Generative 3D world models
spAItial Echo-2
Echo-2 is spAItial’s frontier model for generating 3D-consistent, explorable scenes from text or image inputs. It focuses on physically grounded scene layout, browser-based real-time exploration, and representations that can support downstream assets.
Overview
| Status | Live in Roamscape |
|---|---|
| Access | Runnable in Roamscape |
| Released | 2026 |
| Inputs | text, image, panorama |
| Outputs | 3DGS scene, mesh, point cloud, semantic masks, assets |
| Best for | digital twins, architectural visualization, robotics environments, interactive 3D scenes, single-image world creation |
Why it matters
Echo-2 gives Roamscape a second major world model family with a different modeling philosophy: physically grounded, 3D-consistent scene generation rather than just a passive video-like output.
Roamscape use
Live / planned integration for generating explorable 3D worlds and comparing outputs with Marble.
Strengths
- physically grounded scene layout
- real-time browser exploration
- single-image to 3D world workflow
- scene decomposition and editing direction
Limitations
- provider terms may allow service/model improvement processing
- availability and API behavior can change
- output fidelity varies by input and scene type