Pin the world into a shared 3D scene, then walk through what others left behind.
splatGraph turns every place into a single public Gaussian splat. Contributors anchor photos in that shared space, points inside those photos open into generated child scenes and wiki entries, and every journey becomes a replayable trajectory.
One splat, shared coordinates
Every place is a public Gaussian splat. Contributors pin photos to spatial anchors in that single shared coordinate system — no duplicated scenes.
Points become portals
Tap a point inside a pinned photo to open a wiki entry, another photo, or a freshly generated child splat scene — a universe in a raindrop.
Journeys become trajectories
Movements through the graph are recorded in DynamoDB, compiled into durable trajectories in Aurora, and published as shareable manifests in DSQL.
A graph you can edit, a runtime you can record, a wiki you can publish.
Canonical graph
Amazon Aurora PostgreSQL
Places, splats, anchors, photo contributions, image-point links, child scenes, wiki entries, and the edges between them. Durable and relational.
Runtime journeys
Amazon DynamoDB
High-frequency session events, camera segments, and live generation progress. The append-only record of how people actually move.
Published read model
Amazon Aurora DSQL
Versioned, hash-stamped public manifests of places and trajectories — the shareable wiki, decoupled from the editable graph.