Lisbon AI event spotlights a system that turns ordinary video into walkable 3D scenes

dscape · x · 2026-07-29

A Lisbon AI event post highlights work on generative 3D and spatial intelligence.

The quoted text argues that the physical world is fundamentally 3D and that teaching machines to capture and generate it remains hard. It points to a startup/lab building a system that can take ordinary video and turn it into a photorealistic scene you can move through from any angle, with outputs light enough to run in a browser or headset. The examples mentioned include volumetric video, radiance fields, and hyperreal human avatars.

Original post →

More from Multimodal

Multimodal channel →