pith. sign in

arxiv: 2605.21766 · v1 · pith:PBDVWQR3new · submitted 2026-05-20 · 💻 cs.CV · cs.GR

BodyReLux: Temporally Consistent Full-Body Video Relighting

Li Ma , Mingming He , Xueming Yu , David M. George , Ahmet Levent Ta\c{s}el , Paul Debevec , Julien Philip This is my paper

Pith reviewed 2026-05-22 08:48 UTC · model grok-4.3

classification 💻 cs.CV cs.GR
keywords full-body video relightingtemporally consistent relightingvideo diffusion modellighting conditioningdynamic performance captureOLAT capturesubject-specific model
0
0 comments X

The pith

A subject-specific video diffusion model can relight full-body human performances consistently across frames under new lighting.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper seeks to establish that full-body video relighting can be achieved photorealistically and without temporal flickering by training a diffusion model on specially captured aligned lighting pairs. This would matter for post-production in film and content creation, where changing lights on existing human footage saves time and resources. The method combines static OLAT captures with a dynamic setup of rapidly interleaved lighting sequences to build training data, then adapts a pretrained text-to-video model with token-based light conditioning and masked attention to handle changing lights. Data augmentation further improves robustness across viewpoints and performances.

Core claim

BodyReLux is a subject-specific video diffusion-based framework that relights full-body human performances in a temporally consistent way. It trains on a hybrid dataset of pixel-aligned video relighting pairs obtained from traditional static One-Light-at-a-Time capture plus a novel dynamic capture that rapidly interleaves two smoothly varying lighting sequences above the flicker-fusion threshold. The model starts from a pretrained text-to-video diffusion model, adds a lighting conditioning method that represents each light source as a token, and uses masked attention over lighting sequences to support dynamic control, together with a data augmentation pipeline.

What carries the argument

Token representation of each light source together with masked attention over sequences of lighting, which supplies precise dynamic control inside the video diffusion model.

If this is right

  • Human performance videos can be edited with arbitrary dynamic lighting changes after capture while preserving motion and detail.
  • The hybrid capture approach supplies enough aligned training pairs to adapt generative video models to lighting tasks without large new datasets.
  • Token-based lighting control enables fine-grained, frame-by-frame lighting edits that respect physical light sources.
  • Data augmentation during training increases robustness to viewpoint and performance variation in the relit results.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The interleaving capture technique could supply paired data for other video tasks that require exact alignment under varying conditions, such as material or weather changes.
  • If inference speed improves, the same conditioning approach might support live relighting in virtual production or augmented reality settings.
  • Extending the light tokens to include color temperature or shadow parameters could allow more complete scene editing beyond simple relighting.

Load-bearing premise

Rapidly interleaving two smoothly varying lighting sequences above the human flicker-fusion threshold yields accurate pixel-aligned relighting pairs without visible artifacts or motion disruption.

What would settle it

Relit output videos that exhibit flickering, lighting mismatches, or visible artifacts when tested on new performances under changing lights would show the temporal consistency and alignment claims do not hold.

Figures

Figures reproduced from arXiv: 2605.21766 by Ahmet Levent Ta\c{s}el, David M. George, Julien Philip, Li Ma, Mingming He, Paul Debevec, Xueming Yu.

Figure 1
Figure 1. Figure 1: Input and relighting results of BodyReLux. Given an input video of a subject with arbitrary lighting, and a target lighting condition, BodyReLux allows relighting the input video to the target lighting with a high level of photorealism and temporal consistency. The technique works for any framing (full body, upper body, and closeup), any resolution, and any frame length, including casually captured videos.… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the method. We capture static OLAT data and bi-packed video data of a subject moving inside a large LED sphere, resulting in a dataset of video relighting training tuples that consists of two pixel-aligned videos under different lighting conditions and the corresponding lighting sequences. We train a video diffusion model with a novel lighting conditioning module that supports dynamic lighting … view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of bi-pack lighting sequence. A bi-pack sequence consists of two lighting sequences that vary smoothly every 1 second, while rapidly alternating between the two at 120 Hz. Because the switching fre￾quency exceeds the human flicker-fusion threshold, it appears as a mixture of two lightings evolving at 1 Hz. a dynamic bi-pack capture, with a total capture time of 30 minutes per subject. The det… view at source ↗
Figure 5
Figure 5. Figure 5: Relighting results. We show input videos and relighting results under novel performance and lighting conditions. Each result is a single from from a relit video. Our method achieves photorealistic relighting under directional lighting, multiple point lights, gradient illumination, image-based lighting, and manually edited HDRI maps for various input aspect ratios and resolutions. The resolution is shown on… view at source ↗
Figure 6
Figure 6. Figure 6: In-the-wild relighting results. We show video relighting results for in-the-wild captures. 4.2 Comparisons We compare our method with several baselines. Closest to ours is DifFRelight [He et al. 2024], an image diffusion-based, subject￾specific relighting model for facial performances. We re-implement DifFRelight using our network backbone and retrain it on our data to align the experiment settings; this b… view at source ↗
Figure 7
Figure 7. Figure 7: Video Relighting results. We show video relighting results of BodyReLux under static lighting and dynamic lighting conditions. Note that our method also works for input video with dynamic lighting conditions, while still being able to produce consistent relighting results. Input GT Ours DifFRelight+ SwitchLight3 LuxPostfacto Allfreq [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative comparisons. We compare ground truth and predicted relighting results. Ours achieves the most photorealistic relighting results. DifFRelight+ achieves reasonable results, but tends to produce artifacts in fast-moving regions, as highlighted in red boxes. Other generalized relighting models produce larger errors compared to the ground truth. , Vol. 1, No. 1, Article . Publication date: May 2026 … view at source ↗
Figure 10
Figure 10. Figure 10: OLAToken conditioning. Our OLAToken conditioning method helps improve the lighting accuracy. Target Lighting Frame 0 w/o dyn. cond. Full Frame 20 Frame 40 Frame 60 [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Dynamic lighting conditioning. Without using dynamic lighting conditioning, lighting conditions tend to leak across frames. 2025; Zhang et al. 2025], we implement the HDRI-based condition￾ing by encoding a 2D HDRI into a 1D embedding using a shallow convolutional neural network (CNN), which is then fed to the DiT backbone via cross-attention. To support dynamic lighting, we en￾code HDRI sequences frame by… view at source ↗
Figure 9
Figure 9. Figure 9: Ablation of different types of training data. Without bi-packed video data, the model fails to produce plausible results when trained only on pseudo motion derived from static OLAT captures. Without OLAT data, the model cannot accurately relight scenes under extreme lighting conditions. Ablation on training data. We evaluate the necessity of using bi￾packed video captures and static OLAT captures (w/o vide… view at source ↗
Figure 13
Figure 13. Figure 13: Effectiveness of pretrained weight. Without loading the pre￾trained weight, the relighting result tends to be blurry with fewer specular reflections and lower relighting accuracy. ℛ(LA) ℛ(LB) ℛ(LA + LB ℛ(L ) = ℛ(L) A)+ ℛ(LB) ℛ(L/16) ℛ(L/4) ℛ(L*4) ℛ(L*16) ℛ(L)/16 ℛ(L)/4 ℛ(L)*4 ℛ(L)*16 [PITH_FULL_IMAGE:figures/full_fig_p010_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Linearity test of our relighting model. Our relighting model produces visually linear results with respect to the input lighting conditions. Here R indicates our video relighting model. results indicate that our method exhibits near-linear behavior under both lighting addition and exposure scaling. Repeatability of results. We assess the repeatability of our method by conducting three independent training… view at source ↗
Figure 16
Figure 16. Figure 16: Relighting results on multi-subject shot. We compare relighting results on a two-subject video (first row), with relighting them individually by cropping (second row). The model is able to generalize to multi-subject shot even if it’s only trained on single-subject data. GT Models trained on 4 subjects Subject Erik Model trained on Erik [PITH_FULL_IMAGE:figures/full_fig_p011_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Relighting results from out-of-distribution identities. When inferred on unseen identities, the model tends to bake in subject-specific features such as cloth colors, hair color, and facial features. we train our model for one subject, and infer on a different subject. Interestingly, it still produces reasonable relighting, but it transfers features from the training subject to the new subject, including … view at source ↗
read the original abstract

Being able to relight human performance is a fundamental task for post production and content creation. We present BodyReLux, a subject-specific video diffusion-based framework for relighting full-body human performances in a temporally consistent way. Our model is trained on a hybrid dataset of pixel-aligned video relighting pairs, covering a diverse combination of lighting conditions, performances and viewpoints. To acquire such dataset, we combine traditional static One-Light-at-a-Time (OLAT) capture and a novel dynamic performance capture in which two smoothly varying lighting sequences are rapidly interleaved. Because the lighting operates above the human flicker-fusion threshold, the interleaving does not appear to strobe. We train our video relighting model from a pretrained text-to-video model to fully leverage the generative priors for producing high quality videos. To achieve accurate lighting control, we introduce a new lighting conditioning method that represents each light source as a token. We further condition on sequences of lighting using masked attention to support dynamic lighting control. Together with a carefully designed data augmentation pipeline, we achieve photorealistic, robust, and temporally consistent video relighting of subject-specific human performances.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript presents BodyReLux, a subject-specific video diffusion-based framework for temporally consistent full-body human performance relighting. The model is trained on a hybrid dataset of pixel-aligned video relighting pairs obtained by combining traditional static One-Light-at-a-Time (OLAT) capture with a novel dynamic performance capture that rapidly interleaves two smoothly varying lighting sequences. Because the lighting operates above the human flicker-fusion threshold, the interleaving is claimed to be imperceptible and to yield clean pairs. The approach fine-tunes a pretrained text-to-video diffusion model, introduces a lighting token representation for each light source, and uses masked attention over lighting sequences together with data augmentation to achieve photorealistic, robust, and temporally consistent results.

Significance. If the central claims hold, the work would be a useful contribution to computer vision for post-production and content creation by providing a practical route to subject-specific video relighting data. The hybrid capture strategy and the token-based lighting conditioning that leverages pretrained generative priors are clear strengths. The method addresses a real data bottleneck in dynamic relighting. However, the significance is limited by the absence of detailed quantitative validation of the core data-acquisition assumption in the provided material.

major comments (1)
  1. The central claim that photorealistic and temporally consistent relighting is achieved rests on the quality of the pixel-aligned training pairs produced by the novel dynamic capture. The manuscript states that interleaving two smoothly varying lighting sequences at high speed works because lighting operates above the flicker-fusion threshold and therefore does not appear to strobe. This perceptual argument does not automatically guarantee the technical properties required for the claim, such as exact per-frame lighting isolation, absence of motion-induced lighting bleed, or sub-frame temporal alignment between the interleaved sequences and the performance. Any residual misalignment or bleed would directly degrade the diffusion model's ability to learn accurate relighting, especially under fast motion or complex lighting changes. Quantitative evidence (e.g., lighting-separation error, per
minor comments (1)
  1. The abstract asserts photorealistic results without referencing any quantitative metrics, ablation studies, or baseline comparisons; adding a concise statement of key numerical findings would improve the summary.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive review and for recognizing the strengths of the hybrid capture strategy and token-based lighting conditioning. We address the major comment on the validation of the dynamic capture data below.

read point-by-point responses

  1. Referee: The central claim that photorealistic and temporally consistent relighting is achieved rests on the quality of the pixel-aligned training pairs produced by the novel dynamic capture. The manuscript states that interleaving two smoothly varying lighting sequences at high speed works because lighting operates above the flicker-fusion threshold and therefore does not appear to strobe. This perceptual argument does not automatically guarantee the technical properties required for the claim, such as exact per-frame lighting isolation, absence of motion-induced lighting bleed, or sub-frame temporal alignment between the interleaved sequences and the performance. Any residual misalignment or bleed would directly degrade the diffusion model's ability to learn accurate relighting, especially under fast motion or complex lighting changes. Quantitative evidence (e.g., lighting-separation error, per

    Authors: We agree that the perceptual argument based on the flicker-fusion threshold alone does not automatically guarantee the precise technical properties needed for high-quality training pairs. Our capture setup synchronizes the lighting controller and camera such that each video frame is exposed under a single, consistent lighting condition from one of the interleaved sequences, with the rapid smooth variation ensuring no visible strobing. However, we acknowledge that the original manuscript lacks detailed quantitative validation of lighting isolation error, motion-induced bleed, or sub-frame alignment. We will add a new section with controlled experiments on static scenes (comparing interleaved dynamic captures against traditional OLAT ground truth) reporting per-frame lighting separation error (measured as mean intensity deviation across light sources) and temporal alignment metrics. These additions will directly address the concern and strengthen the central claim. revision: yes

Circularity Check

0 steps flagged

No circularity: method builds on external pretrained model and novel capture without self-referential reductions

full rationale

The paper describes an engineering pipeline for subject-specific video relighting: a hybrid dataset is acquired via static OLAT plus a new interleaved dynamic capture, then used to fine-tune a pretrained text-to-video diffusion model with added lighting-token conditioning and masked attention. No equations, derivations, or fitted parameters are presented that reduce any output quantity to an input quantity by construction. The central claims rest on the empirical effectiveness of the capture technique and conditioning design rather than on any self-definition or self-citation chain that would force the result.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Abstract-only review supplies insufficient detail to enumerate all free parameters or background assumptions; the central additions appear to be the hybrid capture protocol and the token-plus-masked-attention conditioning.

axioms (1)
  • domain assumption Pretrained text-to-video diffusion model supplies generative priors that transfer usefully to relighting
    Paper states it trains from a pretrained text-to-video model to leverage generative priors for high-quality output.
invented entities (1)
  • Lighting token representation no independent evidence
    purpose: Accurate lighting control via per-light tokens
    New conditioning method introduced to represent each light source as a token.

pith-pipeline@v0.9.0 · 5749 in / 1208 out tokens · 53569 ms · 2026-05-22T08:48:34.663148+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

91 extracted references · 91 canonical work pages · 1 internal anchor

  1. [1]

    Lux Post Facto: Learning Portrait Performance Relighting with Conditional Video Diffusion and a Hybrid Dataset , booktitle =

    Mei, Yiqun and He, Mingming and Ma, Li and Philip, Julien and Xian, Wenqi and George, David M and Yu, Xueming and Dedic, Gabriel and Ta. Lux Post Facto: Learning Portrait Performance Relighting with Conditional Video Diffusion and a Hybrid Dataset , booktitle =. 2025 , pages =

    work page 2025

  2. [2]

    DifFRelight: Diffusion-Based Facial Performance Relighting , year =

    He, Mingming and Clausen, Pascal and Ta. DifFRelight: Diffusion-Based Facial Performance Relighting , year =. doi:10.1145/3680528.3687644 , booktitle =

    work page doi:10.1145/3680528.3687644

  3. [3]

    2021 , issue_date =

    Bi, Sai and Lombardi, Stephen and Saito, Shunsuke and Simon, Tomas and Wei, Shih-En and Mcphail, Kevyn and Ramamoorthi, Ravi and Sheikh, Yaser and Saragih, Jason , title =. 2021 , issue_date =. doi:10.1145/3450626.3459829 , journal =

    work page doi:10.1145/3450626.3459829 2021

  4. [4]

    Simulating Spatially Varying Lighting on a Live Performance , booktitle =

    Jones, Andrew and Gardner, Andrew and Bolas, Mark and Mcdowall, Ian and Debevec, Paul , year =. Simulating Spatially Varying Lighting on a Live Performance , booktitle =

    work page

  5. [5]

    arXiv.org , volume =

    Relightable Full-body Gaussian Codec Avatars , author =. arXiv.org , volume =

    work page

  6. [6]

    CVPR , year =

    Shunsuke Saito and Gabriel Schwartz and Tomas Simon and Junxuan Li and Giljoo Nam , title =. CVPR , year =

    work page

  7. [7]

    European Conference on Computer Vision (ECCV) , year=

    Geometry-aware Single-image Full-body Human Relighting , author=. European Conference on Computer Vision (ECCV) , year=

    work page

  8. [8]

    Computer Graphics Forum (Proc

    Daichi Tajima, Yoshihiro Kanamori, Yuki Endo , title =. Computer Graphics Forum (Proc. of Eurographics 2025) , volume =

    work page 2025

  9. [9]

    Relighting human locomotion with flowed reflectance fields , year =

    Chabert, Charles-F\'. Relighting human locomotion with flowed reflectance fields , year =. doi:10.1145/1179849.1179944 , booktitle =

    work page doi:10.1145/1179849.1179944

  10. [10]

    ACM Trans

    Guo, Kaiwen and Lincoln, Peter and Davidson, Philip and Busch, Jay and Yu, Xueming and Whalen, Matt and Harvey, Geoff and Orts-Escolano, Sergio and Pandey, Rohit and Dourgarian, Jason and Tang, Danhang and Tkach, Anastasia and Kowdle, Adarsh and Cooper, Emily and Dou, Mingsong and Fanello, Sean and Fyffe, Graham and Rhemann, Christoph and Taylor, Jonathan...

    work page doi:10.1145/3355089.3356571 2019

  11. [11]

    Deep reflectance fields: high-quality facial reflectance field inference from color gradient illumination , year =

    Meka, Abhimitra and H\". Deep reflectance fields: high-quality facial reflectance field inference from color gradient illumination , year =. doi:10.1145/3306346.3323027 , journal =

    work page doi:10.1145/3306346.3323027

  12. [12]

    Deep relightable textures: volumetric performance capture with neural rendering , year =

    Meka, Abhimitra and Pandey, Rohit and H\". Deep relightable textures: volumetric performance capture with neural rendering , year =. doi:10.1145/3414685.3417814 , journal =

    work page doi:10.1145/3414685.3417814

  13. [13]

    ACM Trans

    Wenger, Andreas and Gardner, Andrew and Tchou, Chris and Unger, Jonas and Hawkins, Tim and Debevec, Paul , title =. ACM Trans. Graph. , month = jul, pages =. 2005 , issue_date =. doi:10.1145/1073204.1073258 , abstract =

    work page doi:10.1145/1073204.1073258 2005

  14. [14]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Sunstage: Portrait reconstruction and relighting using the sun as a light stage , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  15. [15]

    2007 , issue_date =

    Peers, Pieter and Tamura, Naoki and Matusik, Wojciech and Debevec, Paul , title =. 2007 , issue_date =. doi:10.1145/1276377.1276442 , journal =

    work page doi:10.1145/1276377.1276442 2007

  16. [16]

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Andrew Hou and Michel Sarkis and Ning Bi and Yiying Tong and Xiaoming Liu , title =. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    work page

  17. [17]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Kim, Hoon and Jang, Minje and Yoon, Wonjun and Lee, Jisoo and Na, Donghyun and Woo, Sanghyun , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    work page 2024

  18. [18]

    ACM Transactions on Graphics (Proceedings SIGGRAPH) , volume =

    Pandey, Rohit and Orts-Escolano, Sergio and LeGendre, Chloe and Haene, Christian and Bouaziz, Sofien and Rhemann, Christoph and Debevec, Paul and Fanello, Sean , title =. ACM Transactions on Graphics (Proceedings SIGGRAPH) , volume =. 2021 , doi =

    work page 2021

  19. [19]

    Free-viewpoint Indoor Neural Relighting from Multi-view Stereo

    Philip, Julien and Morgenthaler, S\'ebastien and Gharbi, Micha \"e l and Drettakis, George. Free-viewpoint Indoor Neural Relighting from Multi-view Stereo. ACM Transactions on Graphics. 2021

    work page 2021

  20. [20]

    ACM Transactions on Graphics (TOG) , year=

    Learning to Relight Portrait Images via a Virtual Light Stage and Synthetic-to-Real Adaptation , author=. ACM Transactions on Graphics (TOG) , year=

    work page

  21. [21]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

    SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

    work page

  22. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Futschik, David and Ritland, Kelvin and Vecore, James and Fanello, Sean and Orts-Escolano, Sergio and Curless, Brian and S\'ykora, Daniel and Pandey, Rohit , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2023 , pages =

    work page 2023

  23. [23]

    Computer Graphics Forum , volume=

    OutCast: Single Image Relighting with Cast Shadows , author=. Computer Graphics Forum , volume=. 2022 , organization=

    work page 2022

  24. [24]

    Multi-view relighting using a geometry-aware network , year =

    Philip, Julien and Gharbi, Micha\". Multi-view relighting using a geometry-aware network , year =. doi:10.1145/3306346.3323013 , journal =

    work page doi:10.1145/3306346.3323013

  25. [25]

    and Tsai, Yun-Ta and Xu, Zexiang and Yu, Xueming and Fyffe, Graham and Rhemann, Christoph and Busch, Jay and Debevec, Paul and Ramamoorthi, Ravi , title =

    Sun, Tiancheng and Barron, Jonathan T. and Tsai, Yun-Ta and Xu, Zexiang and Yu, Xueming and Fyffe, Graham and Rhemann, Christoph and Busch, Jay and Debevec, Paul and Ramamoorthi, Ravi , title =. 2019 , issue_date =. doi:10.1145/3306346.3323008 , journal =

    work page doi:10.1145/3306346.3323008 2019

  26. [26]

    2020 , issue_date =

    Wang, Zhibo and Yu, Xin and Lu, Ming and Wang, Quan and Qian, Chen and Xu, Feng , title =. 2020 , issue_date =. doi:10.1145/3414685.3417824 , journal =

    work page doi:10.1145/3414685.3417824 2020

  27. [27]

    ACM Transactions on Graphics (TOG) , volume=

    Portrait Shadow Manipulation , author=. ACM Transactions on Graphics (TOG) , volume=

    work page

  28. [28]

    , title =

    Zhou, Hao and Hadap, Sunil and Sunkavalli, Kalyan and Jacobs, David W. , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    work page

  29. [29]

    , title =

    Mei, Yiqun and Zhang, He and Zhang, Xuaner and Zhang, Jianming and Shu, Zhixin and Wang, Yilin and Wei, Zijun and Yan, Shi and Jung, HyunJoon and Patel, Vishal M. , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2023 , pages =

    work page 2023

  30. [30]

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Andrew Hou and Ze Zhang and Michel Sarkis and Ning Bi and Yiying Tong and Xiaoming Liu , title =. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    work page

  31. [31]

    , title =

    Mei, Yiqun and Zeng, Yu and Zhang, He and Shu, Zhixin and Zhang, Xuaner and Bi, Sai and Zhang, Jianming and Jung, HyunJoon and Patel, Vishal M. , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    work page 2024

  32. [32]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Ranjan, Anurag and Yi, Kwang Moo and Chang, Jen-Hao Rick and Tuzel, Oncel , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2023 , pages =

    work page 2023

  33. [33]

    Palette: Image-to-image diffusion models

    Tan, Feitong and Fanello, Sean and Meka, Abhimitra and Orts-Escolano, Sergio and Tang, Danhang and Pandey, Rohit and Taylor, Jonathan and Tan, Ping and Zhang, Yinda , title =. 2022 , isbn =. doi:10.1145/3528233.3530751 , booktitle =

    work page doi:10.1145/3528233.3530751 2022

  34. [34]

    2024 , isbn =

    Hou, Andrew and Shu, Zhixin and Zhang, Xuaner and Zhang, He and Hold-Geoffroy, Yannick and Yoon, Jae Shin and Liu, Xiaoming , title =. 2024 , isbn =. doi:10.1007/978-3-031-73030-6_20 , booktitle =

    work page doi:10.1007/978-3-031-73030-6_20 2024

  35. [35]

    LightIt: Illumination Modeling and Control for Diffusion Models , booktitle =

    Peter Kocsis and Julien Philip and Kalyan Sunkavalli and Matthias Nie. LightIt: Illumination Modeling and Control for Diffusion Models , booktitle =

    work page

  36. [36]

    Computer Graphics Forum , title =

    Poirier-Ginter, Yohan and Gauthier, Alban and Philip, Julien and Lalonde, Jean-François and Drettakis, George , year =. Computer Graphics Forum , title =

    work page

  37. [37]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Ponglertnapakorn, Puntawat and Tritrong, Nontawat and Suwajanakorn, Supasorn , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2023 , pages =

    work page 2023

  38. [38]

    2024 , isbn =

    Zeng, Chong and Dong, Yue and Peers, Pieter and Kong, Youkang and Wu, Hongzhi and Tong, Xin , title =. 2024 , isbn =. doi:10.1145/3641519.3657396 , booktitle =

    work page doi:10.1145/3641519.3657396 2024

  39. [39]

    The Thirteenth International Conference on Learning Representations , year=

    Scaling In-the-Wild Training for Diffusion-based Illumination Harmonization and Editing by Imposing Consistent Light Transport , author=. The Thirteenth International Conference on Learning Representations , year=

    work page

  40. [40]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Ren, Mengwei and Xiong, Wei and Yoon, Jae Shin and Shu, Zhixin and Zhang, Jianming and Jung, HyunJoon and Gerig, Guido and Zhang, He , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    work page 2024

  41. [41]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Zhang, Longwen and Zhang, Qixuan and Wu, Minye and Yu, Jingyi and Xu, Lan , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2021 , pages =

    work page 2021

  42. [42]

    ACM SIGGRAPH 2024 Conference Papers , year =

    Junxuan Li and Chen Cao and Gabriel Schwartz and Rawal Khirodkar and Christian Richardt and Tomas Simon and Yaser Sheikh and Shunsuke Saito , title =. ACM SIGGRAPH 2024 Conference Papers , year =

    work page 2024

  43. [43]

    SIGGRAPH 2024 Conference Proceedings , year=

    VRMM: A Volumetric Relightable Morphable Head Model , author=. SIGGRAPH 2024 Conference Proceedings , year=

    work page 2024

  44. [44]

    2025 , Eprint =

    Nadav Magar and Amir Hertz and Eric Tabellion and Yael Pritch and Alex Rav-Acha and Ariel Shamir and Yedid Hoshen , Title =. 2025 , Eprint =

    work page 2025

  45. [45]

    2025 , journal =

    LBM: Latent Bridge Matching for Fast Image-to-Image Translation , author=. 2025 , journal =

    work page 2025

  46. [46]

    2025 , eprint=

    ReLumix: Extending Image Relighting to Video via Video Diffusion Models , author=. 2025 , eprint=

    work page 2025

  47. [47]

    Advances in Neural Information Processing Systems , year =

    Neural Gaffer: Relighting Any Object via Diffusion , author =. Advances in Neural Information Processing Systems , year =

    work page

  48. [48]

    , year =

    Bharadwaj, Shrisha and Feng, Haiwen and Becherini, Giorgio and Abrevaya, Victoria Fernandez and Black, Michael J. , year =. SIGGRAPH Asia Conference Papers '25 , series =. doi:10.1145/3757377.3763970 , url =

    work page doi:10.1145/3757377.3763970

  49. [49]

    The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Ruofan Liang and Zan Gojcic and Huan Ling and Jacob Munkberg and Jon Hasselgren and Zhi-Hao Lin and Jun Gao and Alexander Keller and Nandita Vijaykumar and Sanja Fidler and Zian Wang , title =. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    work page

  50. [50]

    arXiv preprint arXiv:2506.15673 , year=

    UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting , author=. arXiv preprint arXiv:2506.15673 , year=

    work page arXiv

  51. [51]

    Proceedings of the SIGGRAPH Asia 2025 Technical Communications , articleno =

    Yu, Xueming and He, Mingming and George, David and Joshi, Avani and Debevec, Paul , title =. Proceedings of the SIGGRAPH Asia 2025 Technical Communications , articleno =. 2025 , isbn =. doi:10.1145/3757376.3771405 , abstract =

    work page doi:10.1145/3757376.3771405 2025

  52. [52]

    Proceedings of the 27th Annual Conference on Computer Graphics and Interactive Techniques , pages =

    Debevec, Paul and Hawkins, Tim and Tchou, Chris and Duiker, Haarm-Pieter and Sarokin, Westley and Sagar, Mark , title =. Proceedings of the 27th Annual Conference on Computer Graphics and Interactive Techniques , pages =. 2000 , isbn =. doi:10.1145/344779.344855 , abstract =

    work page doi:10.1145/344779.344855 2000

  53. [53]

    2002 , issue_date =

    Debevec, Paul and Wenger, Andreas and Tchou, Chris and Gardner, Andrew and Waese, Jamie and Hawkins, Tim , title =. 2002 , issue_date =. doi:10.1145/566654.566614 , journal =

    work page doi:10.1145/566654.566614 2002

  54. [54]

    2016 , issue_date =

    LeGendre, Chloe and Yu, Xueming and Liu, Dai and Busch, Jay and Jones, Andrew and Pattanaik, Sumanta and Debevec, Paul , title =. 2016 , issue_date =. doi:10.1145/2897824.2925934 , journal =

    work page doi:10.1145/2897824.2925934 2016

  55. [55]

    Animatable facial reflectance fields , year =

    Hawkins, Tim and Wenger, Andreas and Tchou, Chris and Gardner, Andrew and G\". Animatable facial reflectance fields , year =. Proceedings of the Fifteenth Eurographics Conference on Rendering Techniques , pages =

    work page

  56. [56]

    2024 , isbn =

    Bi, Zoubin and Zeng, Yixin and Zeng, Chong and Pei, Fan and Feng, Xiang and Zhou, Kun and Wu, Hongzhi , title =. 2024 , isbn =. doi:10.1145/3680528.3687576 , booktitle =

    work page doi:10.1145/3680528.3687576 2024

  57. [57]

    SIGGRAPH Asia 2024 Conference Papers , pages=

    OLAT Gaussians for Generic Relightable Appearance Acquisition , author=. SIGGRAPH Asia 2024 Conference Papers , pages=

    work page 2024

  58. [58]

    2009 , isbn =

    Fyffe, Graham , title =. 2009 , isbn =. doi:10.1145/1599301.1599381 , booktitle =

    work page doi:10.1145/1599301.1599381 2009

  59. [59]

    RGB↔X: Image decomposition and synthesis using material- and lighting-aware diffusion models , year =

    Zeng, Zheng and Deschaintre, Valentin and Georgiev, Iliyan and Hold-Geoffroy, Yannick and Hu, Yiwei and Luan, Fujun and Yan, Ling-Qi and Ha. RGB↔X: Image decomposition and synthesis using material- and lighting-aware diffusion models , year =. doi:10.1145/3641519.3657445 , booktitle =

    work page doi:10.1145/3641519.3657445

  60. [60]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Li, Zhengqin and Shafiei, Mohammad and Ramamoorthi, Ravi and Sunkavalli, Kalyan and Chandraker, Manmohan , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    work page

  61. [61]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Zhu, Rui and Li, Zhengqin and Matai, Janarbek and Porikli, Fatih and Chandraker, Manmohan , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2022 , pages =

    work page 2022

  62. [62]

    SIGGRAPH Asia 2022 Conference Papers , articleno =

    Zhu, Jingsen and Luan, Fujun and Huo, Yuchi and Lin, Zihao and Zhong, Zhihua and Xi, Dianbing and Wang, Rui and Bao, Hujun and Zheng, Jiaxiang and Tang, Rui , title =. SIGGRAPH Asia 2022 Conference Papers , articleno =. 2022 , publisher =

    work page 2022

  63. [63]

    Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Kocsis, Peter and Sitzmann, Vincent and Nie\ ss ner, Matthias , title =. Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    work page

  64. [64]

    , author=

    Lightness and retinex theory. , author=. Journal of the Optical Society of America , year=

    work page

  65. [65]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Liang, Zhihao and Zhang, Qi and Feng, Ying and Shan, Ying and Jia, Kui , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    work page 2024

  66. [66]

    CVPR , year=

    NeRV: Neural Reflectance and Visibility Fields for Relighting and View Synthesis , author=. CVPR , year=

    work page

  67. [67]

    CVPR , year=

    Modeling Indirect Illumination for Inverse Rendering , author=. CVPR , year=

    work page

  68. [68]

    Kai Zhang and Fujun Luan and Qianqian Wang and Kavita Bala and Noah Snavely , booktitle=

    work page

  69. [69]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    TensoIR: Tensorial Inverse Rendering , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    work page

  70. [70]

    arXiv preprint , url=

    3D Gaussian Inverse Rendering with Approximated Global Illumination , author=. arXiv preprint , url=

    work page

  71. [71]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Sun, Hanxiao and Gao, Yupeng and Xie, Jin and Yang, Jian and Wang, Beibei , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2025 , pages =

    work page 2025

  72. [72]

    2025 , url=

    HONGZE CHEN and Zehong Lin and Jun Zhang , booktitle=. 2025 , url=

    work page 2025

  73. [73]

    arXiv preprint arXiv:2502.08590 , year=

    Light-A-Video: Training-free Video Relighting via Progressive Light Fusion , author=. arXiv preprint arXiv:2502.08590 , year=

    work page arXiv

  74. [74]

    2025 , eprint=

    RelightVid: Temporal-Consistent Diffusion Model for Video Relighting , author=. 2025 , eprint=

    work page 2025

  75. [75]

    and Chen, Simon S

    Wang, Junying and Liu, Jingyuan and Sun, Xin and Singh, Krishna Kumar and Shu, Zhixin and Zhang, He and Yang, Jimei and Zhao, Nanxuan and Wang, Tuanfeng Y. and Chen, Simon S. and Neumann, Ulrich and Yoon, Jae Shin , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2025 , pages =

    work page 2025

  76. [76]

    Computer Graphics Forum , volume =

    Song, Guoxian and Cham, Tat-Jen and Cai, Jianfei and Zheng, Jianmin , title =. Computer Graphics Forum , volume =. doi:https://doi.org/10.1111/cgf.14384 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1111/cgf.14384 , year =

    work page doi:10.1111/cgf.14384

  77. [77]

    Computer Graphics Forum (Proc

    Daichi Tajima, Yoshihiro Kanamori, Yuki Endo , title =. Computer Graphics Forum (Proc. of Pacific Graphics 2021) , volume =

    work page 2021

  78. [78]

    Single-image Full-body Human Relighting , journal =

    Lagunas, Manuel and Sun, Xin and Yang, Jimei and Villegas, Ruben and Zhang, Jianming and Shu, Zhixin and Masia, Belen and Gutierrez, Diego , keywords =. Single-image Full-body Human Relighting , journal =. doi:10.2312/SR.20211300 , url =

    work page doi:10.2312/sr.20211300

  79. [79]

    2025 , isbn =

    Yu, Xueming and George, David and Millward, John and Debevec, Paul , title =. 2025 , isbn =. doi:10.1145/3721250.3743035 , booktitle =

    work page doi:10.1145/3721250.3743035 2025

  80. [80]

    2023 , url =

    Poly Haven HDRI , author =. 2023 , url =

    work page 2023

Showing first 80 references.