pith. sign in

arxiv: 2606.19495 · v1 · pith:J4GHEOQUnew · submitted 2026-06-17 · 💻 cs.CV

LooseControlVideo: Directorial Video Control using Spatial Blocking

Shariq Farooq Bhat , Niloy J. Mitra , Kalyan Sunkavalli This is my paper

Pith reviewed 2026-06-26 21:03 UTC · model grok-4.3

classification 💻 cs.CV
keywords text-to-video generation3D spatial controloriented bounding boxesocclusion handlingmulti-object video authoringlayout conditioningtrajectory control
0
0 comments X

The pith

Sparse oriented 3D boxes let video models infer realistic occlusions and dynamics from high-level trajectories.

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

The paper introduces LooseControlVideo to control text-to-video generation with sparse oriented 3D boxes rather than dense per-frame signals. Users specify layouts and paths at a high level while the model fills in occlusions, motion, and object interactions. This is done by fine-tuning a backbone on video data annotated with a new encoding that records size, orientation, and depth order. Tests on nuScenes, HO-3D, and BEHAVE show clear gains over 2D-box and flow baselines in trajectory error, rigid-motion consistency, and occlusion accuracy. The work therefore treats oriented 3D primitives as a lightweight geometric prior that simplifies multi-agent scene authoring.

Core claim

Oriented 3D boxes function as an effective blocking proxy: after fine-tuning on DNOCS-annotated videos, the model generates plausible occlusions, dynamics, and interactions directly from sparse 3D size, orientation, and depth-order inputs without requiring dense guidance.

What carries the argument

DNOCS encoding of 3D size, orientation and depth-ordered occlusions, applied as annotation for fine-tuning the generative backbone so that sparse boxes suffice as control signals.

If this is right

  • Users can author multi-object trajectories and layouts with far less manual effort than dense depth or flow maps require.
  • Small local edits to a single object's path or contact can be applied while the rest of the scene stays coherent.
  • The same sparse-box interface yields measurable gains in trajectory accuracy, rigid-motion consistency, and occlusion correctness on the reported benchmarks.

Where Pith is reading between the lines

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

  • The approach could be tested on longer clips or scenes with deformable objects to check whether the geometric prior continues to suffice.
  • Similar sparse 3D annotations might be applied to other video backbones to see if the control benefit transfers.
  • The method suggests a route toward directorial tools that combine 3D blocking with natural-language instructions for hybrid authoring.

Load-bearing premise

Fine-tuning on videos labeled with the new 3D encoding will enable the model to produce realistic occlusions and interactions when given only sparse oriented boxes.

What would settle it

Generate videos from 3D-box sequences whose occlusion patterns or interaction timings fall outside the annotated training distribution and check whether depth ordering or contact events remain correct.

Figures

Figures reproduced from arXiv: 2606.19495 by Kalyan Sunkavalli, Niloy J. Mitra, Shariq Farooq Bhat.

Figure 1
Figure 1. Figure 1: Cinematic video authoring and editing via oriented 3D blocking. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the 3D control space and virtual rendering setup. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Video generation results. Insets visualize the input 3D oriented box se￾quences used as conditioning proxies. Top: High-speed weaving and maneuvering. The oriented boxes guide a vehicle through narrow gaps between trucks, capturing subtle 6-DOF rotations and triggering responsive effects like brake light activation. Middle: Robust occlusion and shadow consistency. A cat maintains its identity and tempo￾ral… view at source ↗
Figure 4
Figure 4. Figure 4: Video motion editing via oriented 3D proxy manipulation. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison. We compare our DNOCs-based oriented box con￾trol against several alternatives, including 2D bounding boxes, 3D box depth, mesh depth, and 2D optical flow. While 2D-centric methods struggle with viewpoint￾consistent orientation and temporal grounding, and dense depth/mesh guidance can over-constrain natural dynamics, our method (bottom) excels at preserving precise 6- DOF choreograph… view at source ↗
read the original abstract

Precise 3D spatial orchestration in text-to-video generation remains a significant challenge, particularly for multi-object scenes where semantic layout and temporal dynamics are often entangled. While existing depth-conditioned models achieve good structural fidelity, they necessitate dense, frame-accurate guidance that is labor-intensive to author for dynamic events involving deformable objects. We present LooseControlVideo, a framework that enables intuitive and expressive control by using sparse, oriented 3D boxes as a "blocking" proxy. This allows users to author high-level layout and trajectory while leveraging a video generative model to generate realistic occlusions, dynamics and interactions. We achieve this by fine-tuning a Wan 2.2 backbone on a video dataset annotated with DNOCS, a novel encoding for 3D size, orientation and depth-ordered occlusions. Furthermore, our method allows for localized refinement, such as adjusting a jump trajectory or adding an interaction, with minimal disruption to the global scene context. Extensive evaluations on the nuScenes, HO-3D, and BEHAVE benchmarks demonstrate that LooseControlVideo significantly outperforms existing 2D-box and flow-based baselines. Our findings indicate a 1.2x to 3x improvement in Trajectory Error; 2x improvement in Rigid Motion Consistency; and a 1.5x to 2x increase in Occlusion Accuracy over current state-of-the-art layout-conditioned models, demonstrating that oriented 3D primitives provide good geometric prior for complex, multi-agent video authoring.

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 introduces LooseControlVideo, a framework for text-to-video generation that uses sparse oriented 3D boxes as a blocking proxy for high-level layout and trajectory control. It fine-tunes a Wan 2.2 backbone on video data annotated with a novel DNOCS encoding (for 3D size, orientation, and depth-ordered occlusions) to generate realistic dynamics, occlusions, and interactions without dense guidance. The central claim is that this yields 1.2x–3x gains in Trajectory Error, 2x in Rigid Motion Consistency, and 1.5x–2x in Occlusion Accuracy over 2D-box and flow-based baselines on the nuScenes, HO-3D, and BEHAVE benchmarks.

Significance. If the empirical claims hold after proper verification of baselines and ablations, the work would demonstrate that oriented 3D primitives supply a useful geometric prior for multi-agent video authoring, reducing the authoring burden relative to dense depth conditioning while supporting localized edits.

major comments (2)
  1. [Abstract] Abstract: the central empirical claim of outperformance (1.2x–3x Trajectory Error, etc.) is stated without any description of baseline implementations, training details, statistical significance, or ablation studies on the DNOCS encoding; this renders the quantitative results unverifiable from the provided text.
  2. [Abstract] Abstract: the DNOCS encoding is introduced as novel but never defined or formalized; without its explicit construction or how it encodes depth-ordered occlusions, it is impossible to assess whether the fine-tuning step actually enables inference from sparse inputs as claimed.
minor comments (1)
  1. [Abstract] The abstract refers to 'Wan 2.2 backbone' and 'DNOCS' without prior definition or citation, which hinders immediate readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed feedback on the abstract. We agree that the abstract can be made more self-contained to improve verifiability of the claims and to provide a high-level definition of DNOCS. We will revise the abstract accordingly while ensuring it remains concise; detailed explanations remain in the main text and supplementary material.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claim of outperformance (1.2x–3x Trajectory Error, etc.) is stated without any description of baseline implementations, training details, statistical significance, or ablation studies on the DNOCS encoding; this renders the quantitative results unverifiable from the provided text.

    Authors: We acknowledge the abstract's conciseness limits immediate verifiability. The main manuscript (Sections 4.1–4.3 and 5) specifies the baselines as 2D-box methods adapted from prior layout-conditioned video models and flow-based approaches, with training on the Wan 2.2 backbone using the DNOCS-annotated dataset; statistical significance is assessed via multiple random seeds and reported with standard deviations. Ablations on the DNOCS components appear in Section 5.2. To address the concern, we will add one sentence to the abstract briefly naming the baseline categories and noting that full implementation and ablation details are in the experiments section. revision: yes

  2. Referee: [Abstract] Abstract: the DNOCS encoding is introduced as novel but never defined or formalized; without its explicit construction or how it encodes depth-ordered occlusions, it is impossible to assess whether the fine-tuning step actually enables inference from sparse inputs as claimed.

    Authors: The abstract provides a brief description ('a novel encoding for 3D size, orientation and depth-ordered occlusions'), but we agree a more explicit high-level formalization would help. Section 3.1 of the manuscript defines DNOCS as a per-frame representation that augments oriented 3D bounding boxes with explicit depth ordering to encode occlusions. We will revise the abstract to include a short clause such as 'DNOCS, which parameterizes oriented 3D boxes with size, rotation, and depth-ordered occlusion masks' to clarify its role in enabling sparse control during fine-tuning. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper's central claims rest on empirical benchmark results (nuScenes, HO-3D, BEHAVE) comparing trajectory error, motion consistency, and occlusion accuracy against external baselines after fine-tuning on DNOCS-annotated data. No equations, parameter fits presented as predictions, self-citation load-bearing premises, or ansatz smuggling appear in the abstract or described derivation. The method is self-contained via standard training and evaluation procedures without internal reduction to its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The approach rests on the unverified assumption that the chosen backbone plus DNOCS fine-tuning transfers geometric priors to realistic video synthesis; no free parameters or additional axioms are stated in the abstract.

invented entities (1)
  • DNOCS encoding no independent evidence
    purpose: Novel encoding for 3D size, orientation and depth-ordered occlusions used as training signal
    Introduced in the paper as the annotation format enabling the method; no independent evidence provided in abstract.

pith-pipeline@v0.9.1-grok · 5801 in / 1171 out tokens · 33057 ms · 2026-06-26T21:03:06.639330+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

49 extracted references · 9 canonical work pages · 5 internal anchors

  1. [1]

    Bhat, S.F., Mitra, N., Wonka, P.: Loosecontrol: Lifting controlnet for generalized depthconditioning.In:ACMSIGGRAPH2024ConferencePapers.pp.1–11(2024) 3, 4

    2024

  2. [2]

    In: CVPR (2022) 3, 10, 13, 14

    Bhatnagar, B.L., Xie, X., Petrov, I., Sminchisescu, C., Theobalt, C., Pons-Moll, G.: BEHAVE: dataset and method for tracking human object interactions. In: CVPR (2022) 3, 10, 13, 14

    2022

  3. [3]

    In: Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Brack, M., Friedrich, F., Kornmeier, K., Tsaban, L., Schramowski, P., Kersting, K., Passos, A.: Ledits++: Limitless image editing using text-to-image models. In: Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8861–8870 (2024) 3

    2024

  4. [4]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Brooks, T., Holynski, A., Efros, A.A.: Instructpix2pix: Learning to follow image editing instructions. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 18392–18402 (2023) 3

    2023

  5. [5]

    OpenAI Blog (2024) 1, 4

    Brooks, T., Peebles, B., Holmes, C., DePue, W., Guo, Y., Jing, L., Schnurr, D., Lecomte, J., Sukhum, A., Senpuru, D., et al.: Video generation models as world simulators. OpenAI Blog (2024) 1, 4

    2024

  6. [6]

    In: CVPR (2020) 3, 9, 12, 14

    Caesar, H., Bankiti, V., Lang, A.H., Vora, S., Liong, V.E., Xu, Q., Krishnan, A., Pan, Y., Baldan, G., Beijbom, O.: nuscenes: A multimodal dataset for autonomous driving. In: CVPR (2020) 3, 9, 12, 14

    2020

  7. [7]

    In: CVPR (2025) 8, 12

    Chen, S., Guo, H., Zhu, S., Zhang, F., Huang, Z., Feng, J., Kang, B.: Video depth anything: Consistent depth estimation for super-long videos. In: CVPR (2025) 8, 12

    2025

  8. [8]

    Google Tech- nical Report (2024) 1, 4

    DeepMind, G.: Veo: Google’s most capable generative video model. Google Tech- nical Report (2024) 1, 4

    2024

  9. [9]

    Gu, Z., Yan, R., Lu, J., Li, P., Dou, Z., Si, C., Dong, Z., Liu, Q., Lin, C., Liu, Z., Wang, W., Liu, Y.: Diffusion as shader: 3d-aware video diffusion for versatile video generation control (2025) 4

    2025

  10. [10]

    In: CVPR (2020) 3, 10, 13

    Hampali, S., Rad, M., Oberweger, M., Lepetit, V.: Honnotate: A method for 3d annotation of hand and object poses. In: CVPR (2020) 3, 10, 13

    2020

  11. [11]

    In: ICLR (2025) 4

    He, H., Xu, Y., Guo, Y., Wetzstein, G., Dai, B., Li, H., Yang, C.: Cameractrl: Enabling camera control for video diffusion models. In: ICLR (2025) 4

    2025

  12. [12]

    Advances in neural information processing systems33, 6840–6851 (2020) 3

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020) 3

    2020

  13. [13]

    Ho, J., Salimans, T., Gritsenko, A., Chan, W., Norouzi, M., Fleet, D.J.: Video diffusion models. pp. 8633–8646 (2022) 4

    2022

  14. [14]

    In: CVPR (2024) 12, 14

    Huang, Z., He, Y., Yu, J., Zhang, F., Si, C., Jiang, Y., Zhang, Y., Wu, T., Jin, Q., Chanpaisit, N., Wang, Y., Chen, X., Wang, L., Lin, D., Qiao, Y., Liu, Z.: VBench: Comprehensive benchmark suite for video generative models. In: CVPR (2024) 12, 14

    2024

  15. [15]

    In: CVPR (2025) 5

    Jeong, H., Huang, C.H.P., Ye, J.C., Mitra, N., Ceylan, D.: Track4gen: Teaching video diffusion models to track points improves video generation. In: CVPR (2025) 5

    2025

  16. [16]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Jiang, Z., Han, Z., Mao, C., Zhang, J., Pan, Y., Liu, Y.: Vace: All-in-one video creation and editing. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 17191–17202 (2025) 8

    2025

  17. [17]

    In: CVPR (2026) 4 16 S

    Kizil, M.B., Sanli, E., Mitra, N.J., Erdem, E., Erdem, A., Ceylan, D.: Lamp: Language-assisted motion planning for controllable video generation. In: CVPR (2026) 4 16 S. F. Bhat et al

    2026

  18. [18]

    In: CVPR (2026) 4

    Lee, Y.C., Zhang, Z., Huang, J., Wang, J.H., Lee, J.Y., Huang, J.B., Shechtman, E., Li, Z.: Generative video motion editing with 3d point tracks. In: CVPR (2026) 4

    2026

  19. [19]

    In: European Conference on Computer Vision

    Li, R., Zheng, C., Rupprecht, C., Vedaldi, A.: Dragapart: Learning a part-level motion prior for articulated objects. In: European Conference on Computer Vision. pp. 165–183. Springer (2024) 3

    2024

  20. [20]

    ACM TOG30(4), 52:1–52:12 (2011) 8

    Li, Y., Wu, X., Chrysanthou, Y., Sharf, A., Cohen-Or, D., Mitra, N.J.: Globfit: Consistently fitting primitives by discovering global relations. ACM TOG30(4), 52:1–52:12 (2011) 8

    2011

  21. [21]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li,Y.,Liu,H.,Wu,Q.,Mu,F.,Yang,J.,Gao,J.,Li,C.,Lee,Y.J.:Gligen:Open-set grounded text-to-image generation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22511–22521 (2023) 3, 4

    2023

  22. [22]

    In: COLM (2024) 4

    Lin, H., Zala, A., Cho, J., Bansal, M.: Videodirectorgpt: Consistent multi-scene video generation via llm-guided planning. In: COLM (2024) 4

    2024

  23. [23]

    In: ECCV (2023) 8

    Liu,S.,Zuo,Z.,Hou,J.,Peng,H.,Li,H.,Hui,J.,Huang,J.,Li,F.,Zhang,L.,etal.: Grounding dino: Marrying dino with grounded pre-training for open-vocabulary object detection. In: ECCV (2023) 8

    2023

  24. [24]

    Luo, G.Y., Luo, Z.H., Gosselin, A., Jolicoeur-Martineau, A., Pal, C.: Ctrl-v: Higher fidelity video generation with bounding-box controlled object motion (2024), https://arxiv.org/abs/2406.056304

    work page arXiv 2024

  25. [25]

    T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models

    Mou, C., Wang, X., Xie, L., Zhang, J., Qi, Z., Shan, Y., Qie, X.: T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. arXiv preprint arXiv:2302.08453 (2023) 3

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  26. [26]

    Peebles, W., Xie, S.: Scalable diffusion models with transformers (2023) 4

    2023

  27. [27]

    In: CVPR (2016) 2, 10

    Perazzi, F., Pont-Tuset, J., McWilliams, B., Gool, L.V., Gross, M., Sorkine- Hornung, A.: A benchmark dataset and evaluation methodology for video object segmentation. In: CVPR (2016) 2, 10

    2016

  28. [28]

    SAM 2: Segment Anything in Images and Videos

    Ravi, N., Gabeur, V., Hu, Y.T., Hu, R., Ryali, C., Ma, T., Khedr, H., Rädle, R., Rolland, C., Gustafson, L., et al.: Sam 2: Segment anything in images and videos. arXiv preprint arXiv:2408.00714 (2024) 8, 10

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  29. [29]

    Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context

    Reid, M., Savinov, N., Teplyashin, D., Coppin, D., Mumtaz, A., Ma, S., Paduraru, C., Paquet, U., Hayes, P., et al.: Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. arXiv preprint arXiv:2403.05530 (2024) 4

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  30. [30]

    In: CVPR (2025) 4

    Ren,X.,Shen,T.,Huang,J.,Ling,H.,Lu,Y.,Nimier-David,M.,Müller,T.,Keller, A., Fidler, S., Gao, J.: Gen3c: 3d-informed world-consistent video generation with precise camera control. In: CVPR (2025) 4

    2025

  31. [31]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10684–10695 (2022) 3

    2022

  32. [32]

    Saha, O., Krs, V., Mech, R., Maji, S., Blackburn-Matzen, K., Gadelha, M.: Sigma- gen: Structure and identity guided multi-subject assembly for image generation (2025) 14

    2025

  33. [33]

    Advances in Neural Information Processing Systems35, 36479–36494 (2022) 3

    Saharia, C., Chan, W., Saxena, S., Li, L., Whang, J., Denton, E.L., Ghasemipour, K., Gontijo Lopes, R., Karagol Ayan, B., Salimans, T., et al.: Photorealistic text- to-image diffusion models with deep language understanding. Advances in Neural Information Processing Systems35, 36479–36494 (2022) 3

    2022

  34. [34]

    Shi, Y., Xue, C., Liew, J.H., Pan, J., Yan, H., Zhang, W., Tan, V.Y., Bai, S.: Dragdiffusion: Harnessing diffusion models for interactive point-based image edit- ing.In:ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPattern Recognition. pp. 8839–8849 (2024) 3 LCV: Directorial Video Control using Spatial Blocking 17

    2024

  35. [35]

    In: ICLR (2022) 4

    Singer, U., Polyak, A., Hayes, T., Yin, X., An, J., Zhang, S., Hu, Q., Yang, H., Ashual, O., Gafni, O., Parikh, D., Gupta, S., Taigman, Y.: Make-a-video: Text-to- video generation without text-video data. In: ICLR (2022) 4

    2022

  36. [36]

    Team, W.: Wan: Open and high-quality video generation with 3d-aware transform- ers,https://arxiv.org/abs/2503.203141, 4

    work page internal anchor Pith review Pith/arXiv arXiv

  37. [37]

    In: ECCV (2020) 12

    Teed, Z., Deng, J.: Raft: Recurrent all-pairs field transforms for optical flow. In: ECCV (2020) 12

    2020

  38. [38]

    In: ACM SIGGRAPH 2023 Conference Proceedings

    Voynov, A., Aberman, K., Cohen-Or, D.: Sketch-guided text-to-image diffusion models. In: ACM SIGGRAPH 2023 Conference Proceedings. pp. 1–11 (2023) 3

    2023

  39. [39]

    In: CVPR (June 2019) 6

    Wang, H., Sridhar, S., Huang, J., Valentin, J., Song, S., Guibas, L.J.: Normalized object coordinate space for category-level 6d object pose and size estimation. In: CVPR (June 2019) 6

    2019

  40. [40]

    arXiv preprint arXiv:2402.01566 (2024),https://arxiv.org/abs/2402.015664

    Wang, J., Zhang, Y., Zou, J., Zeng, Y., Wei, G., Yuan, L., Li, H.: Boxima- tor: Generating rich and controllable motions for video synthesis. arXiv preprint arXiv:2402.01566 (2024),https://arxiv.org/abs/2402.015664

    work page arXiv 2024

  41. [41]

    arXiv preprint arXiv:2205.12952 (2022) 3

    Wang, T., Zhang, T., Zhang, B., Ouyang, H., Chen, D., Chen, Q., Wen, F.: Pretraining is all you need for image-to-image translation. arXiv preprint arXiv:2205.12952 (2022) 3

    work page arXiv 2022

  42. [42]

    Wang, Z., Yuan, Z., Wang, X., Li, Y., Chen, T., Xia, M., Luo, P., Shan, Y.: Mo- tionctrl: A unified and flexible motion controller for video generation (2023) 4

    2023

  43. [43]

    Yang, S., Hou, L., Huang, H., Ma, C., Wan, P., Zhang, D., Chen, X., Liao, J.: Direct-a-video: Customized video generation with user-directed camera movement and object motion (2024) 4

    2024

  44. [44]

    In: ICLR (2025) 4

    Yang, Z., et al.: Cogvideox: Text-to-video diffusion models with an expert trans- former. In: ICLR (2025) 4

    2025

  45. [45]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Zhang, L., Rao, A., Agrawala, M.: Adding conditional control to text-to-image diffusion models. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 3836–3847 (2023) 3

    2023

  46. [46]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Zhang, L., Rao, A., Agrawala, M.: Adding conditional control to text-to-image diffusion models. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 3836–3847 (2023) 8

    2023

  47. [47]

    Zhao, M., Wang, R., Bao, F., Li, C., Zhu, J.: Controlvideo: Conditional control for one-shot text-driven video editing and beyond (2023),https://arxiv.org/abs/ 2305.170984

    work page arXiv 2023

  48. [48]

    Zheng, Z., Peng, X., Yang, T., Shen, C., Li, S., Liu, H., Zhou, Y., Li, T., You, Y.: Open-sora: Democratizing efficient video production for all (2024),https: //arxiv.org/abs/2412.204044

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  49. [49]

    Zhu, H., He, T., Tang, A., Guo, J., Chen, Z., Bian, J.: Compositional 3d-aware video generation with llm director (2024) 4 18 S. F. Bhat et al. Supplementary Material A User Study T able A1:Overall completed-session pairwise preference matrix. Each cell reports the percentage of votes preferring the row method over the column method (64 votes per method p...

    2024