Hiverlab has quietly added a Gaussian Splatting Runtime Editor to SpatialWork, its cloud based digital twin platform. The update lets users bring photo-derived splat models straight into an existing twin, line them up, and begin editing in real time. No pre-bake in a DCC tool, and no lidar pass beforehand. Hiverlab’s documentation calls the underlying pipeline a “multi-splat” method and stresses that the engine can merge several captures into one continuous environment, a workflow aimed at large plants and facilities rather than single object scans.
Instead of converting images into dense triangle meshes, the editor reconstructs a radiance field as millions of view dependent 3D Gaussians. Because each Gaussian stores its own color and opacity, the result can be rendered without a heavyweight polygon stage. SpatialWork keeps these splats editable: individual chunks can be rotated, scaled, or deleted, and multiple captures stitched until seams disappear. The idea is to lower the entry cost of building large scale twins by replacing traditional survey scans with off the shelf photography.
A distinguishing feature is the ability to bind live IoT streams to a splat while it is imported. Temperature, vibration, or throughput readings can appear directly on the geometry, so the twin stays current as the physical asset changes state. For organizations already piping sensor data into a data lake, the editor effectively becomes a spatial front-end instead of a stand-alone model viewer.
SpatialWork was already in production at HP, DB Schenker, and NEOM for more conventional digital twin tasks. Those same companies, according to regional press coverage, are piloting the splat editor in their existing installations. While detailed case studies are still thin, Hiverlab claims the shift from laser scans to photo based splats can shave weeks off data capture schedules and cut survey costs.
Gaussian Splatting has moved quickly from research demo to commercial feature, but most enterprise deployments still rely on custom pipelines. Hiverlab’s update is notable because it places splats inside a broader operational stack that already handles IoT ingest, access control, and web delivery. If the workflow scales, both in model size and in change management discipline, it could signal a broader move away from polygon heavy twins toward radiance field representations optimized for quick capture and iterative updates.
To learn more, check out the Hiverlab website.
New to Gaussian Splatting? Start here
