We were shooting holograms in 2017.

Genlocked 4D capture, years before it had a marketing name.

Our first 4D camera array went up in 2017 — every camera on one clock, recording a performance from every direction at the same instant. We built it for HBO’s Watchmen and shot the show with it that year; its data drove the series’ hologram sequences. The reconstruction methods have changed three times since. The capture problem has not.

[ GENLOCKED / MULTI-VIEW 4K / IN THE ROUND ]

Performer standing inside SCANable's genlocked camera array, ringed by camera modules

WHAT A HOLOGRAM ACTUALLY IS

A performance you can walk around.

On screen, a hologram is not a physics demonstration. It is a volumetric record of a real performance — a person captured from every angle at once, so the shot can move the camera around them, light them differently, or drop them into a scene as a glowing projection without the performance falling apart.

That is why hologram work is a capture problem before it is a rendering problem. If the source is a single camera angle, everything the audience sees from another direction is invented. If the source is the whole volume, it is observed. Every technique below — lightfields, volumetric video, 4D Gaussian splatting — is a different way of reconstructing that same recorded volume.

2017 / THE FIRST ARRAY

One clock, every camera.

We built our first 4D array in 2017 because nothing available did the one thing the work required: genlock. Every camera in the rig runs off a shared clock and exposes on the same frame, to the frame, for the length of the take.

Without that, a multi-camera setup is not capturing an instant — it is capturing a smear. Angles land milliseconds apart, the subject has moved between them, and every reconstruction built from that data has to guess at the difference. You can see it in the result: soft hands, swimming edges, a face that subtly crawls. Genlock is the difference between a volume that was measured and a volume that was averaged, and it is the reason the 2017 data still holds up against methods invented years later.

Contact grid of synchronized camera angles from a single instant of a performance capture

WATCHMEN / HBO

The show that put it to work.

The array’s first job was Watchmen for HBO in 2017, where the capture drove the series’ holographic effects alongside the LiDAR set and location scanning we delivered for its set extensions.

It was the right first test. A hologram in a scene has to read as a recording of a specific person, in a specific costume, doing a specific thing — the moment it reads as a generic CG figure with a blue tint, the effect is dead. What sells it is the detail that only comes from having actually photographed the performance from every side.

Solved 4D facial performance geometry, untextured, mid-sequence

WHERE THE NAME CAME FROM

Why the rig is called C.A.L.V.I.N.N.

The acronym is honest engineering — Camera Array for Lightfields, Volumetrics, Imaging, and Neural Networks. The name it spells is a nod to the show. In Watchmen, the one character who experiences every moment simultaneously spends the season hiding in plain sight under an ordinary name: Cal.

It is the closest thing in fiction to what the rig does. Every camera fires together, so one instant is held from every direction at once instead of being assembled from angles captured moments apart. The name stuck before the spec sheet was finished.

2017 → NOW

The method changed. The capture didn’t.

In 2017 a captured volume became a mesh sequence or a point-cloud sequence, and holding it together over time was most of the work. Then came radiance fields. Then 3D Gaussian splatting, which made photoreal real-time playback of a captured volume ordinary instead of exotic. Now 4D Gaussian splatting — splats that move — is the technique behind the holographic sequences everyone is currently talking about, and the questions land in our inbox the week after each one releases.

Here is the part that matters commercially: none of that changed what has to happen on the day. 4DGS is a reconstruction method, not a capture method. It still needs synchronized multi-view video of a real performance, in the round, calibrated and clean. A rig that can feed 4DGS today is the same rig that fed mesh sequences in 2017 — which is why we have been shooting the input since 2017 and are not learning it on your show.

Gaussian splat reconstruction viewed in a real-time viewport

WHAT A HOLOGRAM SHOT NEEDS

Five things, or it won’t hold.

GenlockEvery camera exposing on the same frame. Without it you are averaging the volume, not measuring it.
Coverage in the roundEnough angles that no direction the shot might travel is being invented after the fact.
Calibration you can trustSolved camera positions and color-matched response, so reconstruction starts from geometry instead of guesswork.
RAW, not delivery codecCompression artifacts become geometry artifacts. What the array records is what the solve inherits.
Controlled, repeatable lightSo a performance captured on Tuesday cuts against one captured in another city on Friday.

[ 150+ CAM 4K / GENLOCKED / LIGHTFIELD + VOLUMETRIC / 4DGS-READY ]

START HERE

Got a hologram in the script?

Tell us the shot and the pipeline it finishes in. We’ll scope the capture around it — and bring the array to your stage.

Request a Quote