Veridian: AI-Generated Semi-Realistic Shaderpack
SHADERAn AI-created Minecraft shader with colored lighting, reflections, and volumetric effects — meant for code inspection, not routine play.
Veridian is a completely AI-generated shaderpack built as an experiment to evaluate the code-generation capabilities of three large language models: Claude Opus 4.8, GPT 5.5, and Deepseek v4 Flash. Every line of the shader code was produced by these AI systems, making this pack a technical curiosity rather than a stable, everyday visual upgrade. It is not recommended for regular gameplay; use it only if you are a developer, researcher, or tinkerer who wants to examine the output of frontier AI models.
Despite its experimental origin, the pack bundles a wide range of semi-realistic rendering techniques. It offers atmospheric sky scattering with Rayleigh/Mie models, soft colored shadows, volumetric clouds, animated reflective water with shoreline foam, screen-space reflections, volumetric sun shafts, and dynamic colored lighting from placed blocks and held torches. Post-processing includes SSAO, bloom, and selectable tonemappers (ACES, Uchimura, Reinhard), all exposed through the Iris in-game settings menu. The pack is compatible with Distant Horizons for long-distance terrain rendering.
- Atmospheric sky with Rayleigh/Mie scattering, warm sunrise/sunset tones, and a starfield.
- Soft shadow mapping with PCF penumbra; shadows can tint through stained glass and water.
- Volumetric raymarched clouds that cast self-shadowing onto themselves and the terrain.
- Animated water featuring waves, shoreline foam, Fresnel-driven reflections, and screen-space reflections (SSR).
- Volumetric sun shafts (god rays) and dynamic colored lighting from placed light sources, with held torch light and Distant Horizons support.
Requirements: Minecraft 1.20.1 or 1.21.1 running on Fabric with Iris Shaders (Sodium strongly recommended). For distant terrain features, install Distant Horizons. Performance is medium to high on modern GPUs; you can reduce the load by lowering cloud steps, SSR steps, volumetric light steps, SSAO samples, and shadow filter samples via the Iris menu. The complete source code is publicly viewable on GitHub for anyone curious about the code patterns these AI models produced.