The Spatial Signal

DIGITAL TWINS · GEOSPATIAL · PHYSICAL AI · Signal 0184 · SEPTEMBER 22, 2026

Niantic Spatial opens beta for relighting captured 3D scenes

10 Sec Scan

Niantic Spatial has launched a design-partner beta that changes lighting in Gaussian splat reconstructions while preserving scene geometry, allowing technical teams to evaluate visual variants of captured environments in simulators.

What Changed

Niantic Spatial has applied established precomputed radiance transfer techniques to its Gaussian splat reconstruction pipeline. Its beta produces relit scene variants offline while retaining the original geometry and associated spatial labels.

The Signal

Teams building simulations of physical environments could vary illumination without rebuilding collision geometry, navigation meshes or semantic labels. That supports more controlled comparisons, although mesh artifacts and unvalidated scene types limit where the beta can be relied upon.

60 Sec Understand

Niantic Spatial is adding an offline, batch relighting workflow to its real-to-simulation pipeline. Gaussian splats normally retain the lighting recorded during capture. The new workflow uses an aligned mesh to simulate illumination, then transfers the results back to the splats by changing their appearance coefficients rather than their geometry. Collision meshes, navigation meshes and semantic labels remain unchanged across variants. This gives simulation teams a way to test different lighting conditions against a consistent spatial foundation without recapturing the location. Access is limited to selected design partners, with Niantic Spatial reviewing source data and producing variants; it is not a self-service API. The company reports early examples spanning time-of-day and weather-related appearances, but quality depends on mesh accuracy. Interiors, glass and sharp edges still need validation, and animated rain or volumetric fog require complementary simulation layers.

The Evidence

Evidence supplied by Niantic Spatial, Inc..

Signal Strength

MODERATE

Evidence Strength

MODERATE

Technologies

Gaussian SplattingPrecomputed Radiance Transfer3D ReconstructionSimulation

Companies

Niantic Spatial

Places

Sources

Niantic Spatial, Inc.

Original Source

Related Signals