to post, where a compositor assembles the pieces and tries to fake the focus pulling from one actor to the other, guessing at what that shift should look like months after the fact.

Behind the scenes on the LED volume stage during the wood-slats shot.

We shot it once. Both actors on set, looking at each other through the slats, reacting to each other in real time. The CG on the LED wall behind them was part of the scene rather than a backdrop waiting to be fixed later, balanced live along with everything else in frame. The focus shift between the two actors, normally faked and second-guessed in post, happened in camera and became part of the storytelling. We were shooting on Panavision Vintage P lenses, a choice from our director of photography, Richard Crudo, and the CG on the wall picked up the characteristics of that glass automatically. There was no lens-matching pass afterward.

Panavision Vintage P lenses, used for principal photography on the LED stage.

The location was closer than you'd think

The shot needed a real location behind our actors. We found one: a decaying structure out in Angels Camp, California. We didn't move the production there, and we didn't build it as CG from scratch either.

Not everything has to be generated in CG, and not everything has to be shot somewhere distant to chase a tax incentive. Our location was about 300 miles from the stage, a day's drive, still in California. We sent a small second unit, a skeleton crew with an Alexa 35 and a Panavision Primo lens, to scan it as a Gaussian splat and bring it back. The Primo was chosen for this job specifically. It's an extremely sharp lens, and sharpness was what we needed to capture as much real-world detail as possible for the scan. Back on stage, the Vintage P lenses would be doing close to the opposite job, which I'll come back to.

The second unit at Angels Camp, California.

The trip wasn't only about the scan. The same crew captured live-action establishing shots on location for the edit itself, shot as day-for-night, and we treated the scan with that same day-for-night look so the two would sit together without a visible seam. We had a few hours at dawn to capture light usable for that, which is exactly the kind of window a small crew is suited to. We brought that lighting back to the stage and spent a full day shooting the scene against it. The principle holds past this one shoot: if you need magic hour, or any similarly narrow lighting window, capture it once and bring the lighting back with you rather than scheduling around that window every time you shoot. The technical reasoning behind the day-for-night choices deserves more space than this piece allows, and we're saving that breakdown for the deep dive on the Chaos blog.

We also scanned the location with the same camera, an Alexa 35, that we used on stage. That kept the scan's color in the same world as the footage it would eventually sit next to.

Panavision Primo lenses, used to capture the Gaussian splat scan at Angels Camp.

Fire, without a fire marshal

The same shot called for fire. Traditionally that means a practical flame on set, which means a fire marshal, a safety crew, and a shoot day organized around managing an open flame near your actors.

We didn't need any of it. The real-time render was fully ray-traced, running on the newest version of Chaos Arena, so we could add CG lighting elements, including fire, and have the Gaussian splat respond to that light correctly, live, on the wall. The building lit up, shadows moved, surfaces caught the glow, all in camera. Our VFX team, Janelle Croshaw Ralla, Kevin Pierce, and myself, came away with light behavior on an actual surface to work from instead of a guess at how fire should fall across that building. What would have been a practical-effects and safety operation became a lighting decision made on the wall.

The fire sequence, lit and relit live on the LED wall.

A shot even VFX couldn't have predicted

Earlier in the same sequence, before the fire was added, we had a different problem: two focal planes needed to hold sharp in the same frame.

The answer was a split diopter, decades-old glass doing the job it has always done. What made it more than a technical fix was what came with it. Shooting through the diopter produced half-moon lens flares, an optical effect that would have been impossible to create in VFX and difficult even to predict beforehand. Nobody storyboarded it. It's in the film.

The half-moon lens flare produced by the split diopter, caught entirely in camera.

The split diopter used to hold two focal planes in a single frame.

A different scene, and the best-practices version of the story

The interior scene is a different case. No fire, no splat work, just the physical set our production designer, Arielle Ness-Cohn, along with her art department and construction coordinator Tom Early built on stage at ISS, with location support from Global Objects, meeting the LED wall standing in for the world outside. This is closer to the way most people use LED volumes, and it's a clean example of the fundamentals.

The exterior needed to read as night, so the wall stayed dark, mostly silhouette, sitting behind the frame instead of competing with it. We did the opposite on an earlier short called Ray Tracing FTW, set on a train, where we deliberately overexposed the background to sell a bright day outside, intentionally mismatched from the darker interior of the train car. The tool was the same; the story called for a different result.

Ray Tracing FTW (2024), © Chaos

There's also a specific optical reason the wall doesn't need to be pixel-perfect in every detail. We talked to Dan Sasaki, Panavision's SVP of Optical Engineering and Lens Strategy, about this. The Vintage P lenses are built from older glass with a much softer falloff and lower resolution than something like the Primo. On an LED wall, that's a practical advantage. A very sharp lens can resolve the individual LEDs as the wall comes into focus, producing a moiré pattern that fights the image, and the Vintage P's lower resolution makes that far less of an issue. So the Vintage P lenses gave us period-correct character and meaningfully more usable depth of field with respect to the wall, in exactly the situation where the Primo's sharpness would have worked against us. Exposure and color still matter most, but the lens choice is doing optical work underneath that, not only stylistic work.

The interior scene, balancing the physical set against the LED wall standing in for the world outside.

Fewer pointless revisions

By the interior scene, we weren't debating hypotheticals. We'd seen actors reacting to each other through the slats. We'd watched a set catch fire on the wall and light the room correctly. We'd caught flares off a split diopter that nobody could have storyboarded. Being able to see a shot working, on the day, cuts down the number of changes that get proposed for no particular reason.

Much of the endless-notes cycle in post exists because nobody in the room knew what the shot looked like when they made the call. Put the shot in front of everyone on set, actors, director, DP, and a lot of the "let's revisit this later" conversations don't happen, not because people stop caring but because there's less to argue about.

It's not a magic bullet

More than one shot from this proof of concept still needs revision, and there's a meaningful amount of post work ahead of us. In-camera didn't always hand us a finished film off the sensor.

At the very least, it gave us is a clear picture of what the final result needs to look like and what has to match. VFX and CG in post work best with real visual references to build against, actual light, actual color, actual behavior on a surface. The approach didn't eliminate post work entirely. It gave post a target instead of a blank page.

The actual case for spending more up front

A shot that would normally have needed three or four setups and heavy compositing, captured once. A location that would have needed a full production move, scanned by a handful of people in a few hours. A fire that would have needed a safety operation, handled as a lighting decision. An optical effect nobody could have designed in VFX, caught because we were rolling camera through real glass. None of it came from cutting corners. It came from spending earlier, on lenses and tools that let us capture things correctly the first time, rather than spending later to fix what was never quite right.

A word on the tools behind this

None of this works without the real-time, ray-traced rendering that let us light and relight the Gaussian splats live on set, built by our partners at Chaos, where Vladimir Koylazov championed the project from the start and Georgi Zhekov, Arena's product manager, helped us push the tools to handle it. We are preparing a deeper dive into that side of the story, the splats, the rendering pipeline, and the engineering that made a scanned location behave like a lit set. Check back here for the link when that post goes live.

All other images © 2026 Monstrous Moonshine.