Under the Hood

The realm runtime

Every realm runs on a GPU-accelerated WebGL rendering pipeline with physically based lighting. Materials behave like the real thing: glass refracts and disperses light into chromatic fringes, metal picks up the environment, fabric scatters softly. A cinematic post stack with filmic tone mapping, bloom, and fine grain sits on top, so realms grade like footage, not like screenshots.

Geometry and assets

Scenes are heavy by nature, so the pipeline is built around streaming. Geometry ships Draco-compressed and decodes on demand, textures stream progressively in GPU-native formats, and video transcodes into adaptive renditions. Flat artwork is not pasted onto planes: logos and vector shapes are traced and extruded into real volumes with rounded, liquid edges, which is exactly how the glass mark on this site is made.

The motion system

Motion in a realm is directed, not decorated. The camera follows a language of moves (push-ins, drifts, reveals) driven by scroll and cursor, with spring-based easing that never snaps. Animation runs outside the framework render cycle on the GPU clock, so interface state changes never stutter the scene.

Adaptive device tiers

At load, the runtime measures the device it woke up on and picks a quality tier: full volumetric effects on strong hardware, a lighter composition of the same story on older phones. Resolution, effect passes, and asset renditions scale together, reduced-motion preferences are honored, and everything pauses the moment it leaves the viewport.

The HTML underlay

Beneath the scene lives a real document: headings, paragraphs, links, and metadata in clean, crawlable HTML. Search engines and screen readers read the underlay, visitors experience the scene, and both always tell the same story.

Next steps