VOXELMIND XR · Shape a space together
A seated spatial workbench: describe a room, review an AI proposal, refine it by hand, and return to your saved design.
Innovation & Creativity
Reviewed AI connects selected objects and explicitly confirmed room surfaces. The miniature fits a confirmed real table; virtual mode provides a stand instead.
Experience Design
Open a curated lounge without login, or describe your own space. Tabletop is default; walking is optional. Local save supports returning without an account.
Technical Implementation
Optional WebXR features, safe spatial fallbacks, interrupted-gesture cancellation, local/Cloud recovery, compressed furniture LOD and measured counters.
Polish & Presentation
Traced cube branding, squared typography, restrained lime accents, textured furniture and optional spatial cues. Submission uses only actual headset footage.
Runs on
Meta Quest browser via WebXR, plus desktop and mobile browsers with the same saved worlds. Optional PWA install adds a home-screen entry and offline shell.
Known limits
Eye tracking is not inferred from head gaze. Surfaces are partial observations, not a full scan or obstacle safety system. Persistent anchors remain diagnostic-only until physical reload alignment is confirmed. AI needs connection and approval. Sharing is an independent read-only snapshot, not live collaboration.
Real footage · 90 seconds
- 0–10s: seated designer and room — “A room should shape the design, not disappear behind it.”
- 10–25s: enter, show detected table and fit miniature; show virtual fallback separately.
- 25–40s: right-hand select, grab, two-hand resize and large central controls.
- 40–60s: point at sofa, speak a change, review, explicitly Apply; show Undo.
- 60–75s: save, end, return, re-detect and re-align. Show persistent-anchor result only if verified.
- 75–90s: accessibility, actual performance HUD, final room and URL/QR.
No generated footage, fake scans, hidden failures or invented FPS.