Reproduce “Reflecting Process Expertise in Procedural Material Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kunal Gupta, Gaurav Joshi, Yen-Ru Chen, Seemandhar Jain, Ishit Mehta, Manmohan Chandraker
MaterialApprentice generates editable procedural materials by retrieving expert process traces and compiling a planned trace into a Blender shader-node graph. It extracts steps, settings, and design intent from tutorials, then combines relevant demonstrations for a new request. In a five-artist study, process retrieval required 7.84 structural edits versus 10.84 for graph retrieval, while parameter tweaks were higher: 5.75 versus 4.83.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.