Niantic Spatial launches Scaniverse

# Niantic Spatial launches Scaniverse A company best known for sending millions of people outside to catch Pokémon is now trying to teach machines how to understand sidewalks, warehouses, stadiums, and job sites. This week, Niantic Spatial launched a rebuilt Scaniverse platform and a new Visual Positioning System 2.0, or VPS 2.0, as part of a push to create machine-readable 3D maps of the physical world. (geekwire.com) (nianticspatial.com) The pitch is simple: latitude and longitude tell you where something is on a map, but they do not tell a robot what is actually around it. A loading dock, a stairwell, a concrete pillar, or a blocked hallway all matter in the real world, and Niantic Spatial says its system is built to capture that scene-level detail in a form software can use. (nianticspatial.com 1) (nianticspatial.com 2) That is the problem Scaniverse is meant to solve. Niantic Spatial describes it as an integrated web and mobile platform that lets companies capture real spaces, upload that data, and generate assets such as textured meshes, Gaussian splats, and visual positioning maps. (nianticspatial.com 1) (nianticspatial.com 2) A textured mesh is the familiar kind of 3D model used in games and design tools: a digital shell with surfaces and images wrapped onto it. A Gaussian splat is a newer way to represent a scene using many tiny fuzzy points, which can make a scanned space look more lifelike and faster to render than a traditional polygon model in some cases. (nianticspatial.com) (nianticspatial.com) Niantic Spatial says Scaniverse can work across multiple devices and is meant to be self-serve rather than a custom consulting product. In practice, that means a field team can scan a site with standard hardware, send the data into Niantic Spatial’s system, and use the output for reconstruction, inspection, planning, or localization. (nianticspatial.com) (roboticsandautomationnews.com) The second half of the launch is VPS 2.0. Visual positioning works by comparing what a camera sees against a visual map of the world, so instead of asking satellites where you are, the system asks the scene itself. (nianticspatial.com) Niantic Spatial says VPS 2.0 now works at global scale and does not require prior scanning to provide improved positioning. In spaces that have been mapped with Scaniverse, the company says the system can deliver near centimeter-accurate six-degree-of-freedom localization, which means both position and orientation: not just where something is, but exactly how it is turned. (nianticspatial.com) (nianticspatial.com) That distinction matters in places where Global Positioning System signals are weak, noisy, or missing. Warehouses, dense urban areas, indoor venues, and other GPS-denied environments are hard on ordinary navigation systems, and Niantic Spatial is explicitly marketing VPS 2.0 as a layer that can correct GPS errors and keep machines oriented when satellite signals degrade. (nianticspatial.com) (nianticspatial.com) The company is aiming this stack at what people in robotics and industrial software often call embodied or real-world artificial intelligence: systems that do not just answer questions on a screen, but move through physical environments. Niantic Spatial says robots, drones, and spatial apps need geometry and grounding together, because a good-looking 3D model is not enough if the machine cannot place itself accurately inside it. (nianticspatial.com) (nianticspatial.com) This launch is also a marker of how far Niantic has moved from its consumer-gaming identity. In March 2025, Niantic announced that Scopely would acquire its games business, while its geospatial artificial intelligence platform would be spun out into a standalone company called Niantic Spatial, led by John Hanke and backed with $250 million in capital. (nianticlabs.com) (scopely.com) (nianticspatial.com) So the bigger change here is not just a product update to an old scanning app. Scaniverse began as a mobile 3D scanning app that Niantic acquired earlier, but Niantic Spatial has now repositioned it as the front door to a broader spatial infrastructure stack for enterprises and developers. (nianticlabs.com) (nianticspatial.com) (nianticspatial.com) Niantic Spatial ties these products to what it calls a Large Geospatial Model, its effort to build a living digital model of the world that machines can query and navigate. The company’s public language suggests it sees Scaniverse as the data-ingestion layer and VPS 2.0 as the localization layer, with both feeding a larger platform for mapping, simulation, and machine perception. (nianticspatial.com) (nianticspatial.com) (nianticspatial.com) The near-term customers sound less like gamers and more like operations teams. Niantic Spatial has pointed to uses in warehouses, entertainment sites, construction, logistics, robotics, and other settings where software needs an exact understanding of real space instead of a rough blue dot on a map. (nianticspatial.com) (tech.yahoo.com) (roboticsandautomationnews.com) There is still a gap between a polished demo and a widely deployed platform. Niantic Spatial’s accuracy claims, scale claims, and ease-of-use claims mostly come from the company itself right now, so the real test will be whether customers can use ordinary phones and cameras to create maps that stay reliable in messy environments with changing light, moving objects, and constant physical change. (nianticspatial.com) (nianticspatial.com) (geekwire.com) Still, the direction is clear. Niantic Spatial is betting that the next important map will not just tell software where Earth is, but what a place looks like, how it is arranged, and exactly where a machine stands inside it. (nianticspatial.com) (nianticspatial.com)