pith. sign in

arxiv: 2606.30003 · v1 · pith:AFBET2XDnew · submitted 2026-06-29 · 💻 cs.CV

GeoEdit: Geometry-Aware Object Editing via Dual-Branch Denoising

Yi He , Jiangming Wang , Xinyu Wang , Mark Fong , Songchun Zhang , Yuxuan Xue , Hai-Tao Zheng , Yue Ma This is my paper

Pith reviewed 2026-06-30 06:46 UTC · model grok-4.3

classification 💻 cs.CV
keywords object editingdiffusion models3D geometryimage manipulationtraining-freedual-branch denoisingpose-aware evaluationstructural depth map
0
0 comments X

The pith

GeoEdit decouples scenes in 3D then applies dual-branch denoising to enforce rigid object geometry while freeing background synthesis.

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

The paper introduces a training-free method to translate, rotate, or scale objects inside a single photograph while obeying 3D physical constraints. Existing 2D diffusion editors produce perspective violations or artifacts because they cannot handle the asymmetric requirements that the moved object must stay rigid yet the revealed background must synthesize freely. GeoEdit lifts the image to 3D, aligns object and scene via point correspondence, renders a geometry-aligned proxy using a structural depth map, and refines it with dual-branch denoising that injects constraints only into the foreground at matching noise variance. This leaves self-attention undisturbed so the object identity is preserved by a video diffusion backbone and the background remains unconstrained. The approach is evaluated on a new pose-aware benchmark covering translation, rotation, and camera movement.

Core claim

We present GeoEdit, a training-free Lift-Manipulate-Render-Denoise pipeline that decouples scene and object in 3D, aligns them through point correspondence, and renders a geometry-aligned proxy with a structural depth map. A Dual-Branch Denoising stage then refines this proxy: a video diffusion backbone preserves object identity, while 3D constraints are injected into the foreground within a narrow denoising window at matching noise variance (variance-homogeneous injection). The background denoises freely. Because the injected signal matches the native latent statistics, self-attention stays undisturbed.

What carries the argument

Dual-Branch Denoising with variance-homogeneous injection, which applies 3D constraints only to the foreground branch inside a narrow denoising window while the background branch remains free.

If this is right

  • Object edits obey rigid 3D geometry without training.
  • Background regions synthesize plausible content without leakage from the object.
  • Identity of the manipulated object is preserved across the edit.
  • The method handles translation, rotation, and camera movement with pose-aware metrics.
  • No additional model training is required beyond the base video diffusion backbone.

Where Pith is reading between the lines

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

  • The variance-homogeneous injection technique could be tested on other asymmetric editing tasks such as adding or removing objects.
  • GeoEditBench may become a reference set for measuring geometric fidelity in future diffusion editors.
  • Extending the narrow denoising window to multiple scales might further reduce residual artifacts in complex scenes.
  • The lift-to-3D step could be replaced by other depth estimators to check robustness of the overall pipeline.

Load-bearing premise

Accurate 3D decoupling via point correspondence and structural depth map rendering can be performed without introducing artifacts that later denoising cannot correct, and the injected signal at matching noise variance leaves self-attention undisturbed.

What would settle it

Running the pipeline on a test image where the rendered proxy has visible misalignment from point correspondence errors, then checking whether the final output still shows perspective violations or ghosting that the dual-branch stage fails to remove.

Figures

Figures reproduced from arXiv: 2606.30003 by Hai-Tao Zheng, Jiangming Wang, Mark Fong, Songchun Zhang, Xinyu Wang, Yi He, Yue Ma, Yuxuan Xue.

Figure 1
Figure 1. Figure 1: Showcase of proposed GeoEdit. In this paper, we propose GeoEdit, a training-free pipeline that lifts editing into 3D for physically plausible object manipulation, without external 3D software or synthetic training data. Abstract Precisely manipulating objects in a single photograph (translation, rotation, scaling) while obeying 3D physical constraints remains unsolved for diffusion-based editors. Current 2… view at source ↗
Figure 2
Figure 2. Figure 2: Comparison with previous approaches on geometry￾aware object manipulation. Given a source image and an instruc￾tion requiring a 120◦ rotation and 1.2× scaling, existing meth￾ods struggle to maintain geometric consistency. In contrast, our method faithfully follows the specified transformation while pre￾serving object identity and producing coherent, realistic results. 2. Related Work Image editing with dif… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the proposed framework. Top: Decoupled 3D reconstruction and precise alignment pipeline. Bottom: Dual￾branch denoising architecture featuring warm-start initialization and variance-homogeneous injection. where p = P comp f g (u) and q = P vis f g(u) denote the matched 3D points. This aligns P comp f g into the global scene space, establishing a unified coordinate frame in which the user can pre… view at source ↗
Figure 4
Figure 4. Figure 4: Visualizing the generative trade-off. Different con￾figurations of initialization (tweak) and injection (tstrong) timesteps dictate whether the model leans towards preserving the rigid ob￾ject skeleton or hallucinating semantic background details. forces strict adherence to the proxy but retains coarse, unre￾alistic background artifacts. We argue that the manipulated foreground and the unedited background … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison of different methods on object manipulation tasks. Our model achieves superior performance compared to state-of-the-art methods in background preservation and geometric consistency [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visual Ablation on Timestep Thresholds. We illus￾trate the fundamental trade-off between semantic realism (left) and geometric structure preservation (right). Relying predomi￾nantly on the generative prior (e.g., tw = 50, ts = 50) grants ex￾cessive freedom, resulting in structural deviation from the proxy. Conversely, excessive proxy injection at low noise levels (e.g., tw = 1, ts = 1) rigidly preserves ge… view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative Ablation on Proposed modules. We demonstrate the visual impact of each core module. The Naive Baseline struggles with both 3D skeleton preservation and back￾ground fidelity, yielding a distorted pose and altered context. Re￾moving the warm-start initialization (w/o Warm-Start) results in an unnatural background synthesis, failing to smoothly harmonize the generated textures with the original sc… view at source ↗
read the original abstract

Precisely manipulating objects in a single photograph (translation, rotation, scaling) while obeying 3D physical constraints remains unsolved for diffusion-based editors. Current 2D methods lack spatial awareness and produce perspective violations. Forcing structural proxies into the latent space also disrupts variance homogeneity, and the resulting self-attention leakage leads to ghosting and background blur. The core difficulty is asymmetric: the relocated object must follow a rigid geometry, yet the uncovered background needs freedom to synthesize plausible content. We present GeoEdit, a training-free Lift-Manipulate-Render-Denoise pipeline that satisfies both constraints. We decouple scene and object in 3D, align them through point correspondence, and render a geometry-aligned proxy with a structural depth map. A Dual-Branch Denoising stage then refines this proxy: a video diffusion backbone preserves object identity, while 3D constraints are injected into the foreground within a narrow denoising window at matching noise variance (variance-homogeneous injection). The background denoises freely. Because the injected signal matches the native latent statistics, self-attention stays undisturbed. We also introduce GeoEditBench, a pose-aware benchmark covering object translation, object rotation, and camera movement with pose-aware evaluation metrics. Experiments confirm consistent gains in geometric accuracy, identity fidelity, and background quality. Our codes are available at https://github.com/Heey731/GeoEdit.

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 / 1 minor

Summary. The paper claims to introduce GeoEdit, a training-free Lift-Manipulate-Render-Denoise pipeline for 3D geometry-aware object editing (translation, rotation, scaling) in single images. It decouples scene/object via 3D point correspondence, renders a geometry-aligned proxy with structural depth map, and applies Dual-Branch Denoising: a video diffusion backbone preserves object identity while variance-homogeneous injection enforces rigid 3D constraints on the foreground within a narrow denoising window, allowing free background synthesis. The key assertion is that matching native latent statistics leaves self-attention undisturbed. It also introduces GeoEditBench with pose-aware metrics and reports consistent gains in geometric accuracy, identity fidelity, and background quality. Code is released at https://github.com/Heey731/GeoEdit.

Significance. If the central claims hold, the work would be significant for diffusion-based editing by addressing the asymmetric constraints of rigid object manipulation versus free background synthesis without requiring training or fine-tuning. The training-free nature, explicit code release, and introduction of a pose-aware benchmark are strengths that support reproducibility and further evaluation.

major comments (2)
  1. [Abstract] Abstract: The claim that 'Because the injected signal matches the native latent statistics, self-attention stays undisturbed' is load-bearing for the Dual-Branch Denoising separation of rigid object vs. free background constraints, yet no equations, attention-map analysis, or ablation is referenced to show that variance matching alone prevents shifts in attention weights from mismatched spatial structure or edge statistics in the depth-rendered proxy.
  2. [Abstract] Abstract (and implied method section): The pipeline relies on accurate 3D decoupling and structural depth map rendering without introducing uncorrectable artifacts, but the abstract supplies no error analysis, failure cases, or quantitative validation of the point correspondence step under the claimed manipulations (translation/rotation/camera movement).
minor comments (1)
  1. [Abstract] The abstract mentions 'consistent gains' but does not name the baselines or report specific metric values; this should be expanded with quantitative results from the experiments section.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below with clarifications from the manuscript and indicate planned revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that 'Because the injected signal matches the native latent statistics, self-attention stays undisturbed' is load-bearing for the Dual-Branch Denoising separation of rigid object vs. free background constraints, yet no equations, attention-map analysis, or ablation is referenced to show that variance matching alone prevents shifts in attention weights from mismatched spatial structure or edge statistics in the depth-rendered proxy.

    Authors: The manuscript explains the variance-homogeneous injection in Section 3.3 as matching both mean and variance of the native latents at each timestep, which preserves the input distribution to self-attention layers. This is supported by the ablation in Section 4.3 (Table 3) showing degraded identity preservation and increased ghosting when variance matching is removed. We agree that direct attention-map evidence would strengthen the claim and will add equations for the injection process plus attention visualization comparisons in the revised method section, with a reference added to the abstract. revision: yes

  2. Referee: [Abstract] Abstract (and implied method section): The pipeline relies on accurate 3D decoupling and structural depth map rendering without introducing uncorrectable artifacts, but the abstract supplies no error analysis, failure cases, or quantitative validation of the point correspondence step under the claimed manipulations (translation/rotation/camera movement).

    Authors: Section 4.2 and Table 2 report quantitative results on GeoEditBench for all three manipulation types using pose-aware metrics that directly measure geometric fidelity after point correspondence and rendering. Qualitative failure cases arising from correspondence errors are shown in the supplementary material. The abstract is concise by design, but we will revise it to reference the benchmark validation and error analysis already present in the full paper. revision: yes

Circularity Check

0 steps flagged

No circularity detected in derivation chain

full rationale

The paper describes a training-free Lift-Manipulate-Render-Denoise pipeline for geometry-aware editing, relying on 3D decoupling via point correspondence, structural depth map rendering, and variance-homogeneous injection into a dual-branch denoising process. No equations, fitted parameters, or quantitative predictions appear in the abstract or described method that reduce by construction to the inputs. Claims about self-attention remaining undisturbed are asserted as following from matching native latent statistics, but without any self-definitional loop, fitted-input renaming, or load-bearing self-citation chain. The introduction of GeoEditBench and empirical results provide external validation points independent of the method's internal logic. This is a standard descriptive methods paper with no detectable circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are quantified. The method implicitly relies on standard diffusion model assumptions and introduces new pipeline stages whose details are not provided.

axioms (1)
  • domain assumption Diffusion models operate on latent spaces where variance-homogeneous signals can be injected without disrupting self-attention
    Invoked to justify the dual-branch stage.
invented entities (2)
  • Dual-Branch Denoising no independent evidence
    purpose: Separate object identity preservation from free background synthesis
    Core new component described in the pipeline.
  • variance-homogeneous injection no independent evidence
    purpose: Inject 3D constraints at matching noise variance to avoid self-attention leakage
    Technique introduced to solve the asymmetric constraint problem.

pith-pipeline@v0.9.1-grok · 5800 in / 1283 out tokens · 33640 ms · 2026-06-30T06:46:29.985864+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

85 extracted references · 30 canonical work pages · 14 internal anchors

  1. [1]

    6, 7, 14

    Gemini 2.5 flash image (nano banana) — google ai stu- dio.https://aistudio.google.com/models/ gemini-2-5-flash-image. 6, 7, 14

  2. [2]

    Blended diffusion for text-driven editing of natural images

    Omri Avrahami, Dani Lischinski, and Ohad Fried. Blended diffusion for text-driven editing of natural images. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 18208–18218, 2022. 3, 5

    2022

  3. [3]

    Stephen Batifol, Andreas Blattmann, Frederic Boesel, Sak- sham Consul, Cyril Diagne, Tim Dockhorn, Jack English, Zion English, Patrick Esser, Sumith Kulal, et al. Flux. 1 kontext: Flow matching for in-context image generation and editing in latent space.arXiv e-prints, pages arXiv–2506,

  4. [4]

    In- structpix2pix: Learning to follow image editing instructions

    Tim Brooks, Aleksander Holynski, and Alexei A Efros. In- structpix2pix: Learning to follow image editing instructions. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 18392–18402, 2023. 2

    2023

  5. [5]

    FreeOrbit4D: Training-Free Arbitrary Camera Redirection for Monocular Videos via Foreground-Complete 4D Reconstruction

    Wei Cao, Hao Zhang, Fengrui Tian, Yulun Wu, Yingying Li, Shenlong Wang, Ning Yu, and Yaoyao Liu. Freeor- bit4d: Training-free arbitrary camera redirection for monoc- ular videos via geometry-complete 4d reconstruction.arXiv preprint arXiv:2601.18993, 2026. 2, 3

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  6. [6]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. InPro- ceedings of the IEEE/CVF international conference on com- puter vision, pages 9650–9660, 2021. 6

    2021

  7. [7]

    ShapeNet: An Information-Rich 3D Model Repository

    Angel X Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, et al. Shapenet: An information-rich 3d model repository.arXiv preprint arXiv:1512.03012, 2015. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  8. [8]

    Blenderfusion: 3d-grounded vi- sual editing and generative compositing.arXiv preprint arXiv:2506.17450, 2025

    Jiacheng Chen, Ramin Mehran, Xuhui Jia, Saining Xie, and Sanghyun Woo. Blenderfusion: 3d-grounded vi- sual editing and generative compositing.arXiv preprint arXiv:2506.17450, 2025. 2, 3

    work page arXiv 2025

  9. [9]

    ContextFlow: Training-free video object editing via adaptive context enrichment.arXiv preprint arXiv:2509.17818, 2025

    Yiyang Chen, Xuanhua He, Xiujun Ma, and Yue Ma. Con- textflow: Training-free video object editing via adaptive con- text enrichment.arXiv preprint arXiv:2509.17818, 2025. 2

    work page arXiv 2025

  10. [10]

    Diffedit: Diffusion-based seman- tic image editing with mask guidance.arXiv preprint arXiv:2210.11427, 2022

    Guillaume Couairon, Jakob Verbeek, Holger Schwenk, and Matthieu Cord. Diffedit: Diffusion-based seman- tic image editing with mask guidance.arXiv preprint arXiv:2210.11427, 2022. 3

    work page arXiv 2022

  11. [11]

    Objaverse: A universe of annotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsani, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A universe of annotated 3d objects. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 13142–13153, 2023. 2

    2023

  12. [12]

    Dit4edit: Dif- fusion transformer for image editing

    Kunyu Feng, Yue Ma, Bingyuan Wang, Chenyang Qi, Haozhe Chen, Qifeng Chen, and Zeyu Wang. Dit4edit: Dif- fusion transformer for image editing. InProceedings of 9 the AAAI Conference on Artificial Intelligence, pages 2969– 2977, 2025. 2

    2025

  13. [13]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981

    Martin A Fischler and Robert C Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981. 13

    1981

  14. [14]

    DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data

    Stephanie Fu, Netanel Tamir, Shobhita Sundaram, Lucy Chai, Richard Zhang, Tali Dekel, and Phillip Isola. Dream- sim: Learning new dimensions of human visual similar- ity using synthetic data.arXiv preprint arXiv:2306.09344,

    work page internal anchor Pith review Pith/arXiv arXiv

  15. [15]

    PAI-Studio: Cinematic Video Background Replacement with Camera-Aware Motion

    Heyuan Gao, Bangxun Tang, Yiren Song, Guian Fang, Zijian He, Jie Yang, and Mike Zheng Shou. Pai-studio: Cinematic video background replacement with camera-aware motion. arXiv preprint arXiv:2606.01399, 2026. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  16. [16]

    Id-animator: Zero-shot identity-preserving human video generation.arXiv preprint arXiv:2404.15275, 2024

    Xuanhua He, Quande Liu, Shengju Qian, Xin Wang, Tao Hu, Ke Cao, Keyu Yan, and Jie Zhang. Id-animator: Zero-shot identity-preserving human video generation.arXiv preprint arXiv:2404.15275, 2024. 2

    work page arXiv 2024

  17. [17]

    Fulldit2: Efficient in-context con- ditioning for video diffusion transformers.arXiv preprint arXiv:2506.04213, 2025

    Xuanhua He, Quande Liu, Zixuan Ye, Weicai Ye, Qi- ulin Wang, Xintao Wang, Qifeng Chen, Pengfei Wan, Di Zhang, and Kun Gai. Fulldit2: Efficient in-context con- ditioning for video diffusion transformers.arXiv preprint arXiv:2506.04213, 2025. 2

    work page arXiv 2025

  18. [18]

    Classifier-Free Diffusion Guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022. 13

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  19. [19]

    Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020. 2, 3

    2020

  20. [20]

    Ex- posure bias can alleviate itself via directional and frequency rectification in flow matching, 2026

    Guanbo Huang, Jingjia Mao, Fanding Huang, Fengkai Liu, Xiangyang Luo, Yaoyuan Liang, Jiasheng Lu, Xiaoe Wang, Pei Liu, Ruiliu Fu, Ruqi Huang, and Shao-Lun Huang. Ex- posure bias can alleviate itself via directional and frequency rectification in flow matching, 2026. 3

    2026

  21. [21]

    Image-to-image translation with conditional adver- sarial networks

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. Image-to-image translation with conditional adver- sarial networks. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 1125–1134,

  22. [22]

    Vace: All-in-one video creation and editing

    Zeyinzi Jiang, Zhen Han, Chaojie Mao, Jingfeng Zhang, Yulin Pan, and Yu Liu. Vace: All-in-one video creation and editing. InProceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 17191–17202, 2025. 4, 5, 13

    2025

  23. [23]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 4401–4410, 2019. 2

    2019

  24. [24]

    Finedance: A fine-grained choreography dataset for 3d full body dance generation

    Ronghui Li, Junfan Zhao, Yachao Zhang, Mingyang Su, Zeping Ren, Han Zhang, Yansong Tang, and Xiu Li. Finedance: A fine-grained choreography dataset for 3d full body dance generation. InProceedings of the IEEE/CVF In- ternational Conference on Computer Vision, pages 10234– 10243, 2023. 2

    2023

  25. [25]

    Lodge: A coarse to fine diffusion network for long dance generation guided by the characteristic dance primitives

    Ronghui Li, YuXiang Zhang, Yachao Zhang, Hongwen Zhang, Jie Guo, Yan Zhang, Yebin Liu, and Xiu Li. Lodge: A coarse to fine diffusion network for long dance generation guided by the characteristic dance primitives. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1524–1534, 2024

    2024

  26. [26]

    Lodge++: High-quality and long dance genera- tion with robust choreography patterns.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    Ronghui Li, Hongwen Zhang, Yachao Zhang, Yuxiang Zhang, Youliang Zhang, Jie Guo, Yan Zhang, Xiu Li, and Yebin Liu. Lodge++: High-quality and long dance genera- tion with robust choreography patterns.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025. 2

    2025

  27. [27]

    Zero-1-to-3: Zero-shot one image to 3d object, 2023

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tok- makov, Sergey Zakharov, and Carl V ondrick. Zero-1-to-3: Zero-shot one image to 3d object, 2023. 2

    2023

  28. [28]

    SyncDreamer: Generating Multiview-consistent Images from a Single-view Image

    Yuan Liu, Cheng Lin, Zijiao Zeng, Xiaoxiao Long, Lingjie Liu, Taku Komura, and Wenping Wang. Syncdreamer: Gen- erating multiview-consistent images from a single-view im- age.arXiv preprint arXiv:2309.03453, 2023. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  29. [29]

    Repaint: Inpainting using denoising diffusion probabilistic models

    Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Repaint: Inpainting using denoising diffusion probabilistic models. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11461–11471, 2022. 2

    2022

  30. [30]

    Follow your pose: Pose- guided text-to-video generation using pose-free videos

    Yue Ma, Yingqing He, Xiaodong Cun, Xintao Wang, Siran Chen, Xiu Li, and Qifeng Chen. Follow your pose: Pose- guided text-to-video generation using pose-free videos. In Proceedings of the AAAI Conference on Artificial Intelli- gence, pages 4117–4125, 2024. 2

    2024

  31. [31]

    Follow-your-emoji: Fine-controllable and expressive freestyle portrait animation

    Yue Ma, Hongyu Liu, Hongfa Wang, Heng Pan, Yingqing He, Junkun Yuan, Ailing Zeng, Chengfei Cai, Heung-Yeung Shum, Wei Liu, et al. Follow-your-emoji: Fine-controllable and expressive freestyle portrait animation. InSIGGRAPH Asia 2024 Conference Papers, pages 1–12, 2024. 2

    2024

  32. [32]

    Magic- stick: Controllable video editing via control handle transfor- mations

    Yue Ma, Xiaodong Cun, Sen Liang, Jinbo Xing, Yingqing He, Chenyang Qi, Siran Chen, and Qifeng Chen. Magic- stick: Controllable video editing via control handle transfor- mations. In2025 IEEE/CVF Winter Conference on Applica- tions of Computer Vision (WACV), pages 9385–9395. IEEE,

  33. [33]

    Controllable video generation: A survey.arXiv preprint arXiv:2507.16869,

    Yue Ma, Kunyu Feng, Zhongyuan Hu, Xinyu Wang, Yucheng Wang, Mingzhe Zheng, Xuanhua He, Chenyang Zhu, Hongyu Liu, Yingqing He, et al. Controllable video generation: A survey.arXiv preprint arXiv:2507.16869,

    work page arXiv

  34. [34]

    Follow-your-creation: Empowering 4d creation through video inpainting.arXiv preprint arXiv:2506.04590, 2025

    Yue Ma, Kunyu Feng, Xinhua Zhang, Hongyu Liu, David Junhao Zhang, Jinbo Xing, Yinhan Zhang, Ayden Yang, Zeyu Wang, and Qifeng Chen. Follow-your-creation: Empowering 4d creation through video inpainting.arXiv preprint arXiv:2506.04590, 2025

    work page arXiv 2025

  35. [35]

    Follow-your-click: Open-domain regional image animation via motion prompts

    Yue Ma, Yingqing He, Hongfa Wang, Andong Wang, Leqi Shen, Chenyang Qi, Jixuan Ying, Chengfei Cai, Zhifeng Li, Heung-Yeung Shum, et al. Follow-your-click: Open-domain regional image animation via motion prompts. InProceed- ings of the AAAI Conference on Artificial Intelligence, pages 6018–6026, 2025

    2025

  36. [36]

    Follow-your-motion: Video motion transfer via efficient spatial-temporal decoupled finetuning.arXiv preprint arXiv:2506.05207, 2025

    Yue Ma, Yulong Liu, Qiyuan Zhu, Ayden Yang, Kunyu Feng, Xinhua Zhang, Zhifeng Li, Sirui Han, Chenyang Qi, and Qifeng Chen. Follow-your-motion: Video motion transfer via efficient spatial-temporal decoupled finetuning.arXiv preprint arXiv:2506.05207, 2025. 10

    work page arXiv 2025

  37. [37]

    Follow-your-emoji-faster: To- wards efficient, fine-controllable, and expressive freestyle portrait animation.arXiv preprint arXiv:2509.16630, 2025

    Yue Ma, Zexuan Yan, Hongyu Liu, Hongfa Wang, Heng Pan, Yingqing He, Junkun Yuan, Ailing Zeng, Chengfei Cai, Heung-Yeung Shum, et al. Follow-your-emoji-faster: To- wards efficient, fine-controllable, and expressive freestyle portrait animation.arXiv preprint arXiv:2509.16630, 2025

    work page arXiv 2025

  38. [38]

    Group editing: Edit multiple im- ages in one go.arXiv preprint arXiv:2603.22883, 2026

    Yue Ma, Xinyu Wang, Qianli Ma, Qinghe Wang, Mingzhe Zheng, Xiangpeng Yang, Hao Li, Chongbo Zhao, Jixuan Ying, Harry Yang, et al. Group editing: Edit multiple im- ages in one go.arXiv preprint arXiv:2603.22883, 2026

    work page arXiv 2026

  39. [39]

    Fastvmt: Eliminat- ing redundancy in video motion transfer.arXiv preprint arXiv:2602.05551, 2026

    Yue Ma, Zhikai Wang, Tianhao Ren, Mingzhe Zheng, Hongyu Liu, Jiayi Guo, Mark Fong, Yuxuan Xue, Zixi- ang Zhao, Konrad Schindler, et al. Fastvmt: Eliminat- ing redundancy in video motion transfer.arXiv preprint arXiv:2602.05551, 2026. 2

    work page arXiv 2026

  40. [40]

    SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations

    Chenlin Meng, Yutong He, Yang Song, Jiaming Song, Jia- jun Wu, Jun-Yan Zhu, and Stefano Ermon. Sdedit: Guided image synthesis and editing with stochastic differential equa- tions.arXiv preprint arXiv:2108.01073, 2021. 3, 4

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  41. [41]

    Object 3dit: Language-guided 3d-aware image editing.Advances in Neural Information Processing Systems, 36:3497–3516,

    Oscar Michel, Anand Bhattad, Eli VanderBilt, Ranjay Kr- ishna, Aniruddha Kembhavi, and Tanmay Gupta. Object 3dit: Language-guided 3d-aware image editing.Advances in Neural Information Processing Systems, 36:3497–3516,

  42. [42]

    Accelerating autoregressive video diffusion via history-guided cache and residual correc- tion

    Kepan Nan, Wangbo Zhao, Penghao Zhou, Jun Li, Zhenheng Yang, Jian Yang, and Ying Tai. Accelerating autoregressive video diffusion via history-guided cache and residual correc- tion. InCVPR, pages 43740–43750, 2026. 2

    2026

  43. [43]

    Drag your gan: Interactive point-based manipulation on the generative image manifold

    Xingang Pan, Ayush Tewari, Thomas Leimk ¨uhler, Lingjie Liu, Abhimitra Meka, and Christian Theobalt. Drag your gan: Interactive point-based manipulation on the generative image manifold. InACM SIGGRAPH 2023 conference pro- ceedings, pages 1–11, 2023. 2

    2023

  44. [44]

    Diffusion handles enabling 3d edits for diffusion models by lifting ac- tivations to 3d

    Karran Pandey, Paul Guerrero, Matheus Gadelha, Yannick Hold-Geoffroy, Karan Singh, and Niloy J Mitra. Diffusion handles enabling 3d edits for diffusion models by lifting ac- tivations to 3d. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7695– 7704, 2024. 3

    2024

  45. [45]

    Styleclip: Text-driven manipulation of stylegan imagery

    Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. Styleclip: Text-driven manipulation of stylegan imagery. InProceedings of the IEEE/CVF inter- national conference on computer vision, pages 2085–2094,

    2085

  46. [46]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 4195–4205,

  47. [47]

    Fatezero: Fus- ing attentions for zero-shot text-based video editing

    Chenyang Qi, Xiaodong Cun, Yong Zhang, Chenyang Lei, Xintao Wang, Ying Shan, and Qifeng Chen. Fatezero: Fus- ing attentions for zero-shot text-based video editing. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 15932–15942, 2023. 2

    2023

  48. [48]

    U2-net: Go- ing deeper with nested u-structure for salient object detec- tion.Pattern recognition, 106:107404, 2020

    Xuebin Qin, Zichen Zhang, Chenyang Huang, Masood De- hghan, Osmar R Zaiane, and Martin Jagersand. U2-net: Go- ing deeper with nested u-structure for salient object detec- tion.Pattern recognition, 106:107404, 2020. 13

    2020

  49. [49]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. InInternational conference on machine learning, pages 8748–8763. PmLR, 2021. 6

    2021

  50. [50]

    Vi- sion transformers for dense prediction

    Ren ´e Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vi- sion transformers for dense prediction. InProceedings of the IEEE/CVF international conference on computer vision, pages 12179–12188, 2021. 2

    2021

  51. [51]

    SAM 2: Segment Anything in Images and Videos

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chloe Rolland, Laura Gustafson, et al. Sam 2: Segment anything in images and videos.arXiv preprint arXiv:2408.00714, 2024. 13

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  52. [52]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 2, 3

    2022

  53. [53]

    Ctrl&shift: High-quality geometry-aware object manipula- tion in visual generation.arXiv preprint arXiv:2602.11440,

    Penghui Ruan, Bojia Zi, Xianbiao Qi, Youze Huang, Rong Xiao, Pichao Wang, Jiannong Cao, and Yuhui Shi. Ctrl&shift: High-quality geometry-aware object manipula- tion in visual generation.arXiv preprint arXiv:2602.11440,

    work page arXiv

  54. [54]

    Geodiffuser: Geometry-based image editing with diffusion models

    Rahul Sajnani, Jeroen Vanbaar, Jie Min, Kapil D Katyal, and Srinath Sridhar. Geodiffuser: Geometry-based image editing with diffusion models. InProceedings of the Winter Confer- ence on Applications of Computer Vision, pages 472–482,

  55. [55]

    In- terpreting the latent space of gans for semantic face editing

    Yujun Shen, Jinjin Gu, Xiaoou Tang, and Bolei Zhou. In- terpreting the latent space of gans for semantic face editing. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 9243–9252, 2020. 2

    2020

  56. [56]

    Follow-your-preference: Towards preference- aligned image inpainting.arXiv preprint arXiv:2509.23082,

    Yutao Shen, Junkun Yuan, Toru Aonishi, Hideki Nakayama, and Yue Ma. Follow-your-preference: Towards preference- aligned image inpainting.arXiv preprint arXiv:2509.23082,

    work page arXiv

  57. [57]

    Dragdiffusion: Harnessing diffusion models for interactive point-based image editing

    Yujun Shi, Chuhui Xue, Jun Hao Liew, Jiachun Pan, Han- shu Yan, Wenqing Zhang, Vincent YF Tan, and Song Bai. Dragdiffusion: Harnessing diffusion models for interactive point-based image editing. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8839–8849, 2024. 2

    2024

  58. [58]

    Time-to-move: Training-free motion controlled video generation via dual-clock denoising, 2025

    Assaf Singer, Noam Rotstein, Amir Mann, Ron Kimmel, and Or Litany. Time-to-move: Training-free motion controlled video generation via dual-clock denoising.arXiv preprint arXiv:2511.08633, 2025. 2, 3, 5

    work page arXiv 2025

  59. [59]

    Pro- cesspainter: Learning to draw from sequence data

    Yiren Song, Shijie Huang, Chen Yao, Hai Ci, Xiaojun Ye, Jiaming Liu, Yuxuan Zhang, and Mike Zheng Shou. Pro- cesspainter: Learning to draw from sequence data. InSIG- GRAPH Asia 2024 Conference Papers, pages 1–10, 2024. 2

    2024

  60. [60]

    StreamingEffect: Real-Time Human-Centric Video Effect Generation

    Yiren Song, Cheng Liu, Yuxin Jiang, and Mike Zheng Shou. Streamingeffect: Real-time human-centric video effect gen- eration.arXiv preprint arXiv:2605.17019, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  61. [61]

    VISTA: Triplet-Supervised Video Style Transfer with Diffusion Transformers

    Yiren Song, Wangzi Yao, Haofan Wang, and Mike Zheng Shou. Vista: Triplet-supervised video style transfer with dif- fusion transformers.arXiv preprint arXiv:2605.17312, 2026. 2 11

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  62. [63]

    Gemini: A Family of Highly Capable Multimodal Models

    Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean- Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, Katie Millican, et al. Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023. 6

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  63. [64]

    An image inpainting technique based on the fast marching method.Journal of graphics tools, 9(1): 23–34, 2004

    Alexandru Telea. An image inpainting technique based on the fast marching method.Journal of graphics tools, 9(1): 23–34, 2004. 4, 13

    2004

  64. [65]

    Sv3d: Novel multi-view syn- thesis and 3d generation from a single image using latent video diffusion

    Vikram V oleti, Chun-Han Yao, Mark Boss, Adam Letts, David Pankratz, Dmitry Tochilkin, Christian Laforte, Robin Rombach, and Varun Jampani. Sv3d: Novel multi-view syn- thesis and 3d generation from a single image using latent video diffusion. InEuropean Conference on Computer Vi- sion, pages 439–457. Springer, 2024. 3, 5, 13

    2024

  65. [66]

    Wan: Open and Advanced Large-Scale Video Generative Models

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al. Wan: Open and advanced large-scale video gen- erative models.arXiv preprint arXiv:2503.20314, 2025. 2, 4, 5, 13

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  66. [67]

    Cove: Unleashing the diffusion fea- ture correspondence for consistent video editing.Advances in Neural Information Processing Systems, 37:96541–96565,

    Jiangshan Wang, Yue Ma, Jiayi Guo, Yicheng Xiao, Gao Huang, and Xiu Li. Cove: Unleashing the diffusion fea- ture correspondence for consistent video editing.Advances in Neural Information Processing Systems, 37:96541–96565,

  67. [68]

    arXiv preprint arXiv:2411.04746 (2024)

    Jiangshan Wang, Junfu Pu, Zhongang Qi, Jiayi Guo, Yue Ma, Nisha Huang, Yuxin Chen, Xiu Li, and Ying Shan. Tam- ing rectified flow for inversion and editing.arXiv preprint arXiv:2411.04746, 2024. 2

    work page arXiv 2024

  68. [69]

    Vggt: Vi- sual geometry grounded transformer

    Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny. Vggt: Vi- sual geometry grounded transformer. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 5294–5306, 2025. 3, 5, 13

    2025

  69. [70]

    Synsin: End-to-end view synthesis from a sin- gle image

    Olivia Wiles, Georgia Gkioxari, Richard Szeliski, and Justin Johnson. Synsin: End-to-end view synthesis from a sin- gle image. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7467–7477,

  70. [71]

    Qwen-Image Technical Report

    Chenfei Wu, Jiahao Li, Jingren Zhou, Junyang Lin, Kaiyuan Gao, Kun Yan, Sheng-ming Yin, Shuai Bai, Xiao Xu, Yilei Chen, et al. Qwen-image technical report.arXiv preprint arXiv:2508.02324, 2025. 6, 7, 14

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  71. [72]

    Holistically-nested edge de- tection, 2015

    Saining Xie and Zhuowen Tu. Holistically-nested edge de- tection, 2015. 2

    2015

  72. [73]

    Smrabooth: Subject and motion representation alignment for customized video generation

    Xuancheng Xu, Yaning Li, Sisi You, and Bing-Kun Bao. Smrabooth: Subject and motion representation alignment for customized video generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16130–16141, 2026. 2

    2026

  73. [74]

    Contranerf: Gen- eralizable neural radiance fields for synthetic-to-real novel view synthesis via contrastive learning

    Hao Yang, Lanqing Hong, Aoxue Li, Tianyang Hu, Zhen- guo Li, Gim Hee Lee, and Liwei Wang. Contranerf: Gen- eralizable neural radiance fields for synthetic-to-real novel view synthesis via contrastive learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16508–16517, 2023. 2

    2023

  74. [75]

    VideoCoF: Unified Video Editing with Temporal Reasoner

    Xiangpeng Yang, Ji Xie, Yiyuan Yang, Yan Huang, Min Xu, and Qiang Wu. Unified video editing with temporal reasoner. arXiv preprint arXiv:2512.07469, 2025. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  75. [76]

    Unic: Unified in-context video editing.ICLR 2026, 2025

    Zixuan Ye, Xuanhua He, Quande Liu, Qiulin Wang, Xintao Wang, Pengfei Wan, Di Zhang, Kun Gai, Qifeng Chen, and Wenhan Luo. Unic: Unified in-context video editing.ICLR 2026, 2025. 2

    2026

  76. [77]

    Image sculpting: Precise ob- ject editing with 3d geometry control

    Jiraphon Yenphraphai, Xichen Pan, Sainan Liu, Daniele Panozzo, and Saining Xie. Image sculpting: Precise ob- ject editing with 3d geometry control. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4241–4251, 2024. 6, 7, 14

    2024

  77. [78]

    Objectmover: Gener- ative object movement with video prior

    Xin Yu, Tianyu Wang, Soo Ye Kim, Paul Guerrero, Xi Chen, Qing Liu, Zhe Lin, and Xiaojuan Qi. Objectmover: Gener- ative object movement with video prior. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17682–17691, 2025. 3

    2025

  78. [79]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023. 2, 5

    2023

  79. [80]

    3ditscene: Editing any scene via language-guided disen- tangled gaussian splatting

    Qihang Zhang, Yinghao Xu, Chaoyang Wang, Hsin-Ying Lee, Gordon Wetzstein, Bolei Zhou, and Ceyuan Yang. 3ditscene: Editing any scene via language-guided disen- tangled gaussian splatting. InInternational Conference on Learning Representations, pages 2760–2775, 2025. 3

    2025

  80. [81]

    Objectadd: adding objects into image via a training-free diffusion modification fashion.Pattern Recog- nition, page 112807, 2025

    Ziyue Zhang, Mingbao Lin, Quanjian Song, Yuxin Zhang, and Rongrong Ji. Objectadd: adding objects into image via a training-free diffusion modification fashion.Pattern Recog- nition, page 112807, 2025. 2

    2025

Showing first 80 references.