Niantic Spatial‘s Reconstruction API now accepts drone imagery from Spexi Geospatial’s capture network and returns geo-referenced 3D gaussian splats, the two companies announced today. The output is viewable and measurable directly inside the Spexi World platform, with coordinate grounding that lets individual scan tiles slot into city-scale composite maps.
Spexi operates an aerial imagery network with more than 10,000 drone pilots who have mapped over six million acres across more than 200 cities in the United States and Canada at 2.8-centimeter ground sample distance. Each flight follows an autonomous, standardized protocol optimized for machine learning pipelines. A single 25-acre Spexigon takes roughly seven minutes to complete. Niantic Spatial has calibrated its reconstruction pipeline specifically to Spexi’s capture workflows, meaning every flight is processed under conditions tuned to match the imagery’s geometry and acquisition parameters.
Customers of either company can commission drone captures through Spexi and receive high fidelity 3D Gaussian splat reconstructions through Niantic Spatial’s API, with the resulting models embedded in the Spexi World viewer alongside a measurement tool. Separately, Spexi has been designated a preferred imagery provider for training Niantic Spatial’s real-world foundation models, the Large Geospatial Model the company has been building since its spin-out from Niantic.
Inhi Cho Suh, who joined as CEO earlier this year, framed the scale jump as the central problem the partnership addresses. Getting high quality 3D reconstruction has largely operated at the object or building level, and this is an attempt to extend that to city scale. The metric-scale reconstructions stitched from multiple drone scans are grounded in geometry, making them applicable across simulation, visual positioning, and AI training, the three core product lines Niantic Spatial has been building out. Recently the company had released SPZ V4.0 format for compressed Gaussian splats and its broader pivot to physical AI through Scaniverse.
Announced use cases span infrastructure inspection, insurance risk assessment, energy site analysis, asset management, and 3D measurement. Adding a crowdsourced drone network as the upstream capture layer is a step toward making on demand city-scale reconstruction a repeatable workflow.
