A newly published US patent filing from Adobe describes a system that builds navigable 3D environments from a text description.
How it would work
It runs in two stages. First an AI model creates a spherical, 360-degree image of the scene from your prompt. A second model then turns that image into real 3D geometry that you can view from different angles and move through.
Read it as a direction, not a product
A patent isn't a launch date, and companies file plenty of ideas they never ship. Still, it fits a pattern: Adobe has published dozens of related filings, covering AI-generated camera angles, matching materials across 3D models and virtual try-on for clothing.
Adobe's Substance apps, including Substance 3D Painter and Modeler, don't have Firefly tools yet. If this ever ships, that's the obvious home for it.
Sources
This story is our summary of reporting by Creative Bloq. Read the original for full detail.