
Sony Music · 2026
AURORA
A 45-minute generative concert where 80,000 neural avatars respond in real-time to music, mood and crowd energy.
[ Index · 06 of ∞ ]
Six selected projects from the last three years of generative spatial computing. Each one trained, shipped and shaped by AI.

Sony Music · 2026
A 45-minute generative concert where 80,000 neural avatars respond in real-time to music, mood and crowd energy.

Apple Vision Pro · 2025
An AI-powered design system of 120+ spatial components that adapt layout, depth and voice based on user gaze.

Off-White · 2025
A floating retail world where every visitor meets a unique AI stylist trained on Off-White's full design history.

Nike · 2024
Diffusion-model-powered sneaker customisation. 14M sessions. Each shoe generated uniquely per user in under 200ms.

IIT Bombay · 2024
A multi-user spatial chemistry sim powered by LLM tutoring agents, used in 80 Indian universities.

Björk · 2023
An album experienced as a generative sculpture garden where diffusion models compose new visuals every playthrough.
[ FAQ ]
Generative VR worlds, neural avatar products, LLM-driven spatial agents, gen-AI AR try-on, and AI tooling for design and research teams. If it requires a custom model running inside a spatial product, it is in scope.
Yes — AURORA, our Sony Music concert, ran AI-generated visuals for 80,000 attendees in real time. We design for live failure modes from day one: fallbacks, watchdogs and deterministic seeds for repeatability.
Three layers: model evals (quality, latency, safety), product metrics (session length, return rate, conversion) and creative metrics (press, awards, brand lift). Every project ships with a dashboard for all three.
Both. We ship to WebXR for reach, Apple Vision Pro and Quest 3 for fidelity, and custom installations for events. The model stack is portable across all of them.