HeadLighter: Disentangling Illumination in Generative 3D Gaussian Heads via Lightstage Captures

TL;DR

HeadLighter introduces a supervised framework for disentangling illumination from appearance in 3D head generative models, enabling controllable relighting while maintaining photorealism and real-time rendering.

Contribution

It proposes a dual-branch architecture and a progressive training strategy for physically grounded illumination disentanglement in 3D head models, using controlled light stage captures.

Findings

Preserves high-quality, real-time head generation.

Enables explicit lighting and viewpoint editing.

Achieves physically plausible illumination decomposition.

Abstract

Recent 3D-aware head generative models based on 3D Gaussian Splatting achieve real-time, photorealistic and view-consistent head synthesis. However, a fundamental limitation persists: the deep entanglement of illumination and intrinsic appearance prevents controllable relighting. Existing disentanglement methods rely on strong assumptions to enable weakly supervised learning, which restricts their capacity for complex illumination. To address this challenge, we introduce HeadLighter, a novel supervised framework that learns a physically plausible decomposition of appearance and illumination in head generative models. Specifically, we design a dual-branch architecture that separately models lighting-invariant head attributes and physically grounded rendering components. A progressive disentanglement training is employed to gradually inject head appearance priors into the generative architecture, supervised by multi-view images captured under controlled light conditions with a light stage setup. We further introduce a distillation strategy to generate high-quality normals for realistic rendering. Experiments demonstrate that our method preserves high-quality generation and real-time rendering, while simultaneously supporting explicit lighting and viewpoint editing. We will publicly release our code and dataset.