TAAFT
Free mode
100% free
Freemium
Free Trial
Deals
Companies
SpAItial
Models
Echo 2

Echo 2

By SpAItial
Echo-2 is SpAItial’s latest model for generating spatially persistent 3D worlds rather than sequential video. It takes a text prompt or a single image and predicts a physically grounded 3D representation that captures geometry, appearance, and semantic scene structure in a consistent layout, then converts it into a renderable format for real-time exploration, with the web demo using 3D Gaussian Splatting. SpAItial positions it for interactive world building, digital twins, scene editing, architecture, and robot training, and reports that it outperforms Marble 1.1, HY-World 2.0, and Lyra 2.0 on its cited WorldScore benchmark.
New Multimodal Gen 3
Released: April 28, 2026

Overview

Echo-2 is SpAItial’s frontier 3D world-generation model for creating physically grounded, explorable environments from text or image inputs. Instead of generating a video, it produces a persistent 3D-consistent scene representation that can be rendered in real time on any device, targeting digital twins, robotics, architecture, gaming, and immersive 3D creation.

🎨Scene creation🧩Spatial intelligence🤖Robotics🏛Architectural visualization🎥3D scenes

About SpAItial

SpAItial is an AI startup building Spatial Foundation Models — a class of AI that generates and reasons about the appearance and physics of 3D environments. Its frontier model Echo-2 transforms a single image or text prompt into a spatially-persistent, navigable 3D Gaussian Splat scene rendered in real time, with applications in gaming, digital twins, architecture, robotics simulation, and AR/VR.

Industry: Artificial Intelligence
Company Size: 35
Location: London, England, GB
View Company Profile

Tools using Echo 2

No tools found for this model yet.

Related Models

Last updated: May 3, 2026
0 AIs selected
Clear selection
#
Name
Task