LightIt: Illumination Modeling and Control for Diffusion Models

TL;DR

LightIt introduces a novel approach for explicit lighting control in image generation by conditioning on shading and normal maps, enabling high-quality, controllable, and consistent lighting effects in generated images.

Contribution

The paper presents a new method that allows explicit illumination control in diffusion models through shading and normal map conditioning, including a dataset and a relighting model.

Findings

High-quality lighting control in generated images

Effective shading and normal map conditioning

Competitive relighting performance

Abstract

We introduce LightIt, a method for explicit illumination control for image generation. Recent generative methods lack lighting control, which is crucial to numerous artistic aspects of image generation such as setting the overall mood or cinematic appearance. To overcome these limitations, we propose to condition the generation on shading and normal maps. We model the lighting with single bounce shading, which includes cast shadows. We first train a shading estimation module to generate a dataset of real-world images and shading pairs. Then, we train a control network using the estimated shading and normals as input. Our method demonstrates high-quality image generation and lighting control in numerous scenes. Additionally, we use our generated dataset to train an identity-preserving relighting model, conditioned on an image and a target shading. Our method is the first that enables the generation of images with controllable, consistent lighting and performs on par with specialized relighting state-of-the-art methods.