One module, one source of truth
Every material exports its name, its TSL apply() function, and the readable source fragment displayed in the Lab.
Shader Lab is Aurelius' public rendering workbench. Pick a material, drag the geometry, adjust the clock, and inspect the source that built the result. Every material here is a reusable module verified on both WebGPU and WebGL2 before it appears.
Interactive workbench
DRAG TO ROTATE · SCROLL TO ZOOM
Loading the pinned renderer and verified material modules.
These are not screenshots or separate shader forks. The material selected above is the same ES module used by the benchmark, the landing-page preview, and this dedicated workbench. The source pane is exported by that module, so it cannot silently diverge from the implementation.
Every material exports its name, its TSL apply() function, and the readable source fragment displayed in the Lab.
The bench renders frozen parity frames, live animation, and a mobile advisory pass on WebGPU and WebGL2. Console errors and invalid frames fail the gate.
Each chip carries its measured cost class and baseline provenance. Expensive materials remain visible, but never hidden behind a performance claim.
Three new materials broaden the library without adding redundant variations:
An iridescent shell exposes view-dependent thin-film color alongside a drifting, circuit-like facet field.
Animated radial pulses illuminate Truchet traces, turning a low-cost pattern primitive into a readable surface language.
Curl noise drives color bands and filaments. It is intentionally expensive and visibly badged as the right reference for a real flow-field shader.
Modern browser rendering spans two backends: WebGPU uses WGSL and WebGL2 uses GLSL. Three.js Shading Language lets Aurelius author one node graph and compile it for both. That reduces parallel maintenance, but only if the graph is tested on both paths. The Lab exists to make that testable work public.
Materials receive a frozen clock for screenshot comparison, then run live under each backend. Differences beyond a material-specific tolerance block release.
Cost classes come from a documented Intel UHD 630 baseline. The live readout is your machine's current frame time, not a borrowed benchmark number.
Drag the knot, change the clock, and switch materials. Direct interaction is a useful test of behavior that static imagery cannot provide.