Photo to splat to game-ready asset

A capture begins with the dullest possible act: photographs. Hundreds of them, from every angle, in even light, the car standing still while we walk a slow orbit. Nothing about it looks like motorsport yet — it looks like an insurance inspection. But every frame is a measurement, and measurements are the only thing that survive.

From frames to a field

Photogrammetry turns that orbit into geometry: a dense point cloud where every point is a place the cameras agreed with each other. We then train a Gaussian splat — instead of hard polygons, the car becomes a cloud of soft, view-dependent blobs that reproduce paint, chrome, and carbon weave under any light. A splat is uncanny the first time you orbit one. It is the car, not a model of the car.

From a field to an asset

A splat is beautiful but heavy. The last step retopologises it into a game-ready mesh — clean quads, baked materials, a sane polycount, the kind of asset an engine runs at a hundred and twenty frames a second. The same capture now lives in three forms: the measurement, the likeness, and the playable.

What makes it rTracer and not a render farm is what rides along: provenance. Each asset carries a birth record — when it was captured, from which car, by whom — hashed so it cannot be quietly swapped later. The asset is renderable because it happened, and provably happened because the capture is on the record. A car you can drive, with a paper trail.

The grid fills this way, one slow orbit at a time. Every car attested; every number earned.