Reality, replayed.
VoxaVerse turns real moments into living volumetric scenes you can walk around — captured on devices you already own, reconstructed by our own engine, streamed to any browser or headset. No app. No download.
* Current demo scenes come from the synthetic D-NeRF research dataset — far less detail than a full multi-camera rig carries. Richer real-world scenes are being added; see the benchmarks page for full-fidelity 21-camera reconstructions.
Every camera at once. Then any camera you want.
A ring of time-synchronized phones films the same seconds from every side. The engine fuses those views into millions of moving 4D Gaussians — a scene where the viewpoint is no longer baked in, and the camera becomes a decision you make on playback.
Capture. Create. View.
One continuous pipeline — capture on iPhone, direct from iPad, create on Mac, then step inside on the web or in Apple Vision Pro. The viewer is live today; the rest ship next.
Beacon
iPhone · with Relay for iPad
WiFi-mesh time-synchronized RGB+Depth capture across every iPhone in the room — 100% on-device. Its iPad companion, Relay, gives the director live framing, monitoring, and take control on set.
Preview capture →Nexus
macOS studio
A from-scratch reconstruction and editing suite built on the engine behind every scene on this site — no COLMAP, no cloud queue. The engine already runs; the studio around it ships next.
Preview Nexus →Vortex
Web today · Vision Pro next
Stream full 4D scenes to any browser tab or headset — WebGPU-first, real WebXR entry, honest fallbacks everywhere. It’s what’s rendering above. Native Apple Vision Pro viewing lands next.
Open the viewer →Real numbers. Dated, sourced, re-runnable.
Every figure below was measured on our own engine and hardware, carries a capture date, and anything not yet independently reproduced is flagged — never buried.
Ahead where it counts. Honest where it matters.
Same XR-first entry as the category leaders — plus pipeline provenance, standards-native export, and documented fallbacks instead of a black screen.
| Capability | VoxaVerse | 4dv.ai | Luma AI | Gracia.ai |
|---|---|---|---|---|
| Full 4D scenes on the open web | Yes — live on this site | Yes (player) | No | Yes (player) |
| WebXR entry (feature-detected VR/AR) | Yes | Headset browser | No | Yes |
| Published, dated benchmarks page | Yes — with methodology | No | No | No |
| Standards-native export (open formats) | PLY · SPZ · SOG · OpenUSD | Partial | Partial | Plugin-based |
| On-device capture (nothing uploaded) | Yes — Beacon, coming soon | Studio rigs | Cloud | Rig capture |
| Documented device fallbacks | Yes — per-device matrix | No | N/A | Partial |
Standards-native. Never locked in.
glTF's Gaussian-splatting extension reached Release Candidate in February 2026, and OpenUSD shipped a first-class splat schema in March. VoxaVerse exports to open formats from day one — your captures outlive any single vendor, including us.