LumiX: Structured and Coherent Text-to-Intrinsic Generation

TL;DR

LumiX introduces a structured diffusion framework that jointly generates coherent intrinsic scene maps from text prompts, ensuring physical consistency and outperforming previous methods in alignment and preference scores.

Contribution

The paper proposes Query-Broadcast Attention and Tensor LoRA for stable joint training and unified generation of multiple intrinsic maps from text prompts.

Findings

Achieves 23% higher alignment than state-of-the-art.

Produces physically meaningful and coherent intrinsic maps.

Can perform image-conditioned intrinsic decomposition.

Abstract

We present LumiX, a structured diffusion framework for coherent text-to-intrinsic generation. Conditioned on text prompts, LumiX jointly generates a comprehensive set of intrinsic maps (e.g., albedo, irradiance, normal, depth, and final color), providing a structured and physically consistent description of an underlying scene. This is enabled by two key contributions: 1) Query-Broadcast Attention, a mechanism that ensures structural consistency by sharing queries across all maps in each self-attention block. 2) Tensor LoRA, a tensor-based adaptation that parameter-efficiently models cross-map relations for efficient joint training. Together, these designs enable stable joint diffusion training and unified generation of multiple intrinsic properties. Experiments show that LumiX produces coherent and physically meaningful results, achieving 23% higher alignment and a better preference score (0.19 vs. -0.41) compared to the state of the art, and it can also perform image-conditioned intrinsic decomposition within the same framework.