Frequently Asked Questions

Meridian 4D is a fully mobile, battery-powered 4D volumetric capture stage built by Radiant Images. It captures a performance from every angle at once and turns it into an explorable 3D asset you can view from any perspective — not just the one a single camera saw.

Unlike traditional volumetric stages that are fixed installations, Meridian is untethered and modular: it packs into airline-checkable cases, deploys on location the same day, processes in real time, and reshapes to almost any size or layout on the fly. The result is studio-grade capture in a footprint you can fly anywhere in the world.

Gaussian Splatting is the rendering technique behind Meridian's explorable assets. Instead of building a scene from polygons and textures, it represents the captured subject as a vast cloud of tiny, soft, colored 3D points — "splats" — that together reproduce the look of the original photography from any angle.

The payoff is photoreal, view-dependent results — soft edges, reflections, and fine detail — that render in real time on everything from a workstation to a headset. Meridian generates Gaussian Splats (and Bullet Time) on-device for instant preview, and our final pipeline produces temporally stable splats that hold up cleanly frame to frame.

The name has two parts:

• "Gaussian" — each building block is a 3D Gaussian: a soft, ellipsoidal blob that's brightest at its center and fades smoothly toward its edges, following the bell-curve (Gaussian) function. Each one carries a position, a shape/size, a color, and a transparency. That soft falloff is why splats blend seamlessly instead of looking like hard dots or polygon edges.
• "Splatting" — a long-standing computer-graphics term for rendering by projecting a 3D primitive onto the 2D screen so it leaves a soft footprint that blends with its neighbors — a bit like tossing a soft paintball at glass. The technique renders a scene by "splatting" millions of these Gaussians together.

Terms you'll hear:

• Gaussian Splatting (GS) — the general technique, and the representation it produces.
• GS3D / 3DGS — 3D Gaussian Splatting: a static, single-moment 3D scene.
• GS4D / 4DGS — 4D Gaussian Splatting: a moving version that adds time as the fourth dimension, so the splats change over time to reproduce motion — a performance you can play back and move around in.

You'll see both orderings — "GS3D / GS4D" and "3DGS / 4DGS" — used interchangeably for the same thing.

In plain terms: it turns many photographs of a subject into a cloud of millions of tiny, soft 3D blobs that, viewed together, look exactly like the real thing from any angle. Step by step:

1. Capture the subject from many synchronized angles at the same instant — exactly what Meridian's camera array does.
2. Solve the cameras. The software works out where each camera was and builds a rough 3D point cloud from the overlapping images (photogrammetry / structure-from-motion).
3. Seed the splats. Each point becomes a 3D Gaussian — a translucent ellipsoid with a position, size, orientation, color, and transparency. The color can shift with viewing angle, which is what makes reflections and highlights look right.
4. Optimize. The system renders the splats from each known camera position, compares the result to the real photo, and nudges every splat's parameters to reduce the difference — repeating for thousands of iterations, adding splats where detail is missing and pruning ones that aren't needed.
5. Result: a splat cloud that renders in real time from any viewpoint.

For 4D (moving performances), this runs across frames with temporal consistency, so the splats track smoothly through time instead of flickering.

Yes and no — and the distinction really matters for production work:

• It uses machine-learning optimization. The splats are "trained" with gradient descent — the same math that trains neural networks — to learn the best representation of your scene. In that sense, yes, it's AI/ML.
• It is not generative AI. It doesn't invent, hallucinate, or imagine detail. Every splat is optimized to match the actual photography your cameras recorded, so the output is a faithful reconstruction of what really happened on set — which matters for VFX continuity, likeness, and clearance.
• It's also not a "black box" neural model. Unlike NeRF, which stores a scene inside a neural network, Gaussian Splatting stores it as explicit 3D splats you can render directly and fast.

Some steps around the core reconstruction may use AI assistance — for example denoising footage shot at high ISO, segmentation and clean-up, or upscaling — but these support the capture rather than fabricate content. Bottom line: it's data-driven reconstruction of your real capture, not AI-generated imagery.

First, an important clarification: they overlap. Gaussian Splatting is itself a (newer) way to produce volumetric video, so it isn't strictly one-or-the-other. The useful comparison is between the traditional mesh-based approach and the splat-based approach:

• Traditional volumetric video reconstructs each frame as textured 3D geometry — polygon meshes (sometimes point clouds or voxels) with mapped textures. It's mature and integrates cleanly with classic CG pipelines (rigging, relighting, collision), but it can struggle with fine detail, hair, transparency, reflections, and soft edges, and often needs heavy data and manual clean-up.
• Gaussian Splatting represents the scene as soft, view-dependent splats instead of a mesh. That tends to look more photoreal — handling hair, soft edges, and reflections gracefully — and renders efficiently in real time, including on headsets. The trade-off is that splats are newer: they behave more like "captured light" than editable geometry, and tooling for editing, rigging, and relighting them is still catching up.

Meridian can deliver Gaussian-Splat output for maximum photorealism, and we'll advise on the right representation for your downstream pipeline.

Traditional cameras record a flat, fixed viewpoint. Volumetric capture records the subject as 3D geometry; "4D" simply means that 3D capture is moving through time — a living, explorable performance rather than a static scan.

Because the moment is reconstructed in 3D, you can move the camera anywhere after the shoot — orbit it, dolly in, freeze it for a bullet-time move, or place a viewer inside it. In technical terms it delivers true 6DoF (six degrees of freedom): the audience can move and look in any direction. One capture produces infinite views.

360-degree video (also called spherical video) is filmed from a single fixed point — one camera or rig records everything around that spot. The viewer can look around in any direction, but can't move through the scene: lean or step and the image doesn't respond. It's a flat panorama mapped onto a sphere, so it's 3DoF (rotation only).

Volumetric video captures the scene as real 3D geometry from many viewpoints, so the viewer gets full 6DoF — they can look around and move through and around the space, with correct parallax (near objects shift against far ones as you move). In short: 360 video lets you look around from one fixed point; volumetric lets you move around inside a real 3D space.

It's captured by recording the subject at the same instant from many synchronized cameras placed all around it. Those overlapping views are then reconstructed into a moving 3D representation — a mesh, or in Meridian's case Gaussian Splats — using photogrammetry and multi-view reconstruction.

Three things make it work: enough cameras with good coverage and overlap; precise synchronization (genlock and timecode) so every camera captures the exact same moment; and calibration so the system knows where each camera sits. Meridian does exactly this — an array of nodes, each holding several iPhone cameras, records the performance from every angle and processes it into explorable 4D assets.

Motion capture records only movement — the positions of markers or joints over time — which is then used to drive a separate 3D character model you build elsewhere. You get the performance's motion, but the appearance (face, skin, wardrobe) comes from a CG model, not the capture.

Volumetric video captures the complete visual appearance and the motion together: the real person's face, body, clothing, and movement are recorded as they actually looked and reconstructed in 3D — no separate character model or rigging required. In short, mocap gives you data to animate a model; volumetric gives you the real performance, fully reconstructed.

Most volumetric stages are permanent installations: large purpose-built rooms ringed with dozens or hundreds of machine-vision cameras, hard-wired to racks of capture and processing hardware, in a fixed location and a fixed shape. They produce excellent results, but you have to bring the production to the stage, and reconfiguring or relocating them is slow and costly.

Meridian inverts that model. It's a mobile, modular, battery-powered array that you bring to the production — any location, reshaped to any size on the fly, deployed and capturing the same day, processing in real time and fully untethered. You trade a fixed mega-array for go-anywhere flexibility without giving up studio-grade image quality.

Crucially, the entire system is automated, and each mobile device acts as a distributed micro edge computer — capturing, processing, and offloading on board rather than streaming raw data back to a central rack of hardware. The impact is a workflow that's simpler and faster, with far less demand on every resource: power, footprint, mobility, and crew.

Meridian is built around a different premise than most volumetric offerings. Where many companies run a single fixed studio you have to book and travel to, Meridian is a mobile system we bring to you. A few things set it apart:

• Truly mobile and untethered: battery-powered with no fixed installation — it packs into 28 airline-checkable cases, deploys and captures the same day, and travels worldwide.
• Real-time, on-device processing: we generate Bullet Time and Gaussian Splat previews on set and can evaluate a take within minutes, rather than waiting days to learn whether a capture worked.
• Adaptive volume: the stage reshapes to almost any size or shape — from a single performer to a large group or environment — instead of a one-size studio.
• Flexible camera approach: the iPhone-based array delivers deep depth of field and a fully wireless workflow, and the system is camera-agnostic (RED, ARRI, Sony, Canon, and Blackmagic) when a project calls for it.
• One capture, every format: a single Meridian capture scales across cinema/VFX, immersive/XR, and live broadcast from one pipeline.
• Decades of experience: Radiant has 20+ years across cinema, immersive, and live, and has worked alongside the field's first professional volumetric companies — so Meridian is backed by deep, hands-on systems building, not a single product.

The result is studio-grade volumetric capture with the flexibility, speed, and reach that fixed installations can't match.

Both. Meridian is built to capture people and environments, and its adaptive volume scales to any shape, size, or form. You can wrap a single performer, a group, a set or location, or objects and props — we simply tailor the node layout to whatever you're capturing.

The same best practices apply across subjects: good coverage and overlap, and avoiding large reflective or transparent surfaces. For bigger environments we plan camera placement to balance full coverage against the detail trade-off that comes with larger volumes.

The standard Meridian configuration uses Apple iPhone 17 Pro cameras. Each one is far more than a camera: it's a camera, recorder, wireless transmitter, and battery in a single device — capturing 48 MP class imagery and recording professional codecs on-device, so capture, processing, and upload all happen right at the camera. That's what makes a fully untethered, real-time array possible, with no web of cables or external recorders.

Compared with the machine-vision or cinema cameras most volumetric stages use, the iPhone wins on three fronts:

• Depth of field (the big one). The iPhone's comparatively small sensor produces a naturally deep depth of field — far more of the subject stays in sharp focus at a given aperture than with a large cinema or full-frame sensor. Volumetric reconstruction wants the entire subject crisp, not a shallow plane of focus with soft falloff. Getting equivalent depth of field out of a big-sensor camera means stopping the lens way down and pouring in much more light.
• Self-contained. Each iPhone bundles the camera, recording, wireless sync, and power into one device — so there's no external power, cabling, genlock, or recording infrastructure to wire up for every camera, which is exactly what makes traditional arrays heavy, complex, and slow.
• No compromise on quality. The iPhone 17 Pro records pro codecs (ProRes 422 HQ / ProRes RAW HQ), open gate, up to 60 fps, in Apple Log 2 HDR — so you get the depth-of-field and logistics benefits without sacrificing image quality.

Most existing volumetric stages rely on machine-vision cameras (and sometimes cinema cameras), and they all hit the same walls: lengthy setup and poor mobility. We can still build a Meridian around cinema cameras when a project demands it (see the next question), but the iPhone configuration is what unlocks same-day, untethered deployment.

Yes. Meridian is camera-agnostic. Radiant offers configurations built around motion-picture cameras including RED, ARRI, Sony, Canon, and Blackmagic, alongside the iPhone-based system. The trade-off is mobility and setup time: cinema-camera arrays can push maximum image quality, while the iPhone system delivers the deep depth of field, wireless workflow, and same-day deployment Meridian is known for. We'll recommend the right balance for your project.

Beyond the mobile system, Radiant also operates a permanent, cinema-camera-based 4D volumetric stage as part of our HQ studio operations, which we can make available for projects that call for a fixed, studio-based setup.

Meridian captures 4K open-gate (~4:3 aspect ratio) imagery with wide-gamut HDR color, giving the maximum pixel area and creative flexibility for volumetric solves.

Four full-frame-equivalent lenses are available per camera: Ultra-Wide 13mm (f/2.2), Main 24mm (f/1.78), Main Crop 48mm (f/1.78), and Telephoto 100mm (f/2.8).

The image quality is genuinely cinema-grade. Each iPhone 17 Pro records ProRes 422 HQ and ProRes RAW HQ (10- to 12-bit) in wide-gamut HDR with an Apple Log 2 profile, up to 60 fps, and around 13 stops of dynamic range. That log/RAW pipeline gives real latitude to grade, pull detail from highlights and shadows, and match the look to other cameras on the shoot — which is what makes it viable for high-fidelity volumetric work.

The honest trade-offs versus a large-sensor cinema camera:

• Sensor size & focus: the small sensor's deep depth of field is an advantage for volumetric capture (the whole subject stays sharp), but it gives you less of the shallow, cinematic background blur — and less selective-focus control — than a large sensor.
• Low light: a smaller sensor gathers less light, so very dim scenes are noisier than on a cinema camera. We light the scene where we can and, if we have to raise ISO, budget extra denoising time in processing.
• Dynamic range & highlights: ~13 stops is strong for this format but a notch below the newest cinema cameras, and small-sensor highlight roll-off is less forgiving — so we expose carefully and protect highlights.
• Lenses & aperture: fixed focal lengths and apertures (no iris control), so framing and exposure are handled through lens choice, lighting, and camera placement rather than on-lens adjustment.
• Computational processing: phones can apply sharpening and noise reduction; shooting ProRes/Log keeps the signal clean and film-like with minimal baked-in processing.

For volumetric capture, the iPhone's deep depth of field, wireless workflow, and cinema codecs outweigh these trade-offs. Where a project demands the absolute maximum image quality, we can build the array around RED, ARRI, Sony, Canon, or Blackmagic cameras instead.

Every camera is locked together with genlock and timecode, and the system performs pixel-level alignment through precision calibration. That synchronization is essential for volumetric reconstruction — the solve depends on every camera seeing the exact same instant from its own angle.

The QR code markers are fiducial reference points placed around the capture volume, and they're how we calibrate the camera array. Before a take, the system reads the markers from every camera to solve each camera's exact position and orientation in 3D space, and to lock in a shared coordinate system, origin, and real-world scale.

That calibration is essential to the solve — it's what lets dozens of separate camera views align precisely into a single, accurate 3D reconstruction (the pixel-level alignment Meridian is built on). Because each marker is unique and machine-readable, the cameras detect and identify the reference points automatically and reliably, which keeps setup fast and the alignment robust — even as you reshape the volume on location.

Yes — and it's worth unpacking what that means. HDRI image-based lighting (IBL) uses a high-dynamic-range image of a real environment — a 360° map that records light across a very wide brightness range — as the light source. Instead of a few hard fixtures, the subject is bathed in the full spread of soft, directional light from that environment, so the result looks natural and can be matched to the scene the capture will ultimately live in.

On Meridian this lighting is built in, wirelessly controllable via DMX, and battery-powered, like the rest of the system. That lets you shape consistent, even, reconstruction-friendly illumination around the volume — without running mains power or lighting cable to the stage — which is a big part of what makes same-day, off-grid deployment possible.

The individual mounting systems that hold the cameras are called nodes. You arrange them around your subject to build the capture volume. Each node measures roughly 2 ft 8 in wide × 2 ft 8 in deep × 8 ft 4 in tall and rolls on heavy-duty 4" custom leveling caster wheels built for most environments.

It depends on how much of the subject you need to wrap and how large the capture area is. As a guide:

• Full 360° wraparound: we recommend our standard 14 nodes, covering a minimum volume of about 12 ft round up to a maximum of about 60 ft round.
• Smaller setups: we've had great success with as few as 4 nodes for a 90° solve, and 8 nodes for a 360° solve.

That 60 ft guideline is based on our standard 14-node configuration — spread those same cameras across a much larger area and you retain less detail in the solve. If you need to capture a bigger volume, the fix is simple: we add more nodes to keep coverage and detail high.

If we know your scene in advance, we can dial in the right configuration.

We recommend at least 3 cameras per node, and as many as 9 per node. Adding cameras improves the solve, but there's a point of diminishing returns beyond which extra cameras won't make a meaningful difference. The right number depends on the scene and how much vertical coverage and overlap you need.

There isn't a fixed formula yet — the technology is developing rapidly, and the answer depends heavily on what you're capturing. One person versus four people in the same stage size can require very different camera counts and arrangements, mostly because of occlusion (people blocking the cameras' view of one another).

The best approach is to share your scene ahead of time. We can run internal tests and generate synthetic data in previs to guide the configuration before a single case ships. If you know you'll need to zoom or dolly into a region, we can add tighter-lensed cameras on that area to preserve detail.

Each camera attaches with a quick-release, toolless mount, and because the whole array — cameras and their wireless sync alike — is 100% untethered and battery-powered, we can place and reposition cameras almost anywhere with no cabling to route.

That flexibility lets us integrate the cameras into small truss systems, mount them on the walls of a room or stage, or deploy them across a large space like a stadium — whatever the scene and venue call for. With no fixed rig to build and nothing to wire, installation and integration could not be easier.

Yes. Because the system is 100% untethered and battery-powered — with no cables or external recorders tying it down — it can be rigged to a moving vehicle. And since we can control and monitor the entire array in real time from anywhere in the world over the 5G cellular network, the rig doesn't need a hardwired operator station alongside it. We'll work out the right mounting and stabilization for your vehicle and your shot.

This is currently on our roadmap. We're doing extensive testing to find drone-mounting solutions that hold to Radiant's standards for quality and reliability before we offer it. If drone-based capture is important to your project, reach out and we'll share where things stand.

Yes. We've developed a fully carbon-fiber mounting-arm system whose arms fold into each other for fast, single-person, backpack-portable deployment. It's called Parallel. For more information, contact the Radiant Images team.

Yes — audio capture is available when your shoot needs it. The system records multi-channel, wireless 32-bit float audio for up to two people. The 32-bit-float format captures an extremely wide dynamic range, which keeps dialogue clean and makes levels very forgiving on set. Let us know if you need sync sound and we'll configure the audio channels for your shoot. If additional audio support is required, the Radiant team can assist.

Meridian was designed to fly as standard checked airline baggage. The entire system packs tightly into 28 Pelican cases and can be shipped anywhere in the world.

For ground moves, there's also a custom mounting solution for the inside of a standard 16 ft high-clearance box truck with a lift gate. Shipped that way, the system travels mostly pre-built and can be ready for capture within 2–4 hours of arrival.

Yes. The Meridian is highly portable and works internationally. We've completed several feature films and live broadcasts where the system traveled overseas as hand-checked luggage on commercial airlines, and others where we shipped it in a cargo van.

Because it's self-sufficient — battery-powered, untethered, and packed into airline-checkable cases — our crew and equipment can deploy and capture in almost any country in the world.

We handle the travel logistics, but the transportation costs — for both equipment and crew — are the client's responsibility. Contact the Radiant Images team and we'll help plan the logistics and provide cost estimates for equipment and crew transportation.

Yes — going anywhere is the whole point. Meridian is untethered, battery-powered, and self-sufficient, with built-in lighting, so it doesn't depend on a studio or wall power. We've captured everything from controlled stages to near-blackout environments.

Outdoor shoots are possible too; the main considerations are managing changing daylight, weather, and very bright or harsh light, which can introduce artifacts — so we plan lighting and timing with you. Indoors or out, the system deploys and captures the same day.

Often, yes. The system is adaptive and scales down to a small footprint, so it can set up in a large room, office space, garage, studio, or warehouse. A full 360° volume is about 12 ft across at minimum, and smaller partial-wrap configurations (for example, a 4-node 90° solve) take even less room.

The main thing to check is ceiling height. Each node stands about 8 ft 4 in tall, and average ceilings — in both homes and offices — are around 9 ft, which comfortably fits a full-height node with a little room to spare. The main exceptions are spaces with lower or sloped ceilings, so it's worth confirming the height of your specific room. Access is rarely an issue — the system breaks down into 28 Pelican cases that move through standard doorways.

Tell us the room dimensions and ceiling height and we'll confirm the right configuration, or adjust the layout to fit your space.

As few as 2 people can build and assemble the full stage. If it needs to be ready by end of day or sooner, we recommend 3–4 people, who can unpack and be shoot-ready in under 6–8 hours.

When the system arrives pre-built in the box-truck configuration, setup drops to roughly 2–4 hours. And because Meridian deploys the same day you land on location, there's no multi-day stage build.

Meridian runs with a small crew throughout. A two-to-four-person team handles the build and teardown (see the setup question for timing), and during the shoot itself a lean Radiant crew handles calibration, capture, and on-set processing.

Because the system is automated and each device is a self-contained capture-and-processing unit, you don't need the large technical crew a traditional hard-wired stage demands. Exact crew scales with the volume size, shoot length, and processing turnaround — smaller partial-wrap setups need even fewer people — and we'll confirm the right team when we scope your project.

The system is self-sufficient on standard cine batteries — regular V-Lock or Anton Bauer Gold Mount packs. Depending on how many accessories are connected, expect roughly 3–4 hours of runtime per charge, with built-in lighting also powered by the system. Swapping batteries keeps you running indefinitely with no wall power required.

Heat management is handled in layers. Each iPhone uses Apple's built-in vapor-chamber ("liquid") cooling, and Meridian adds active cooling on top: add-on gaming-style radiators and fans, plus thermoelectric (Peltier/TEC) heatsink-and-fan units that clamp or MagSafe onto the back of each device. Together this keeps cameras recording reliably through long, high-bitrate takes.

A lot — volumetric capture means many camera streams recording at once, so plan storage and offload accordingly. Per-stream rates for a single iPhone camera:

Total system bandwidth = per-stream rate × number of cameras, plus about 20% overhead — so a multi-camera, 14-node stage adds up fast. Because Meridian processes and uploads on-device in real time, you're not stuck downloading everything before you can work, but it's still worth sizing your media and transfer pipeline in advance. We'll help you budget storage for your specific configuration.

Meridian is flexible across a wide range of conditions, but a few guidelines produce the cleanest reconstructions:

Lighting: Avoid harsh back lighting, which tends to bake into edges and create artifacts. If you want a dark look, it's usually better to light the scene properly and darken it later with LUTs rather than shoot in true darkness — though we've successfully captured even near-blackout moments (e.g. a nightclub project). When in doubt, more soft, even light is the safe bet.

Wardrobe: Most clothing is fine — even dresses with small shiny details. Try to avoid large reflective or transparent surfaces, which is where artifacts appear: glasses (just pop the lenses out and keep the frame), mirrors, and big shiny objects like a glossy guitar (a quick matte spray helps). Small shiny accents are fine. For sheer or semi-transparent fabrics, keep the lighting extra soft so it isn't baked in.

Note that for some looks — a hologram effect, for instance — minor artifacts may be invisible or even add to the effect. Our on-set processing lets us evaluate all of this within minutes of capture and adjust on the spot.

It depends on the action, but for crisp motion in fast fight choreography we generally recommend a very narrow shutter: around 10° or faster at 60 fps, or under 4° at 24 fps. A narrow shutter freezes each frame, which the solve relies on for sharp reconstruction.

The trade-off is light: narrow shutters need more illumination, or higher ISO in a pinch. If we have to push ISO, we plan for extra denoising time in processing. We'll work through the right balance for your specific action ahead of the shoot.

Meridian processes on-device, in real time, with no need to download footage first. On set you can trim clips, color grade, transcode, and upload — every step on the device itself — and generate real-time Bullet Time and Gaussian Splat previews.

That means we can evaluate a take within minutes of capturing it and make creative or technical decisions on the spot, rather than discovering issues days later in post.

Reducing flicker and temporal artifacts is a core part of our final processing pipeline. At a high level, we process frames in batches rather than one at a time — which lets us track the splats through time and remove flickering, so a moving 4D capture stays smooth and stable from frame to frame. There's a lot more going on under the hood.

Review runs in stages, each one adding confidence and fidelity:

• Capture & sync confirmation — under 30 seconds. We confirm that every camera captured and that we have 100% common sync across all cameras, so you know immediately the take is usable.
• 3DGS preview & quality check — 3–5 minutes from capture to review. We process a 3D Gaussian Splat solve and confirm there are no technical issues, such as gaps in coverage, and assess the overall quality of the solve.
• 4DGS final asset — about 24 hours. The full, temporally stable 4D Gaussian Splat takes longer to process; for footage under three minutes, plan on roughly a day, with longer pieces scaling up from there.

With advance planning — knowing your needs 2–3 days before the footage upload so we can scale our team — we can often commit to 2–3 days for final asset delivery. Sharing your creative vision early (floor removal, LUTs, color scheme, etc.) helps us plan accurately.

You receive explorable volumetric assets — Gaussian Splat sequences and Bullet Time / free-viewpoint clips — along with access to the underlying high-quality footage (4K ProRes 422 HQ / ProRes RAW HQ).

These assets are designed to slot into cinema (VFX), immersive/XR, and live-broadcast workflows, so we deliver in formats suited to your pipeline and target platform. Tell us your tools and destination — game engine, compositor, headset, or broadcast — and we'll confirm exact deliverable formats and any conform steps up front.

Resolution: A Gaussian Splat isn't a fixed-pixel image like a video frame — it's a 3D representation that's effectively resolution-independent. It renders in real time at whatever resolution your target needs, from a headset's native display to 4K and beyond on screen. What actually drives fidelity is the splat count/density and the resolution of the source capture: Meridian shoots 4K open gate, and more (and denser) splats plus tighter-lensed coverage on the area you'll push into yield finer detail.

Because that detail comes from the capture, zooming into a small region is limited by how many capture pixels covered it — so we plan lens choices around where you intend to dolly or zoom.

File formats: Gaussian Splats are most commonly delivered as .PLY (the de-facto standard for storing 3D Gaussians), with compact or streaming variants such as .splat or .spz when a target platform calls for them, and as splat sequences for 4D/moving captures. The underlying footage is ProRes (QuickTime MOV). Because splat tooling is evolving fast and varies by engine and headset, the next question covers the specific viewing options for each platform.

We deliver Gaussian Splats ready for the platform you'll actually view them in:

• VFX / DCC tools: splats brought into 3D and compositing pipelines as .PLY (and, increasingly, USD) for tools like Houdini, Blender, and Nuke — or, when you just need a finished shot, pre-rendered camera moves delivered as EXR / ProRes image sequences to drop straight into comp.
• Unity: delivered for Unity's Gaussian Splatting renderers, so the asset runs in real time in your project across desktop, mobile, and XR (OpenXR / Quest, etc.).
• Unreal Engine: delivered for Unreal's Gaussian Splatting plugins, for real-time use in virtual production, in-engine cinematics, and XR.
• Web player: a WebGL / WebGPU viewer that plays in a standard browser — no app install — across most mobile, desktop, and WebXR-capable headsets (HMDs). It's the easiest "send a link, view anywhere" option; common formats include .splat, .ply, .spz, and engine-specific compressed variants (e.g. PlayCanvas).

The web player is usually the simplest way to share a capture for review or audience delivery, while the Unity, Unreal, and VFX deliveries are for teams building the splat into a larger production.

Please contact Radiant Images and we'll walk you through finished projects. We can also provide sample RAW data as well as Gaussian Splat processed footage, so you can evaluate the output for your own pipeline.

In film, TV, and commercials, a Meridian capture becomes an explorable 3D asset the production controls in post. Instead of locking in one camera angle on the day, you can re-frame, orbit, dolly, and create Bullet Time or "frozen moment" moves after the shoot; drop a real performance into a virtual environment for VFX; or generate free-viewpoint replays.

Because the asset is photoreal and view-dependent, it cuts together with live action and lowers the cost and risk of reshoots — you captured every angle once. The same capture also scales across cinema, immersive, and live broadcast from a single pipeline, so content made once can travel across formats.

Meridian captures a real person's performance — face, body, wardrobe, and motion — as a volumetric asset, rather than reconstructing it from a stylized 3D model or a marker-based mocap rig. That makes it valuable anywhere you need an authentic human likeness in 3D: digital doubles and de-aging for film, virtual production, performance and likeness libraries, and sports or dance analysis.

Multi-camera, genlocked capture preserves the timing, expression, and fine detail that sell a genuine performance — the nuance that hand-animated or approximated characters tend to lose.

The same explorable, photoreal assets are powerful for training and simulation, where realism drives how well skills transfer to the real world. Captured human performances and environments can be dropped into simulators and procedural training scenarios — medical, industrial, defense and first-responder, aviation, sports — so trainees rehearse against true-to-life people and spaces they can view from any angle.

Because Meridian is mobile, you can capture real subjects and real locations in the field and bring them into the simulation, rather than approximating them with built-from-scratch CG. That shortens content production and raises fidelity at the same time.

Gaussian Splats are well suited to VR, AR, and mixed reality because they're photoreal and render efficiently in real time, and because volumetric capture delivers true 6DoF — viewers can move through and around the scene, not just look at a flat 360° video from a fixed point.

That enables immersive experiences that place audiences inside a captured moment, free-viewpoint replays, virtual attendance at live events, and AR/MR content where a real person or place appears convincingly in the user's own space. One Meridian capture can feed cinema, broadcast, and XR from the same asset.

Radiant Images is an award-winning company with more than two decades of deep experience across cinema, immersive, and live production. We've focused specifically on 6DoF volumetric video for over a decade — and in that time we've delivered hundreds and hundreds of hours of volumetric-based video, with end-to-end experience across the full pipeline: capture, final delivery, and distribution.

We've worked since the earliest days of the field alongside the first professional volumetric companies, including Uncorporeal, 8i, Microsoft, and Sony. Radiant developed the award-winning AXA system, professionally installed at Sony Pictures, and was hired by Sony to build the world's first portable light-field and volumetric capture system — powered by Sony RX0 cameras — which debuted at NAB 2018. That system was the genesis of Meridian. In 2023, Radiant was again commissioned by Sony to develop a portable NeRF stage, which debuted at NAB 2023.

Along the way we've built volumetric systems for movie studios and companies around the world, and created training and simulation programs with Harvard University for NASA and the U.S. Department of Defense. In 2025 we introduced the highly portable Meridian 4D system powered by Apple iPhones.

We continue to combine decades of hands-on experience with new methods to push the capture and processing of volumetric, 6DoF video forward — and to make a real impact on what the format can do.

Yes. Meridian is currently offered as a full service, run by the Radiant Images team — we handle transport, setup (typically a 2–4 person build), calibration, capture, on-set processing, and final delivery. You focus on the creative; we run the system.

The more notice you can give us, the better we can plan crew, configuration, and any pre-shoot tests — though same-day deployment means we can also move quickly when needed. (We're also finalizing features that will let us offer Meridian as a supported purchase for teams who want to operate it themselves.)

No — the crew stay on Radiant's books, not yours. Everyone who builds and runs Meridian is covered under Radiant Images' insurance and workers' compensation, so you don't take on that liability or payroll burden.

Crew work is based on a 10-hour day, portal to portal — meaning the workday is counted door to door, from departure to return, rather than only on-set hours. We'll confirm the specifics for your schedule when we scope the project.

Equipment insurance is the client's responsibility. The company hiring Radiant Images provides insurance coverage for the Meridian system while it's in transit and at offsite locations. This is separate from the crew, who are covered under Radiant's own insurance and workers' compensation.

Please contact Radiant Images' accounts team to assist with the coverage requirements — reach out early so the coverage is in place before the shoot.

You do. Radiant Images works on a work-for-hire basis, so the rights to the footage we capture and the final volumetric assets we deliver belong to the company that hires us. Specific usage and delivery terms are confirmed in your project agreement; if you have particular licensing, confidentiality, or data-handling requirements, let us know early and we'll build them into the engagement.

Meridian is currently available as a service — rental plus capture, delivered by the Radiant Images team. We're finalizing several key features that will let us also offer the system as a supported purchase.

Pricing depends on a number of factors — volume size, configuration, shoot length, processing scope, and delivery timeline — so we put together detailed pricing once we understand your project. Reach out and we'll walk through the options.