pith. sign in

arxiv: 2606.01362 · v1 · pith:USF5KWO2new · submitted 2026-05-31 · 💻 cs.GR · cs.CV

AlbedoEdit: Unified Instance-Level Video Editing with Albedo Guidance

Xilong Zhou , Bao-Huy Nguyen , Zheng Zeng , Jacob Munkberg , Jon Hasselgren , Thomas Leimk\"uhler , Nima Kalantari , Milo\v{s} Ha\v{s}an
show 1 more author
Christian Theobalt
This is my paper

Pith reviewed 2026-06-28 15:53 UTC · model grok-4.3

classification 💻 cs.GR cs.CV
keywords video editingalbedo mapgenerative modelsobject insertionobject removaltexture editinginstance-level editingvideo synthesis
0
0 comments X

The pith

AlbedoEdit conditions video generation on a user-edited first-frame albedo map to perform object insertion, removal, and texture editing while learning to add consistent lighting effects.

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

The paper introduces AlbedoEdit as a single generative model that accepts a source RGB video and an albedo map edited only in the first frame, then outputs an edited RGB video. The albedo map's lighting invariance lets users specify precise appearance changes without dealing with shadows or reflections themselves. Training occurs on a new synthetic paired dataset that covers insertion, removal, and texture tasks, allowing the model to implicitly learn harmonization and effects such as specular highlights and mirror reflections. Prior work either offered only coarse semantic control or built separate pipelines for each editing type. If the approach holds, creators gain one tool that handles multiple fine-grained video edits with automatic real-world visual consistency.

Core claim

AlbedoEdit is a unified generative video editing framework that jointly supports object insertion, object removal, and texture editing by conditioning on a user-edited first-frame albedo map and implicitly learning to harmonize edited contents and simulate complex real-world visual effects triggered by editing operations, including specular highlights, soft shadows, and mirror reflections.

What carries the argument

The albedo map, an intrinsic image invariant to lighting with no specularity, shadowing or inter-reflections, acts as the conditioning signal that specifies the desired appearance edit for the entire video sequence.

If this is right

  • A single fine-tuned model handles object insertion, object removal, and texture editing without task-specific architectures.
  • Edited videos automatically include realistic specular highlights, soft shadows, and mirror reflections triggered by the albedo change.
  • The approach outperforms prior unified and task-specific video editing methods on both qualitative and quantitative metrics.
  • Training on synthetic paired data transfers to real videos by learning implicit harmonization of edited content with the scene.

Where Pith is reading between the lines

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

  • The first-frame-only conditioning could support interactive tools where users paint albedo changes on one frame and receive a full consistent video.
  • The method might extend to other intrinsic properties such as surface normals to control additional effects like bump mapping in video.
  • If the synthetic-to-real gap proves small, similar albedo conditioning could apply to longer or higher-resolution sequences without retraining from scratch.
  • The implicit learning of lighting effects suggests the framework could be adapted for video relighting tasks where only albedo is modified.

Load-bearing premise

A user-provided edited albedo map for only the first frame plus training on synthetic paired data is sufficient for the model to correctly infer and render all subsequent-frame lighting interactions without visible artifacts.

What would settle it

Run the model on a source video containing a mirror surface, edit the first-frame albedo to introduce a new reflective object, and check whether the output video shows the expected mirror reflection and consistent shadows across later frames.

Figures

Figures reproduced from arXiv: 2606.01362 by Bao-Huy Nguyen, Christian Theobalt, Jacob Munkberg, Jon Hasselgren, Milo\v{s} Ha\v{s}an, Nima Kalantari, Thomas Leimk\"uhler, Xilong Zhou, Zheng Zeng.

Figure 1
Figure 1. Figure 1: AlbedoEdit is a unified generative video editing framework supporting object insertion, object removal, and texture editing. We leverage the intrinsic albedo map as an effective and user-friendly mechanism for specifying fine-grained appearance edits. We show object insertion, object removal, and an additional insertion example demonstrating lighting harmonization capabilities of the model. The source and … view at source ↗
Figure 2
Figure 2. Figure 2: Method overview. Given an input RGB video [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Our method takes a source video, extracts an albedo [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Editing pairs from our training dataset. The top section [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 6
Figure 6. Figure 6: Demonstration of VTE on a synthetic dataset. Our [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Demonstration of VOI tasks on the benchmark dataset with GT videos. Our method is compared against three universal video [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Demonstration of free-form edits on in-the-wild videos. From left to right, we demonstrate VOI, VOR, and VTE results, [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Left: We demonstrate the effect of inserting objects [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Limitation of our method. As shown in the figure, [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
read the original abstract

Video generative models have achieved remarkable progress in synthesizing photorealistic video sequences. However, enabling broader and more creative downstream applications requires fine-grained instance-level video editing, including object insertion, object removal, and texture editing, which has emerged as a prominent yet challenging problem. Existing approaches either propose unified generative frameworks with only coarse semantic control, or design task-specific frameworks for individual editing tasks, limiting their flexibility and applicability across diverse real-world scenarios. To address these limitations, we propose AlbedoEdit, a unified generative video editing framework that jointly supports object insertion, object removal, and texture editing. Our key insight is that the intrinsic albedo map, which is invariant to lighting and contains no specularity, shadowing and inter-reflection effects, provides an effective and user-friendly mechanism for specifying fine-grained appearance edits. Built upon video foundation models, AlbedoEdit is fine-tuned to translate source RGB videos into edited RGB videos, conditioned on a user-edited first-frame albedo. Trained on a new paired synthetic dataset covering all three editing tasks, AlbedoEdit implicitly learns to harmonize edited contents and simulate complex real-world visual effects triggered by editing operations, including specular highlights, soft shadows, and mirror reflections. AlbedoEdit demonstrates superior performance over state-of-the-art video editing approaches, both qualitatively and quantitatively. Project webpage is https://vcai.mpi-inf.mpg.de/projects/AlbedoEdit/.

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

2 major / 2 minor

Summary. The manuscript proposes AlbedoEdit, a unified generative framework for instance-level video editing that supports object insertion, removal, and texture editing. It conditions a fine-tuned video foundation model on a user-provided edited albedo map of only the first frame, using a new synthetic paired training dataset to implicitly learn harmonization of edited content with the source video and simulation of lighting effects including specular highlights, soft shadows, and mirror reflections.

Significance. If the central empirical claim holds, the work offers a flexible, albedo-based control mechanism that unifies three editing tasks previously handled by separate or coarser methods, with potential impact on creative video applications. The albedo-invariance insight is a clear strength for decoupling appearance specification from lighting.

major comments (2)
  1. [Abstract and §3 (Method)] Abstract and presumed §3 (Method): The load-bearing claim that a single first-frame edited albedo plus synthetic-pair fine-tuning suffices to infer temporally consistent lighting interactions (specular highlights, soft shadows, mirror reflections) across all frames rests on an unverified assumption that the synthetic training distribution matches real-video lighting statistics; no explicit test of out-of-distribution real videos with complex inter-reflections is described, which directly risks falsifying the 'unified generative framework' claim.
  2. [Experiments section] Presumed Experiments section: The abstract asserts superior qualitative and quantitative performance, yet without reported metrics (e.g., specific PSNR/SSIM/FID values, temporal consistency scores, or ablation on albedo vs. RGB conditioning), it is impossible to verify whether gains are attributable to albedo guidance or to the underlying foundation model; this undermines assessment of the harmonization claim.
minor comments (2)
  1. [Abstract] Abstract: The phrase 'implicitly learns to harmonize' is repeated without a concrete definition or loss term; a short clarification of the training objective would improve readability.
  2. [Abstract] Project webpage link: Ensure the supplementary material includes side-by-side comparisons on real videos with strong reflections to support the generalization claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and outline the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract and §3 (Method)] Abstract and presumed §3 (Method): The load-bearing claim that a single first-frame edited albedo plus synthetic-pair fine-tuning suffices to infer temporally consistent lighting interactions (specular highlights, soft shadows, mirror reflections) across all frames rests on an unverified assumption that the synthetic training distribution matches real-video lighting statistics; no explicit test of out-of-distribution real videos with complex inter-reflections is described, which directly risks falsifying the 'unified generative framework' claim.

    Authors: We acknowledge that the core generalization from synthetic pairs to real-video lighting statistics is an assumption that merits stronger support. The video foundation model is pre-trained on large-scale real footage, and our qualitative results on real test videos show plausible handling of specular highlights, soft shadows, and reflections. Nevertheless, we agree that dedicated examples of out-of-distribution real videos with complex inter-reflections are not explicitly presented. In the revision we will add a new subsection (or appendix) with such examples together with a discussion of the synthetic dataset’s coverage and remaining limitations. revision: yes

  2. Referee: [Experiments section] Presumed Experiments section: The abstract asserts superior qualitative and quantitative performance, yet without reported metrics (e.g., specific PSNR/SSIM/FID values, temporal consistency scores, or ablation on albedo vs. RGB conditioning), it is impossible to verify whether gains are attributable to albedo guidance or to the underlying foundation model; this undermines assessment of the harmonization claim.

    Authors: We apologize that the specific numerical values and the requested ablation were not presented with sufficient clarity. The experiments section does contain quantitative comparisons, but we will revise the manuscript to explicitly tabulate PSNR, SSIM, FID, and temporal-consistency scores, and to add an ablation that isolates albedo conditioning from RGB conditioning. These additions will make the contribution of the albedo guidance transparent. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical ML training on synthetic pairs

full rationale

The paper presents a fine-tuned video generative model that takes a user-edited first-frame albedo as conditioning input and is trained end-to-end on a new synthetic paired dataset to produce edited RGB video. No equations, parameter-fitting steps, or derivation chain are described that would reduce any output quantity to the input by construction. The central claim rests on empirical performance after training rather than on self-definition, fitted-input predictions, or load-bearing self-citations of uniqueness theorems. The approach is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach assumes that albedo maps can be reliably extracted or provided by users and that synthetic training data sufficiently covers the distribution of real-world lighting interactions triggered by edits.

axioms (2)
  • domain assumption Albedo maps are invariant to lighting and contain no specularity, shadowing or inter-reflection effects.
    Stated in the abstract as the key insight enabling user-friendly control.
  • domain assumption Fine-tuning a video foundation model on synthetic paired data will generalize to produce realistic lighting effects on real videos.
    Implicit in the training description; no explicit validation on real data is mentioned.

pith-pipeline@v0.9.1-grok · 5818 in / 1317 out tokens · 16662 ms · 2026-06-28T15:53:26.094242+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

57 extracted references · 31 canonical work pages · 12 internal anchors

  1. [1]

    Abdal, O

    R. Abdal, O. Patashnik, I. Skorokhodov, W. Menapace, A. Siarohin, S. Tulyakov, D. Cohen-Or, and K. Aberman. Dynamic concepts personalization from single videos. In Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers, pages 1–9, 2025

    2025

  2. [2]

    Adobe photoshop.https://www.adobe

    Adobe Inc. Adobe photoshop.https://www.adobe. com/products/photoshop.html, 2026

    2026

  3. [3]

    C. Bai, Z. Shao, G. Zhang, D. Liang, J. Yang, Z. Zhang, Y . Guo, C. Zhong, Y . Qiu, Z. Wang, et al. Anything in any scene: Photorealistic video object insertion.arXiv preprint arXiv:2401.17509, 2024

    work page arXiv 2024

  4. [4]

    J. Bai, S. Bai, Y . Chu, Z. Cui, K. Dang, X. Deng, Y . Fan, W. Ge, Y . Han, F. Huang, et al. Qwen technical report.arXiv preprint arXiv:2309.16609, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  5. [5]

    Batifol, A

    S. Batifol, A. Blattmann, F. Boesel, S. Consul, C. Diagne, T. Dockhorn, J. English, Z. English, P. Esser, S. Kulal, et al. Flux. 1 kontext: Flow matching for in-context image genera- tion and editing in latent space.arXiv e-prints, pages arXiv– 2506, 2025

    2025

  6. [6]

    Y . Bian, Z. Zhang, X. Ju, M. Cao, L. Xie, Y . Shan, and Q. Xu. Videopainter: Any-length video inpainting and editing with plug-and-play context control. InProceedings of the Special Interest Group on Computer Graphics and Interactive Tech- niques Conference Conference Papers, pages 1–12, 2025

    2025

  7. [7]

    Deitke, D

    M. Deitke, D. Schwenk, J. Salvador, L. Weihs, O. Michel, E. VanderBilt, L. Schmidt, K. Ehsani, A. Kembhavi, and A. Farhadi. Objaverse: A universe of annotated 3d objects. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 13142–13153, 2023

    2023

  8. [8]

    Dhariwal and A

    P. Dhariwal and A. Nichol. Diffusion models beat gans on image synthesis.Advances in neural information processing systems, 34:8780–8794, 2021

    2021

  9. [9]

    Y . Fang, T. Wu, V . Deschaintre, D. Ceylan, I. Georgiev, C.- H. P. Huang, Y . Hu, X. Chen, and T. Y . Wang. V-rgbx: Video editing with accurate controls over intrinsic properties.arXiv preprint arXiv:2512.11799, 2025

    work page arXiv 2025

  10. [10]

    Y . Fu, Y . Zheng, Z. Dai, and H. Ding. Effecterase: Joint video object removal and insertion for high-quality effect erasing.arXiv preprint arXiv:2603.19224, 2026

    work page arXiv 2026

  11. [11]

    X. Gao, R. Li, X. Chen, Y . Wu, S. Feng, Q. Yin, and Z. Tu. Pisco: Precise video instance insertion with sparse control. arXiv preprint arXiv:2602.08277, 2026

    work page arXiv 2026

  12. [12]

    Garibi, S

    D. Garibi, S. Yadin, R. Paiss, O. Tov, S. Zada, A. Ephrat, T. Michaeli, I. Mosseri, and T. Dekel. Tokenverse: Versa- tile multi-concept personalization in token modulation space. ACM Transactions On Graphics (TOG), 44(4):1–11, 2025

    2025

  13. [13]

    Z. Gu, R. Yan, J. Lu, P. Li, Z. Dou, C. Si, Z. Dong, Q. Liu, C. Lin, Z. Liu, et al. Diffusion as shader: 3d-aware video dif- fusion for versatile video generation control. InProceedings of the Special Interest Group on Computer Graphics and In- teractive Techniques Conference Conference Papers, pages 1–12, 2025

    2025

  14. [14]

    LTX-2: Efficient Joint Audio-Visual Foundation Model

    Y . HaCohen, B. Brazowski, N. Chiprut, Y . Bitterman, A. Kvochko, A. Berkowitz, D. Shalem, D. Lifschitz, D. Moshe, E. Porat, et al. Ltx-2: Efficient joint audio-visual foundation model.arXiv preprint arXiv:2601.03233, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  15. [15]

    H. He, Y . Xu, Y . Guo, G. Wetzstein, B. Dai, H. Li, and C. Yang. Cameractrl: Enabling camera control for text-to- video generation.arXiv preprint arXiv:2404.02101, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  16. [16]

    K. He, R. Liang, J. Munkberg, J. Hasselgren, N. Vijaykumar, A. Keller, S. Fidler, I. Gilitschenski, Z. Gojcic, and Z. Wang. Unirelight: Learning joint decomposition and synthesis for video relighting.arXiv preprint arXiv:2506.15673, 2025

    work page arXiv 2025

  17. [17]

    J. Ho, A. Jain, and P. Abbeel. Denoising diffusion proba- bilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  18. [18]

    W. Hong, M. Ding, W. Zheng, X. Liu, and J. Tang. Cogvideo: Large-scale pretraining for text-to-video gener- ation via transformers.arXiv preprint arXiv:2205.15868, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  19. [19]

    Huang, Y

    Z. Huang, Y . He, J. Yu, F. Zhang, C. Si, Y . Jiang, Y . Zhang, T. Wu, Q. Jin, N. Chanpaisit, Y . Wang, X. Chen, L. Wang, D. Lin, Y . Qiao, and Z. Liu. VBench: Comprehensive bench- mark suite for video generative models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  20. [20]

    Jiang, Z

    Z. Jiang, Z. Han, C. Mao, J. Zhang, Y . Pan, and Y . Liu. Vace: All-in-one video creation and editing. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 17191–17202, 2025

    2025

  21. [21]

    H. Jin, H. Jang, J. Kim, J. Hyung, K. Kim, D. Kim, H. Choi, H. Kim, and J. Choo. Insertanywhere: Bridging 4d scene geometry and diffusion models for realistic video object in- sertion.arXiv preprint arXiv:2512.17504, 2025

    work page internal anchor Pith review arXiv 2025

  22. [22]

    D. P. Kingma and M. Welling. Auto-encoding variational bayes.arXiv preprint arXiv:1312.6114, 2013

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  23. [23]

    W. Kong, Q. Tian, Z. Zhang, R. Min, Z. Dai, J. Zhou, J. Xiong, X. Li, B. Wu, J. Zhang, et al. Hunyuanvideo: A sys- tematic framework for large video generative models.arXiv preprint arXiv:2412.03603, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  24. [24]

    M. Ku, C. Wei, W. Ren, H. Yang, and W. Chen. Anyv2v: A tuning-free framework for any video-to-video editing tasks. arXiv preprint arXiv:2403.14468, 2024

    work page arXiv 2024

  25. [25]

    Y .-C. Lee, Z. Zhang, J. Huang, J.-H. Wang, J.-Y . Lee, J.- B. Huang, E. Shechtman, and Z. Li. Generative video motion editing with 3d point tracks.arXiv preprint arXiv:2512.02015, 2025

    work page arXiv 2025

  26. [26]

    X. Li, H. Xue, P. Ren, and L. Bo. Diffueraser: A diffusion model for video inpainting.arXiv preprint arXiv:2501.10018, 2025

    work page arXiv 2025

  27. [27]

    Liang, Z

    R. Liang, Z. Gojcic, H. Ling, J. Munkberg, J. Hasselgren, C.- H. Lin, J. Gao, A. Keller, N. Vijaykumar, S. Fidler, et al. Dif- fusion renderer: Neural inverse and forward rendering with video diffusion models. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 26069– 26080, 2025

    2025

  28. [28]

    C. Lin, H. Liu, Q. Lin, Z. Bright, S. Tang, Y . He, M. Liu, L. Zhu, and C. Le. Objaverse++: Curated 3D Object Dataset with Quality Annotations. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Work- shops, pages 6813–6822, 2025

    2025

  29. [29]

    Flow Matching for Generative Modeling

    Y . Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le. Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  30. [30]

    Decoupled Weight Decay Regularization

    I. Loshchilov and F. Hutter. Decoupled weight decay regu- larization.arXiv preprint arXiv:1711.05101, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  31. [31]

    L. Lyu, V . Deschaintre, Y . Hold-Geoffroy, M. Ha ˇsan, J. S. Yoon, T. Leimk¨uhler, C. Theobalt, and I. Georgiev. Intrin- sicedit: Precise generative image manipulation in intrinsic space.ACM Transactions on Graphics (TOG), 44(4):1–13, 2025

    2025

  32. [32]

    X. Ma, V . Deschaintre, M. Haˇsan, F. Luan, K. Zhou, H. Wu, and Y . Hu. Materialpicker: Multi-modal dit-based mate- rial generation.ACM Transactions on Graphics, 44(4):1–12, July 2025. ISSN 1557-7368. doi: 10.1145/3731199. URL http://dx.doi.org/10.1145/3731199

    work page doi:10.1145/3731199 2025

  33. [33]

    C. Miao, Y . Feng, J. Zeng, Z. Gao, H. Liu, Y . Yan, D. Qi, X. Chen, B. Wang, and H. Zhao. Rose: Remove objects with side effects in videos.arXiv preprint arXiv:2508.18633, 2025

    work page arXiv 2025

  34. [34]

    Motamed, W

    S. Motamed, W. Harvey, B. Klein, L. Van Gool, Z. Yuan, and T.-Y . Cheng. V oid: Video object and interaction deletion. arXiv preprint arXiv:2604.02296, 2026

    work page arXiv 2026

  35. [35]

    Peebles and S

    W. Peebles and S. Xie. Scalable diffusion models with trans- formers. InProceedings of the IEEE/CVF international con- ference on computer vision, pages 4195–4205, 2023

    2023

  36. [36]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. InInternational conference on machine learn- ing, pages 8748–8763. PmLR, 2021

    2021

  37. [37]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Om- mer. High-resolution image synthesis with latent diffu- sion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684– 10695, 2022

    2022

  38. [38]

    Saini, N

    N. Saini, N. Bodla, A. Shrivastava, A. Ravichandran, X. Zhang, A. Shrivastava, and B. Singh. Invi: Object in- sertion in videos using off-the-shelf diffusion models.arXiv preprint arXiv:2407.10958, 2024

    work page arXiv 2024

  39. [39]

    J. Shin, Z. Li, R. Zhang, J.-Y . Zhu, J. Park, E. Shecht- man, and X. Huang. Motionstream: Real-time video gen- eration with interactive motion controls.arXiv preprint arXiv:2511.01266, 2025

    work page arXiv 2025

  40. [40]

    Sohl-Dickstein, E

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli. Deep unsupervised learning using nonequilib- rium thermodynamics. InInternational conference on ma- chine learning, pages 2256–2265. pmlr, 2015

    2015

  41. [41]

    Van Den Oord, O

    A. Van Den Oord, O. Vinyals, et al. Neural discrete represen- tation learning.Advances in neural information processing systems, 30, 2017

    2017

  42. [42]

    T. Wan, A. Wang, B. Ai, B. Wen, C. Mao, C.-W. Xie, D. Chen, F. Yu, H. Zhao, J. Yang, et al. Wan: Open and ad- vanced large-scale video generative models.arXiv preprint arXiv:2503.20314, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  43. [43]

    B. Wang, X. Chen, M. Gadelha, and Z. Cheng. Frame in- n-out: Unbounded controllable image-to-video generation. arXiv preprint arXiv:2505.21491, 2025

    work page arXiv 2025

  44. [44]

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli. Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4): 600–612, 2004

    2004

  45. [45]

    C. Wei, Q. Liu, Z. Ye, Q. Wang, X. Wang, P. Wan, K. Gai, and W. Chen. Univideo: Unified understanding, generation, and editing for videos.arXiv preprint arXiv:2510.08377, 2025

    work page arXiv 2025

  46. [46]

    Video models are zero-shot learners and reasoners

    T. Wiedemer, Y . Li, P. Vicol, S. S. Gu, N. Matarese, K. Swersky, B. Kim, P. Jaini, and R. Geirhos. Video mod- els are zero-shot learners and reasoners.arXiv preprint arXiv:2509.20328, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  47. [47]

    Z. Yang, J. Teng, W. Zheng, M. Ding, S. Huang, J. Xu, Y . Yang, W. Hong, X. Zhang, G. Feng, et al. Cogvideox: Text-to-video diffusion models with an expert transformer. arXiv preprint arXiv:2408.06072, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  48. [48]

    Z. Ye, X. He, Q. Liu, Q. Wang, X. Wang, P. Wan, D. Zhang, K. Gai, Q. Chen, and W. Luo. Unic: Unified in-context video editing.arXiv preprint arXiv:2506.04216, 2025

    work page arXiv 2025

  49. [49]

    M. Yu, W. Hu, J. Xing, and Y . Shan. Trajectorycrafter: Redi- recting camera trajectory for monocular videos via diffusion models. InProceedings of the IEEE/CVF international con- ference on computer vision, pages 100–111, 2025

    2025

  50. [50]

    X. Yu, T. Wang, S. Y . Kim, P. Guerrero, X. Chen, Q. Liu, Z. Lin, and X. Qi. Objectmover: Generative object move- ment with video prior. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 17682–17691, 2025

    2025

  51. [51]

    Z. Zeng, V . Deschaintre, I. Georgiev, Y . Hold-Geoffroy, Y . Hu, F. Luan, L.-Q. Yan, and M. Ha ˇsan. Rgb¡-¿x: Image decomposition and synthesis using material- and lighting-aware diffusion models. InACM SIGGRAPH 2024 Conference Papers, SIGGRAPH ’24, New York, NY , USA, 2024. Association for Computing Machinery. ISBN 9798400705250. doi: 10.1145/3641519.36...

    work page doi:10.1145/3641519.3657445 2024

  52. [52]

    Zhang, A

    L. Zhang, A. Rao, and M. Agrawala. Adding conditional control to text-to-image diffusion models. InProceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023

    2023

  53. [53]

    Zhang, P

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang. The unreasonable effectiveness of deep features as a percep- tual metric. InProceedings of the IEEE conference on com- puter vision and pattern recognition, pages 586–595, 2018

    2018

  54. [54]

    Diffusionharmonizer: Bridging neural reconstruction and photorealistic simulation with online diffusion enhancer.arXiv preprint arXiv:2602.24096, 2026

    Y . Zhang, K. T´othov´a, Z. Wang, K. Yin, H. Turki, R. de Lu- tio, Y .-Y . Chang, O. Litany, S. Fidler, and Z. Gojcic. Dif- fusionharmonizer: Bridging neural reconstruction and pho- torealistic simulation with online diffusion enhancer.arXiv preprint arXiv:2602.24096, 2026

    work page arXiv 2026

  55. [55]

    Q. Zhao, Z. Ma, and P. Zhou. Dreaminsert: Zero-shot image- to-video object insertion from a single image.arXiv preprint arXiv:2503.10342, 2025

    work page arXiv 2025

  56. [56]

    S. Zhou, C. Li, K. C. Chan, and C. C. Loy. Propainter: Im- proving propagation and transformer for video inpainting. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10477–10486, 2023

    2023

  57. [57]

    B. Zi, S. Zhao, X. Qi, J. Wang, Y . Shi, Q. Chen, B. Liang, R. Xiao, K.-F. Wong, and L. Zhang. Cococo: Improving text- guided video inpainting for better consistency, controllability and compatibility. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 11067–11076, 2025

    2025