MatLoom: Layered Text-to-Material Generation in a Compact Program Space

Name: MatLoom: Layered Text-to-Material Generation in a Compact Program Space
Uploaded: 2026-10-04T23:10:42.406845+00:00
Duration: 198 s
Description: MATLOOM introduces compact, layer-oriented programs to generate material assets from text prompts. It combines explicit authoring structure with text-to-material fidelity, allowing language models to write and repair programs that define PBR channels. This method successfully generates materials that align well with prompts compared to diffusion baselines.

MATLOOM introduces compact, layer-oriented programs to generate material assets from text prompts. It combines explicit authoring structure with text-to-material fidelity, allowing language models to write and repair programs that define PBR channels. This method successfully generates materials that align well with prompts compared to diffusion baselines.

Open the full episode