Writing / Conference paper
Demokritos: Interactive, Community-Centred, Self-Improving Shader Generation using Large Language Models
Abstract
Demokritos is an in-browser research prototype for generating, rendering, and evaluating WebGL shaders from natural-language descriptions. It treats large-language-model generation as a community-centred iterative process: users inspect and render generated GLSL programs, evaluate their results, and contribute successful prompt–shader pairs to a growing shared corpus. The project investigates how collective evaluation can improve specialised program generation rather than treating generated code as automatically correct.
Citation
Ferrando, Santiago, Christian Clark and Tomas Laurenzo. Demokritos: Interactive, Community-Centred, Self-Improving Shader Generation using Large Language Models. To appear in Proceedings of the 25th International Symposium on Electronic Arts, ISEA 2025, May 2025, Seoul, South Korea.