PacTure: Efficient PBR Texture Generation on Packed Views with Visual Autoregressive Models

TL;DR

PacTure is a new framework that efficiently generates high-resolution PBR textures for 3D meshes from text descriptions, overcoming limitations of existing methods in speed and global consistency.

Contribution

It introduces view packing for higher resolution multi-view generation and integrates fine-grained control within an autoregressive framework for improved efficiency.

Findings

PacTure achieves higher quality textures than state-of-the-art methods.

It significantly reduces inference time for multi-view texture generation.

The method maintains global consistency without sacrificing resolution.

Abstract

We present PacTure, a novel framework for generating physically-based rendering (PBR) material textures for an untextured 3D mesh from a text description. Existing 2D generation-based texturing approaches either generate textures sequentially from different views, resulting in long inference times and globally inconsistent textures, or adopt multi-view generation with cross-view attention to enhance global consistency, which, however, limits the resolution for each view. In response to these weaknesses, we first introduce view packing, a novel technique that significantly increases the effective resolution for each view during multi-view generation, without imposing additional inference cost. Unlike UV mapping, it preserves the spatial proximity essential for image generation and maintains full compatibility with current 2D generative models. To further reduce the inferencing cost, we enable fine-grained control and multi-domain generation within the next-scale prediction autoregressive framework, creating an efficient multi-view PBR generation backbone. Extensive experiments show that PacTure outperforms state-of-the-art methods in both quality and efficiency.