SpAItial Unveils Echo-3, Turning Phone Videos into Photorealistic 3D Worlds

SpAItial released the world model Echo-3 on October 27, calling it the most accurate reconstruction of real spaces to date and the world's strongest 3D world reconstruction model: users simply record a few minutes of slow video by walking around a space with a phone, and can generate photorealistic, metric-scale Gaussian Splat 3D scenes. Multiple authors tested it that day and shared their takes; the product is available now.

Confirmed

• Input requires only ordinary phone (e.g. iPhone) video—no specialized capture hardware
• Reconstruction covers single rooms, multi-story buildings, outdoor and public spaces
• Output is a Gaussian Splat scene with metric consistency (true dimensions preserved, no distortion)
• rmbrualla, former OpenAI researcher, tested it: about 10 minutes of iPhone footage was enough to reconstruct an entire house into a walkable 3D world
• Beta tester ajcbuilds gave the product positive feedback
• Available on launch day

Why it matters

• Lowers the barrier for 3D reconstruction to casual phone scanning, removing the need for professional equipment, potentially greatly expanding applications in architecture, real estate, AR/VR and more
• Metric scale means reconstructions can be used directly for measurement and design, not just visual display
• If the world-model approach can reliably reconstruct real spaces, it provides a data foundation for robot training, spatial computing, and more

2026-10-27 ~ 2026-10-27 · 7 related posts

• [source] Echo-3 turns 10 minutes of iPhone video into a walkable, metric-scale 3D world — rmbrualla · 2026-10-27
• [source] SpAItial launches Echo-3 world model turning phone captures into metric-scale 3D worlds — risingodegua · 2026-10-27
• SpAItial launches Echo-3 world model turning phone video into photorealistic 3D worlds — jon_barron · 2026-10-27

4 near-duplicate retellings: risingodegua · risingodegua · risingodegua · risingodegua