SpAItial Launches Echo to Instantly Generate Fully Navigable 3D Worlds

• Echo generates interactive 3D environments from text or images, with a single, consistent structure built at real-world scale.
• Scenes can be explored and edited in real time directly in the browser, using fast rendering that runs even on low-end hardware.

SpAItial has launched Echo, a world model that turns an image or text prompt into a full 3D environment. The scenes are built at real-world scale and can be explored, edited, and expanded directly in the browser. When starting from text alone, the model generates a reference image before constructing the full 3D world.

Echo currently uses 3D Gaussian Splatting for fast, GPU-efficient rendering, but its underlying scene representation is format-agnostic and designed to adapt as rendering technologies evolve. Users can move the camera freely and explore scenes from any angle, even on low-end hardware. Behind the scenes, the model builds a single, consistent 3D structure at metric scale, so every view and depth map comes from the same scene, not from separate images. According to the company, this is what makes Echo reliable for simulation, space planning, and visual design, including industrial tasks like construction mapping and factory layout planning.

Echo lets users change the look and layout of a scene without breaking its 3D structure. Users can swap materials, move objects, or try different visual styles. The model also understands and segments different parts of a scene, like walls, chairs, or tables, which makes localized edits easier to apply. According to the company, these features support a range of uses, like digital twins, space design, game development, and robotics.

A browser-based demo of Echo is publicly available. SpAItial is also offering early access through a closed beta.

🌀 Tom’s Take:

A lot of models generate scenes that look 3D but aren’t fully connected. Echo creates one consistent world you can move through, edit, and explore in real time.

Source: SpAItial