The /lab/ page is an experiment, not ten scientifically validated cloud
presets.
Takram
- Repository: three-geospatial.
-
@takram/three-clouds0.7.6,@takram/three-atmosphere0.19.1,@takram/three-geospatial0.9.1. - MIT license: download LICENSE.txt.
- Cloud assets: revision
45a1c6c1bb9fd38b3680fd120795ff4c32df68ff. -
Atmosphere lookup tables: revision
eac103980f20c0956f2d3215833e73514be08462. - STBN data: revision
9627216cc50057994c98a2118f3c4a23765d43b9. - STBN texture data originates from NVIDIA's Spatiotemporal Blue Noise SDK. Its upstream license is included separately; do not assume this asset is covered by Takram's MIT license.
- Volume rendering follows the official vanilla Three.js example. The application integrates wind and evolution offsets separately, with explicit pause, time scaling and stepping.
Photon
- Copyright © 2021–2025 Benjamin Stott (SixthSurge).
- Repository: Photon.
- Source revision:
15458c0937f8647c37eb6a501bef5eb3bf3da31b. -
Source files:
shaders/include/sky/clouds/cirrus.glsl,common.glsl,shaders/include/utility/random.glsl,shaders/image/noise.png. - Full custom license: download LICENSE.txt. Photon code/assets are not MIT licensed. Redistribution is subject to the upstream license, including its monetization restrictions.
-
src/lab/photon.frag.glslretains the earlier cirrus and cirrocumulus density functions and curl noise, retaining the original noise texture. The lab now uses it only for cirrocumulus; cirrus uses the separate model below. Independent evolution time, browser coordinates, a thin layer and simplified lighting are adaptations. Solar elevation and azimuth drive all lab modes; thin-cloud sunlight colour is sampled from Takram's atmospheric transmittance LUT. This does not reproduce Photon's complete lighting pipeline or imply identical visuals. -
src/shaders/cirrocumulus.glslalso adapts Photon density, curl noise and multi-scale erosion for the home atlas, using the same original noise texture. Separate coverage, local ripple organisation, ground-view projection, a smooth height profile, four-sample optical-depth integration and mipmap filtering are project adaptations. Home lighting remains simplified; it is not the complete Photon renderer. - Gradient-noise derivative credited upstream to Inigo Quilez.
- Hash credited upstream to Dave Hoskins.
Cirrus morphology research
- Visual targets: WMO International Cloud Atlas, Cirrus fibratus and uncinus, and Met Office, Cloud types for observers, printed page 33. No photographs are bundled with the application.
- Hogan, R. J. and Kew, S. F. (2005), A 3D stochastic cloud model for investigating the radiative properties of inhomogeneous cirrus clouds, Q. J. R. Meteorol. Soc., 131, 2585–2608.
-
src/lab/cirrus.frag.glslis an original morphology-driven approximation inspired by fallstreak displacement, wind shear and the spread of ice fall speeds. Sparse generating regions, height-dependent analytical trajectories and anisotropic texture detail define the thin volume. It does not implement the paper's Fourier model, ice microphysics or a numerical weather solver. - No source code or datasets from the GPL-2.0 Cloudgen implementation were incorporated. The model reuses the existing Takram shape texture and atmospheric lookup tables; it adds no external runtime assets.
-
src/lab/CirrusEffect.tsintegrates the thin volume at reduced resolution and filters four sampling phases with a short history. Edits and seeks reset the history. Solar elevation and azimuth use the lab's shared controls.
All reference assets are served locally; the home atlas and experiment make no runtime requests to GitHub or other external asset hosts.