StyLitGAN: Image-Based Relighting via Latent Control
Anand Bhattad, James Soole, David A. Forsyth
Abstract
Generated Image Relit -1 (w + + d 1 ) Relit -2 (w + + d 2 ) Relit -3 (w + + d 3 ) Relit -4 (w + + d 4 ) Relit -5 (w + + d 5 ) Figure 1 . StyLitGAN identifies directional vectors (di) within StyleGAN's style space (W + ) which, when added to the w + style code, effectively modify the lighting of generated images while preserving their geometry and albedo. This process eliminates the need for per-image search or model fine-tuning. The first column displays images generated from StyleGAN2; subsequent columns illustrate the same scene, each relit using a specific direction. These relighting directions (di) are derived through a forward selection method, ensuring diversity and avoiding cherry-picking. The directional effects are consistent across different scenes: for instance, d1 activates an orange-tinged bedside lamp, d2 a less intense white-tinged lamp, d3 introduces strong directional light from the window, and so on, demonstrating diverse relighting capabilities of StyLitGAN.
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext 7a561b69-5dea-44f1-aec2-c3f1769adcf6Cited by top-tier papers14
- UniRelight: Learning Joint Decomposition and Synthesis for Video RelightingKai He, Ruofan Liang, Jacob Munkberg, Jon Hasselgren et al.NeurIPS 2025 · 42 citations
- IllumiNeRF: 3D Relighting Without Inverse RenderingXiaoming Zhao, Pratul P. Srinivasan, Dor Verbin, Keunhong Park et al.NeurIPS 2024 · 34 citations
- Latent Intrinsics Emerge from Training to RelightXiao Zhang, William Gao, Seemandhar Jain, Michael Maire et al.NeurIPS 2024 · 21 citations
- Visual Jenga: Discovering Object Dependencies via Counterfactual InpaintingAnand Bhattad, Konpat Preechakul, Alexei A. EfrosNeurIPS 2025 · 13 citations
- DreamLight: Towards Harmonious and Consistent Image RelightingYong Liu, Wenpeng Xiao, Qianqian Wang, Junlin Chen et al.NeurIPS 2025 · 9 citations
Builds on20
- Alias-Free Generative Adversarial NetworksTero Karras, Miika Aittala, Samuli Laine, Erik Härkönen et al.NeurIPS 2021 · 2,126 citations
- Unsupervised Discovery of Interpretable Directions in the GAN Latent SpaceAndrey Voynov, Artem BabenkoICML 2020 · 459 citations
- Omnidata: A Scalable Pipeline for Making Multi-Task Mid-Level Vision Datasets from 3D ScansAinaz Eftekhar, Alexander Sax, Jitendra Malik, Amir ZamirICCV 2021 · 422 citations
- On Aliased Resizing and Surprising Subtleties in GAN EvaluationGaurav Parmar, Richard Zhang, Jun-Yan ZhuCVPR 2022 · 250 citations
- EditGAN: High-Precision Semantic Image EditingHuan Ling, Karsten Kreis, Daiqing Li, Seung Wook Kim et al.NeurIPS 2021 · 248 citations
Related papers
- StyleGAN knows Normal, Depth, Albedo, and MoreAnand Bhattad, Daniel McKee, Derek Hoiem, David A. ForsythNeurIPS 2023 · 61 citations
- Navigating the GAN Parameter Space for Semantic Image EditingAnton Cherepkov, Andrey Voynov, Artem BabenkoCVPR 2021
- ScribbleLight: Single Image Indoor Relighting with ScribblesJun Myeong Choi, Annie Wang, Pieter Peers, Anand Bhattad et al.CVPR 2025
- GANSpace: Discovering Interpretable GAN ControlsErik Härkönen, Aaron Hertzmann, Jaakko Lehtinen, Sylvain ParisNeurIPS 2020 · 1,049 citations
- LumiNet: Latent Intrinsics Meets Diffusion Models for Indoor Scene RelightingXiaoyan Xing, Konrad Groh, Sezer Karaoglu, Theo Gevers et al.CVPR 2025
