Why VoxaVerse — the economics of volumetric video

The thesis

Volumetric video is inevitable.
The economics finally arrived.

For a decade, capturing a real moment in 3D meant a soundstage, a camera sphere, and five figures per finished minute. Three curves just crossed — capture hardware, reconstruction silicon, and playback devices — and VoxaVerse is built exactly at the intersection. Every number on this page is sourced and dated.

The cost collapse

From a soundstage to your pocket.

What one minute of walkable, free-viewpoint video costs to produce — the numbers the industry publishes, next to the pipeline this site runs on.

Approach	Capture hardware	Reported cost	Where you can shoot
Legacy volumetric stage	30–100+ mounted cameras, dedicated soundstage	"~$10k per second" end-to-end (Metastage, 2020)¹ · $5–10k per finished minute (Intel Studios)²	One room, one city
Splat-capture studios today	~60 shutter-synchronized cameras in a sphere³	Session + per-minute processing fees³	Fixed studio
Depth-sensor kits	$16k–27k sensor packages + $3k/month software⁴	Hardware + subscription	Portable, sensor-bound
VoxaVerse	The iPhones already in the room + one Mac	Hardware you likely own	Anywhere life happens

Studio rate cards are quote-only and vary by project; the cited figures are the operators’ own published statements, linked in the sources below. The point isn’t a specific rival’s price — it’s that professional volumetric capture has been priced like a satellite launch, and the underlying hardware no longer justifies that.

Silicon economics

CUDA is dominant. It just isn’t efficient here.

Gaussian-splat reconstruction is memory-bound: millions of splats, gradients, and multi-view frames living in one working set. That's the exact workload where discrete-GPU architecture — copying across PCIe into separate VRAM pools — burns money and watts, and unified memory quietly wins.

$13,250One RTX PRO 6000 Blackwell 96 GB, July 2026 street price — a 4-GPU rig is ~$45–53k in GPUs alone, before the workstation around it⁵

2,400 WGPU power draw alone of that 4-card rig (600 W × 4)⁶ — versus 480 W maximum for an entire Mac Studio⁸

512 GBUnified memory in one M3 Ultra Mac Studio at $9,499 — one address space, no PCIe copies, ~$18.5/GB vs ~$138/GB for discrete VRAM⁵⁷

~1/8thThe rough total-cost ratio: one Mac Studio versus a quad-Blackwell workstation of comparable memory capacity⁵⁷

This is not an anti-NVIDIA argument — for batch training farms, CUDA remains the default. It’s an architecture argument. A 4D Gaussian scene wants to be resident in one memory space — captured frames, optimizer state, and the live preview all touching the same splats. On discrete GPUs that means sharding, copying, and 384 GB of VRAM priced at roughly 7× more per gigabytethan Apple’s unified memory⁵⁷. The VoxaVerse engine is Metal-native on Apple silicon: the same chip family that sits in the room where the capture happened. That single fact collapses the studio into a desk.

Honesty notes: GPU street prices are currently inflated ~55% by the GDDR7 memory shortage, so the dollar gap may narrow⁵ — the watts gap won’t. And demand for large unified-memory configurations is so strong that Apple withdrew the 512 GB option from sale in March 2026; the memory-per-dollar figures cite its launch pricing⁷.

One engine, two markets

Pro-grade B2B. Prosumer B2C.

The same economics that let a studio replace a capture stage let a creator own the whole pipeline. Professional tools that fit on consumer hardware serve both markets with one product — that's the category playbook that built the modern editing, music, and photography industries.

B2B

Studios, sports, heritage, advertising

Teams already paying volumetric-stage prices¹²³ get an owned, repeatable pipeline: capture with a managed iPhone rig, reconstruct on Apple silicon they can rack, deliver through open formats their clients’ engines already read. No per-minute processing invoice, no vendor lock on the archive.

B2C

Creators, families, memory-keepers

A creator’s rig is their friends’ phones. A family’s capture stage is the living room. The pipeline that produced every scene on this site runs on hardware from a store, not a soundstage — which is what makes volumetric video a consumer behavior instead of a service you book.

The demand side

Every screen is learning depth.

The playback side of the thesis, with the honest 2026 picture: premium headsets are finding their floor while glasses and open platforms scale — and every one of those surfaces needs content that current cameras can't produce.

7M+Meta × EssilorLuxottica smart glasses sold in 2025 — roughly 3× the prior year, with capacity plans of 10M+/year⁹

+87%IDC’s forecast headset-market rebound for 2026, after the 2025 trough¹⁰

$10.3BVolumetric-video market by 2030 (26% CAGR from ~$2.6B in 2024), corroborated across three research houses¹¹

$480BCreator economy by 2027 (Goldman Sachs)¹² — the population that will make spatial content when the tools let them¹³

The platform shelves are already being built: YouTube shipped a native Vision Pro app with a dedicated Spatial tab in February 2026¹³, Android XR arrived at $1,799 with Samsung’s Galaxy XR¹⁴, and Apple opened its Immersive Video pipeline to third-party creators. Think of the last format transition: the creators who could deliver HD before their audience finished buying HD televisions won the decade. A YouTuber publishing flat video today will publish walkable momentsthe day their platform’s spatial shelf has an audience — ifa camera rig they can afford exists. That rig is the phones already in their pocket, and the render target is any browser tab: the demo on this site’s front page is that claim, running.

Honesty notes: Apple Vision Pro’s own unit volumes remain small and its roadmap cautious — the validated 2026 momentum is in smart glasses, open XR platforms, and the browser itself. That is precisely why Vortex targets the open web first: the largest spatial-capable install base on Earth is the one that needs no headset at all. Evidence that 2D creators are migrating en masse is early-stage and anecdotal today; the infrastructure evidence¹³ is not.

Format tailwind

4DGS is winning the representation war.

Research and standards both moved decisively toward Gaussian splatting: 8K/60 renderers, hierarchical codecs that decode in real time on phones¹⁵, a glTF splatting extension in ratification, and MPEG exploring splat-attribute compression. VoxaVerse's engine, formats, and viewer are native to exactly this representation — with open-format export so the bet is hedged for customers even if the standards shift.

The demo is the pitch.

Most vision pages ask you to imagine the product. Ours is rendering one section above — a real 4D capture, streaming into your browser from a pipeline that runs on a desk.

Step inside a scene Talk to us

Sources

Compiled and verified July 2026. Superscripts above link here; each source below links to the original. Figures marked as estimates are labeled as such in-line.

Volumetric video is inevitable.The economics finally arrived.