Finalverse

FinalStorm · Rendering infrastructure

Spatial data.
Made visible.

A rendering runtime for real places, persistent worlds and the next generation of AI applications.

FinalStorm turns published spatial scenes into frames you can inspect. It streams detail as a camera moves, keeps large-world coordinates precise and connects what you see to stable entity identities.

Repository ↗Architecture ↗

Development runtime, October 2026. The standalone renderer works with authenticated SpatialDB publications. It is a source integration preview; a bounded two-user native MutSea proxy path is now tested, while complete Desktop integration and native geometry remain the next milestone. Repository access may be required.

Actual FinalStorm frame from a synthetic Montréal district rendered on Apple M1 Metal
Actual renderer output on Apple M1 Metal. The district is synthetic reference geometry, not a reconstruction of Montréal. This frame is from the standalone runtime.

Load what matters. Keep the identity.

Scene graphs, external subtrees, geometry and texture mips arrive under explicit budgets. Every asset is checked against its published digest. FinalStorm selects the detail it can actually draw; the database offers advice without controlling every frame.

Precision from the planet to the detail

World and camera transforms remain double precision until conversion to camera-relative GPU values. Reversed-Z depth and stable entity picking connect visual detail to the object it represents.

A real rendering foundation

One scene. Clear authority.

SpatialDB owns published representations and their delivery. MutSea owns the persistent live world. Finalverse Desktop keeps its networking, identity and interactions. FinalStorm presents their data; it does not grant permissions or approve world changes.

Real native data, one bounded step

Two independently authenticated MutSea users now feed scoped native inspections into the renderer. Root-entity proxies refresh alongside the SpatialDB scene and clear on logout. Actual proxy pixels were verified on Metal. Full native linksets, meshes, avatars and the complete Desktop experience remain separate acceptance gates.

The Living World comes next

Our product goal brings Earth discovery, published geometry, multiplayer worlds and shared Lumi collaboration into one journey. A person reviews a proposed change; WorldKernel authorizes it; WorldLine records the receipt and Undo. The current renderer tests establish a foundation for that experience, not completion of the whole product.

The private TetraMesh support foundation preserves tetrahedral geometry APIs and adds explicit-unit scientific coordinates. A future volumetric runtime still requires its own representation contract and rendering acceptance.

Neural rendering, GPU-driven submission and alternative volumetric representations remain research directions. We will publish measured implementations before presenting them as supported runtime features.

Build the runtime ↗Explore SpatialDB