Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Individual Content and Motion Dynamics Preserved Pruning for Video Diffusion Models

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The layer-role split in video diffusion U-Nets makes targeted block pruning a viable compression route: shallow blocks are pruned, deep blocks preserved, and a two-part consistency loss recovers quality.

desk verdict Solid pruning recipe and useful block analysis, but the headline I2V speedup is mostly bought by VAE decoder compression, not the proposed U-Net pruning. read the letter →

arxiv 2411.18375 v3 pith:YU5RKH3P submitted 2024-11-27 cs.CV eess.IV

Yiming Wu , Zhenghao Chen , Huan Wang , Dong Xu This is my paper · ORCID
classification cs.CVeess.IV
keywords videodiffusionmodelsmodelcompressionnetworkpruningknowledgedistillationadversariallosstext-to-videogenerationimage-to-videomotiondynamics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that video diffusion models can be made much cheaper to run without visibly losing quality, by pruning blocks according to what they actually do. It reports that the deep layers of a video diffusion U-Net carry the cross-frame motion dynamics, while the shallow layers mostly render individual frame content, so the shallow blocks can be removed and the deep blocks kept. The resulting compressed model, VDMini, is then fine-tuned with a two-part loss: one part distills per-frame features from the teacher model, and one adversarial part keeps the whole video's motion coherent. On three existing video generators the method reports 2.5x, 1.4x, and 1.25x inference speedups with roughly unchanged benchmark scores. If these results hold, pruning video diffusion models by layer role is a practical path to cheaper deployment.

What carries the argument

The load-bearing mechanism is the layer-role split plus the two-part distillation objective. The paper measures block importance by replacing each U-Net block with an identity mapping or a channel-matching convolution and scoring the FVD, which identifies deep blocks as motion-critical and shallow blocks as content-critical. The pruning pattern removes the second ResBlock-Attention pair from most down and up blocks, removes the deepest Down-3, Mid, and Up-0 blocks, and prunes the VAE decoder. Fine-tuning then uses the Individual Content and Motion Dynamics (ICMD) loss: the Individual Content Distillation (ICD) loss aligns frame-level intermediate features between teacher and student, and the Multi-frame Content Adversarial (MCA) loss trains a spatio-temporal discriminator to distinguish student video latents from teacher outputs, which the paper credits with preserving motion dynamics.

What would settle it

Run the same per-block FVD replacement study on a different video diffusion U-Net, such as T2V-Turbo-v2 or another image-to-video model: if the low-importance blocks are not the shallow ones, or if removing the supposedly redundant shallow blocks visibly degrades motion coherence, the transfer assumption behind VDMini is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a video diffusion model's U-Net has a clean division of labor: shallow blocks determine the quality of individual frames, while deep blocks determine the coherence of motion across frames. The paper supports this with per-block importance measurements on the image-to-video model SF-V, replacing each block with an identity mapping or a single convolution and measuring the change in Fréchet Video Distance (FVD). Based on that, it prunes the second ResBlock-Attention pairs in most down and up blocks, removes the lowest-resolution down block, the middle block, and the highest-resolution up block, and also compresses the VAE decoder, producing VDMini. A fine-tuning stage with the proposed Individual Content and Motion Dynamics (ICMD) loss, combining individual-content feature distillation with a multi-frame adversarial loss, transfers the teacher's behavior to the student. The paper reports that VDMini matches the teacher's benchmark scores within a few points while running 2.5x, 1.4x, and 1.25x faster on SF-V, T2V-Turbo-v2, and HunyuanVideo respectively.

Load-bearing premise

The load-bearing premise, stated in Appendix B, is that the block-role pattern found in one image-to-video model (SF-V) transfers to other video diffusion models, so the same blocks can be pruned without re-measuring their importance; if that transfer fails, the quality of the text-to-video variants is not guaranteed by the paper's analysis.

Editorial extensions

If this is right

  • The image-to-video model VDMini-I2V cuts U-Net latency from 512 ms to 345 ms per pass and, together with a compressed VAE decoder, reaches about a 2.5x total speedup over SF-V while keeping FVD on UCF101 at 198.13 versus the teacher's 166.26.
  • For the text-to-video model T2V-Turbo-v2, VDMini-T2V lowers U-Net latency from 2554 ms to 1662 ms, about a 1.4x speedup, with a VBench-T2V Total Score drop from 83.52 to 82.14.
  • For the DiT-based HunyuanVideo, reducing double-stream blocks from 20 to 12 and single-stream blocks from 40 to 28 yields a 1.25x speedup with a Total Score drop of 0.82 percentage points.
  • Ablations show that both components of the ICMD loss matter: neither ICD nor MCA alone reaches the combined FVD of 198.13, and pruning without either loss lands at 299.44.
  • The compressed VAE decoder alone delivers a large share of the speedup, cutting decoder latency by about 70% while keeping PSNR, SSIM, and LPIPS close to the original decoder.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct test the paper leaves implicit is to re-run the per-block FVD importance measurement on each new backbone before applying the same pruning pattern, since Appendix B transfers the SF-V pattern to other models on the assumption that block roles are shared.
  • The component-level latency numbers suggest that on many models the compressed VAE decoder may deliver as much wall-clock savings as the U-Net pruning; separating those gains on additional models would show how much of the headline speedup depends specifically on the layer-role insight.
  • A natural next experiment is to stress-test the motion argument at longer durations and higher resolutions, where cross-frame coherence demands increase and the supposedly motion-critical deep blocks could become even more important.
  • If the motion-preserving effect of the temporal discriminator heads generalizes, the MCA loss could be reused as a fine-tuning objective for other compressed video generators without redesigning the pruning pattern.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper proposes VDMini, a pruned video diffusion model obtained by removing selected U-Net blocks (mostly shallow blocks, plus the mid/up blocks) and by adding an Individual Content and Motion Dynamics (ICMD) consistency loss made of an individual-content feature distillation term (ICD) and a multi-frame content adversarial term (MCA). The authors report speedups of 2.5x for the I2V model SF-V, 1.4x for T2V-Turbo-v2, and 1.25x for HunyuanVideo, with quality comparable to the unpruned teachers on UCF101 and VBench benchmarks. They also compress the VAE decoder of SF-V with layer and channel pruning, and they ablate the loss components and compare against magnitude pruning and DepGraph.

Significance. If the claims held, this would be a useful demonstration that structural pruning can accelerate video diffusion models while largely preserving quality, and the paper would be one of the first to do so across both U-Net and DiT backbones. The work has clear strengths: it evaluates on two tasks and three backbones, includes ablations of the loss terms, and compares against standard pruning baselines. However, the headline I2V speedup is mostly purchased by the separately compressed VAE decoder rather than by the proposed U-Net block pruning, and the quality assessment is partly self-referential because the pruning pattern and hyperparameters are selected on the same benchmark used for final evaluation. With corrected latency accounting, the marginal contribution of the proposed pruning and ICMD loss to I2V speed is small. The T2V results still show genuine total speedups, but those claims depend on an unverified transfer assumption about block roles across architectures.

major comments (5)
  1. [§3.2, Table 2, Table A1] The central I2V speedup claim is not supported by the reported latency accounting. Table A1 gives SF-V total latency as 35+75+512+2832 = 3454 ms and VDMini-I2V total latency as 35+75+345+840.5 = 1295.5 ms, which is a 2.67x end-to-end speedup. If the U-Net is pruned but the original VAE decoder is kept, the total is 35+75+345+2832 = 3287 ms, only a 1.05x speedup over SF-V; if only the VAE decoder is compressed and the U-Net is left at 512 ms, the total is 35+75+512+840.5 = 1462.5 ms, a 2.36x speedup. Thus the proposed block pruning and ICMD loss contribute almost none of the advertised I2V speedup, and the abstract and Introduction overstate the contribution of the pruning method. The paper should report full end-to-end latency for every configuration and attribute the speedup components explicitly.
  2. [Table 2] The 37x speedup versus SVD-16 mixes incompatible latency scopes. VDMini-I2V is credited with 345 ms, which is only the U-Net latency from Table A1, while the SVD-16 number of 12880 ms appears to be a whole-model latency. Using the Table A1 components, the end-to-end VDMini-I2V latency is 1295.5 ms, so the speedup over SVD-16 is about 9.9x if SVD-16 is end-to-end, and if SVD-16 is U-Net-only the comparison still needs to be stated and justified. The 37x figure should be removed or recomputed with consistent latency definitions.
  3. [§4.2, §4.3, Table 5] The pruning pattern is selected using FVD scores computed on UCF101 (Section 4.2, 1200 videos), and the loss weights λ_ICD and λ_MCA are tuned on the same FVD metric in Table 5, while the final quality evaluation in Table 2 is also FVD on UCF101. This selection-on-test-set makes the reported quality comparisons optimistic and complicates interpretation of the claims that quality is 'maintained.' The authors should use a validation split for block-importance analysis and hyperparameter selection, or at minimum state that the final numbers are on the same data used to choose the architecture and loss weights.
  4. [Appendix B] The transfer of the pruning pattern to T2V-Turbo-v2 and HunyuanVideo is an assumption, not a demonstrated result. Appendix B says 'it is reasonable to assume that specific blocks within these U-Nets perform similar roles across both tasks,' and the same pattern is applied without per-model importance re-evaluation. This assumption is load-bearing for the VDMini-T2V and VDMini-T2V-HY quality claims. The authors should provide at least a block-importance or FVD-based validation for the T2V U-Net (or for the Hunyuan DiT) to show that the pruned blocks are indeed the least important ones in those models.
  5. [Appendix C] The I2V teacher and all I2V baselines are a re-implementation of SF-V because the official SVD code is unavailable, and Tables 1-3 compare against this re-implementation. This creates a self-referential evaluation: the pruning pattern is derived from the re-implemented teacher and the quality is measured against the same re-implemented teacher. The authors should clarify how the re-implementation matches the official SF-V numbers, and ideally release the teacher and student code so the comparisons can be checked.
minor comments (6)
  1. [Table 2 / References] The reference for SF-V in Table 2 appears to be [56] (StyleGAN-V), but SF-V is reference [72]; please correct the citation.
  2. [Table 4 and Table 5] Table 4's caption and text use 'Integrated Content Adversarial Loss (L_ICA)' while the method is called Multi-frame Content Adversarial (MCA) loss; Table 5's header lists λ_ICA but should be λ_MCA. Please make the notation consistent.
  3. [§4.1] The text says 'We utilize three datasets' but then lists four (OpenVid-1M, VidGen-1M, WebVid-10M, HD-Mixkit); please correct the count.
  4. [Equation (3)] The consistency distillation loss formula appears malformed in the typeset text (the argument list of the distance function is garbled); please fix the equation.
  5. [Abstract] The phrase 'average 2.5x, 1.4x, and 1.25x speed up' is awkward because there is only one number per model; consider rewording.
  6. [§3 heading] The section heading reads 'METHODOLGY' and Section 3.2 contains 'evaluting'; please fix these typographical errors.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild self-referential benchmark loop in the I2V evaluation; the central pruning-plus-distillation derivation is otherwise independent.

  1. fitted input called prediction [Appendix B (block FVD evaluation), Section 4.1 (Evaluation Metrics), Table 2 and Table 5]
    "For fast evaluation, we sample 1200 videos from the UCF101 dataset as the ground-truth videos and calculate the FVD score between the generated and ground-truth videos."

    The pruning pattern is selected by measuring FVD on UCF101, and the ICMD loss weights are tuned against FVD on the same benchmark (Table 5). The paper then reports FVD on UCF101 as the headline quality evidence for VDMini-I2V (Section 4.1, Table 2). Thus the architecture and hyper-parameters are chosen using the same metric and benchmark used for the final quality claim, making the I2V quality result partly a model-selection artifact rather than an independent test. This is not a full derivation collapse, because the speedups are latency arithmetic and the T2V branches are evaluated on VBench, but the I2V quality comparison is partially self-referential.

full rationale

The central derivation of VDMini is not circular: block pruning is decided by an empirical FVD analysis, the ICMD loss is a feature-matching plus adversarial objective, and the speedup numbers are direct latency measurements. The T2V and Hunyuan branches are evaluated on VBench, an external benchmark, which gives independent support. The one genuine self-referential element is the I2V loop: the pruning pattern is chosen using FVD on UCF101 (Appendix B) and the loss weights are tuned against FVD (Table 5), while the final quality claim for VDMini-I2V is the FVD on the same UCF101 set. That makes the I2V quality comparison partly self-referential but not a derivation that is equivalent to its inputs by construction. Two non-circular correctness risks should also be noted: the SF-V teacher is a re-implementation created for this paper (Appendix C), and Table 2 compares the student U-Net-only latency of 345 ms against a total-latency number for SVD-16 of 12880 ms, whereas the end-to-end latency from Table A1 gives about a 9.9x speedup over SVD-16, not 37x. These affect the strength of the claims but do not constitute circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on an empirical layer-role hypothesis, its transfer across architectures, and a re-implementation of the baseline. Loss weights and pruning patterns are tuned on the evaluation benchmarks. No new physical or conceptual entities are introduced.

free parameters (4)
  • lambda_ICD = 0.1
    Loss weight for Individual Content Distillation, chosen by sensitivity analysis on FVD/VBench (Table 5).
  • lambda_MCA = 1
    Loss weight for Multi-frame Content Adversarial loss, chosen by sensitivity analysis on FVD/VBench (Table 5).
  • block pruning pattern = Remove Down-3, Mid, Up-0; halve blocks in Down-0, Down-1, Up-2, Up-3; for Hunyuan, Dual-stream 20 to 12 and…
    Pruning decisions derived from SF-V block importance analysis and transferred to other models without per-model re-evaluation.
  • VAE decoder pruning ratios = 70% latency reduction, 30% parameter reduction
    Layer and channel pruning to the VAE decoder, chosen to maximize speed while keeping reconstruction metrics comparable (Appendix A).
assumptions (4)
  • domain assumption Deeper layers of VDMs maintain motion dynamics; shallower layers focus on individual content.
    Empirical observation from Figure 1 and Section 3.2; used as the basis for the pruning strategy.
  • domain assumption Block roles transfer from SF-V to T2V-Turbo-v2 and HunyuanVideo.
    Appendix B states 'it is reasonable to assume that specific blocks within these U-Nets perform similar roles across both tasks.' Applied to DiT without evidence.
  • domain assumption The re-implemented SF-V baseline faithfully reproduces the original SF-V.
    Section C says SVD code is not public, so SF-V is re-implemented on OpenVid-1M; the teacher for distillation is this re-implementation.
  • standard math Standard diffusion, consistency, and adversarial training formulations from cited works.
    Equations (1) through (3) and (5) rely on EDM, consistency models, and Diffusion-GAN.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Individual Content and Motion Dynamics Preserved Pruning for Video Diffusion Models." pith.science (2026). https://pith.science/paper/YU5RKH3P

@misc{pith2026241118375,
  author       = {Pith},
  title        = {Pith review of: Individual Content and Motion Dynamics Preserved Pruning for Video Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YU5RKH3P}},
  note         = {Machine review of arXiv:2411.18375}
}
abstract

The high computational cost and slow inference time are major obstacles to deploying Video Diffusion Models (VDMs). To overcome this, we introduce a new Video Diffusion Model Compression approach using individual content and motion dynamics preserved pruning and consistency loss. First, we empirically observe that deeper VDM layers are crucial for maintaining the quality of \textbf{motion dynamics} (\textit{e.g.,} coherence of the entire video), while shallower layers are more focused on \textbf{individual content} (\textit{e.g.,} individual frames). Therefore, we prune redundant blocks from the shallower layers while preserving more of the deeper layers, resulting in a lightweight VDM variant called VDMini. Moreover, we propose an \textbf{Individual Content and Motion Dynamics (ICMD)} Consistency Loss to gain comparable generation performance as larger VDM to VDMini. In particular, we first use the Individual Content Distillation (ICD) Loss to preserve the consistency in the features of each generated frame between the teacher and student models. Next, we introduce a Multi-frame Content Adversarial (MCA) Loss to enhance the motion dynamics across the generated video as a whole. This method significantly accelerates inference time while maintaining high-quality video generation. Extensive experiments demonstrate the effectiveness of our VDMini on two important video generation tasks, Text-to-Video (T2V) and Image-to-Video (I2V), where we respectively achieve an average 2.5 $\times$, 1.4 $\times$, and 1.25 $\times$ speed up for the I2V method SF-V, the T2V method T2V-Turbo-v2, and the T2V method HunyuanVideo, while maintaining the quality of the generated videos on several benchmarks including UCF101, VBench-T2V, and VBench-I2V.

Figures

Figures reproduced from arXiv: 2411.18375 by the authors.

Figure 1
Figure 1. (a) FVD score by removing or replacing the blocks [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The proposed VDMini framework for Video Diffusion Model Compression. Left: The retraining process with the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Qualitative results of VDMini-I2V based on SVD. From top to bottom, the rows correspond to videos generated by SVD [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Qualitative Results of VDMini-T2V. For the first two rows, the Prompt is [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. On-device Sora: Enabling Training-Free Diffusion-based Text-to-Video Generation for Mobile Devices

    cs.CV 2025-02 conditional novelty 6.0 of 10

    A training-free pipeline makes diffusion text-to-video generation run on an iPhone 15 Pro with quality close to GPU output, at the cost of slower generation.

Reference graph

Works this paper leans on

75 extracted references · 56 canonical work pages · cited by 1 Pith paper

  1. [1]

    Kling AI. 2023. Kling AI. https://klingai.com/

    work page 2023

  2. [2]

    Max Bain, Arsha Nagrani, Gül Varol, and Andrew Zisserman. 2021. Frozen in time: A joint video and image encoder for end-to-end retrieval. InProc. IEEE Conf. CVPR. 1728–1738

    work page 2021

  3. [3]

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram Voleti, Adam Letts, Individual Content and Motion Dynamics Preserved Pruning for Video Diffusion Models MM ’25, October 27–31, 2025, Dublin, Ireland et al. 2023. Stable video diffusion: Scaling latent video diffusion models to ...

    arXiv 2023

  4. [4]

    Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dockhorn, Seung Wook Kim, Sanja Fidler, and Karsten Kreis. 2023. Align your latents: High-resolution video synthesis with latent diffusion models. In Proc. IEEE Conf. CVPR . 22563– 22575

    work page 2023

  5. [5]

    Cristian Buciluˇa, Rich Caruana, and Alexandru Niculescu-Mizil. 2006. Model compression

    work page 2006

  6. [7]

    Thibault Castells, Hyoung-Kyu Song, Tairen Piao, Shinkook Choi, Bo-Kyeong Kim, Hanyoung Yim, Changgwun Lee, Jae Gon Kim, and Tae-Ho Kim. 2024. Edge- Fusion: On-Device Text-to-Image Generation. arXiv preprint arXiv:2404.11925 (2024)

    arXiv 2024

  7. [8]

    Haoxin Chen, Menghan Xia, Yingqing He, Yong Zhang, Xiaodong Cun, Shaoshu Yang, Jinbo Xing, Yaofang Liu, Qifeng Chen, Xintao Wang, et al. 2023. Videocrafter1: Open diffusion models for high-quality video generation. arXiv preprint arXiv:2310.19512 (2023)

    arXiv 2023

  8. [9]

    Haoxin Chen, Yong Zhang, Xiaodong Cun, Menghan Xia, Xintao Wang, Chao Weng, and Ying Shan. 2024. Videocrafter2: Overcoming data limitations for high-quality video diffusion models. In Proc. IEEE Conf. CVPR . 7310–7320

    work page 2024

Show all 75 references
  1. [10]

    Jierun Chen, Dongting Hu, Xijie Huang, Huseyin Coskun, Arpit Sahni, Aarush Gupta, Anujraaj Goyal, Dishani Lahiri, Rajesh Singh, Yerlan Idelbayev, et al. 2025. Snapgen: Taming high-resolution text-to-image models for mobile devices with efficient architectures and training. In ...

    2025

  2. [11]

    Tsai-Shien Chen, Aliaksandr Siarohin, Willi Menapace, Ekaterina Deyneka, Hsiang-wei Chao, Byung Eun Jeon, Yuwei Fang, Hsin-Ying Lee, Jian Ren, Ming- Hsuan Yang, and Sergey Tulyakov. 2024. Panda-70M: Captioning 70M Videos with Multiple Cross-Modality Teachers. In Proc. IEEE Conf. CVPR

    2024

  3. [12]

    Zhenghao Chen, Shuhang Gu, Guo Lu, and Dong Xu. 2022. Exploiting intra- slice and inter-slice redundancy for learning-based lossless volumetric image compression. IEEE Transactions on Image Processing 31 (2022), 1697–1707

    2022

  4. [13]

    Zhenghao Chen, Guo Lu, Zhihao Hu, Shan Liu, Wei Jiang, and Dong Xu. 2022. LSVC: A learning-based stereo video compression framework. In Proc. IEEE Conf. CVPR. 6073–6082

    2022

  5. [14]

    Zhenghao Chen, Lucas Relic, Roberto Azevedo, Yang Zhang, Markus Gross, Dong Xu, Luping Zhou, and Christopher Schroers. 2023. Neural video compression with spatio-temporal cross-covariance transformers. In Proc. ACM Multimedia. 8543–8551

    2023

  6. [15]

    Zhenghao Chen, Luping Zhou, Zhihao Hu, and Dong Xu. 2024. Group-aware parameter-efficient updating for content-adaptive neural video compression. In Proc. ACM Multimedia. 11022–11031

    2024

  7. [16]

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. 2024. Scaling rectified flow transformers for high-resolution image synthesis. In Proc. ICML

    2024

  8. [17]

    Patrick Esser, Robin Rombach, and Bjorn Ommer. 2021. Taming transformers for high-resolution image synthesis. In Proc. IEEE Conf. CVPR . 12873–12883

    2021

  9. [18]

    Gongfan Fang, Xinyin Ma, Mingli Song, Michael Bi Mi, and Xinchao Wang. 2023. DepGraph: Towards Any Structural Pruning. In Proc. IEEE Conf. CVPR

    2023

  10. [19]

    Gongfan Fang, Xinyin Ma, and Xinchao Wang. 2023. Structural Pruning for Diffusion Models. In Proc. NeurIPS

    2023

  11. [20]

    Songwei Ge, Aniruddha Mahapatra, Gaurav Parmar, Jun-Yan Zhu, and Jia-Bin Huang. 2024. On the Content Bias in Fréchet Video Distance. In Proc. IEEE Conf. CVPR. 7277–7288

    2024

  12. [21]

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2020. Generative adversarial networks. Commun. ACM 63, 11 (2020), 139–144

    2020

  13. [22]

    Yuwei Guo, Ceyuan Yang, Anyi Rao, Zhengyang Liang, Yaohui Wang, Yu Qiao, Maneesh Agrawala, Dahua Lin, and Bo Dai. 2024. AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning. In Proc. ICLR

    2024

  14. [23]

    Song Han, Jeff Pool, John Tran, and William J Dally. 2015. Learning both weights and connections for efficient neural network. In Proc. NeurIPS

    2015

  15. [24]

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2014. Distilling the knowledge in a neural network. In Proc. NeurIPS Workshop

    2014

  16. [25]

    Jonathan Ho and Tim Salimans. 2021. Classifier-Free Diffusion Guidance. In Proc. NeurIPS Workshop

    2021

  17. [26]

    Jonathan Ho, Tim Salimans, Alexey Gritsenko, William Chan, Mohammad Norouzi, and David J Fleet. 2022. Video diffusion models. In Proc. NeurIPS , Vol. 35. 8633–8646

    2022

  18. [27]

    Torsten Hoefler, Dan Alistarh, Tal Ben-Nun, Nikoli Dryden, and Alexandra Peste

  19. [28]

    Ziqi Huang, Yinan He, Jiashuo Yu, Fan Zhang, Chenyang Si, Yuming Jiang, Yuanhan Zhang, Tianxing Wu, Qingyang Jin, Nattapol Chanpaisit, Yaohui Wang, Xinyuan Chen, Limin Wang, Dahua Lin, Yu Qiao, and Ziwei Liu. 2024. VBench: Comprehensive Benchmark Suite for Video Generative Mod...

    2024

  20. [29]

    Ziqi Huang, Fan Zhang, Xiaojie Xu, Yinan He, Jiashuo Yu, Ziyue Dong, Qianli Ma, Nattapol Chanpaisit, Chenyang Si, Yuming Jiang, et al . 2024. Vbench++: Comprehensive and versatile benchmark suite for video generative models.arXiv preprint arXiv:2411.13503 (2024)

    2024 arXiv

  21. [30]

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. 2022. Elucidating the design space of diffusion-based generative models. In Proc. NeurIPS , Vol. 35. 26565–26577

    2022

  22. [31]

    Tero Karras, Miika Aittala, Jaakko Lehtinen, Janne Hellsten, Timo Aila, and Samuli Laine. 2024. Analyzing and improving the training dynamics of diffusion models. In Proc. IEEE Conf. CVPR . 24174–24184

    2024

  23. [32]

    Levon Khachatryan, Andranik Movsisyan, Vahram Tadevosyan, Roberto Hen- schel, Zhangyang Wang, Shant Navasardyan, and Humphrey Shi. 2023. Text2video-zero: Text-to-image diffusion models are zero-shot video genera- tors. In Proc. IEEE Conf. CVPR . 15954–15964

    2023

  24. [33]

    Bo-Kyeong Kim, Hyoung-Kyu Song, Thibault Castells, and Shinkook Choi. 2024. BK-SDM: A Lightweight, Fast, and Cheap Version of Stable Diffusion. In Proc. ECCV

    2024

  25. [34]

    Diederik P Kingma. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 (2013)

    2013 arXiv

  26. [35]

    Weijie Kong, Qi Tian, Zijian Zhang, Rox Min, Zuozhuo Dai, Jin Zhou, Jiangfeng Xiong, Xin Li, Bo Wu, Jianwei Zhang, et al. 2024. Hunyuanvideo: A systematic framework for large video generative models. arXiv preprint arXiv:2412.03603 (2024)

    2024 arXiv

  27. [36]

    PKU-Yuan Lab and Tuzhan AI etc. 2024. Open-Sora-Plan. doi:10.5281/zenodo. 10948109

    2024 doi

  28. [37]

    Luma Labs. 2024. Dream Machine. https://lumalabs.ai/dream-machine

    2024

  29. [38]

    Hao Li, Asim Kadav, Igor Durdanovic, Hanan Samet, and Hans Peter Graf. 2017. Pruning filters for efficient convnets. In Proc. ICLR

    2017

  30. [39]

    Jiachen Li, Qian Long, Jian Zheng, Xiaofeng Gao, Robinson Piramuthu, Wenhu Chen, and William Yang Wang. 2024. T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design. arXiv preprint arXiv:2410.05677 (2024)

    2024

  31. [40]

    Yanyu Li, Huan Wang, Qing Jin, Ju Hu, Pavlo Chemerys, Yun Fu, Yanzhi Wang, Sergey Tulyakov, and Jian Ren. 2023. SnapFusion: Text-to-Image Diffusion Model on Mobile Devices within Two Seconds. In Proc. NeurIPS

    2023

  32. [41]

    Bin Lin, Yunyang Ge, Xinhua Cheng, Zongjian Li, Bin Zhu, Shaodong Wang, Xianyi He, Yang Ye, Shenghai Yuan, Liuhan Chen, et al. 2024. Open-sora plan: Open-source large video generation model. arXiv preprint arXiv:2412.00131 (2024)

    2024 arXiv

  33. [42]

    Pengyang Ling, Jiazi Bu, Pan Zhang, Xiaoyi Dong, Yuhang Zang, Tong Wu, Huaian Chen, Jiaqi Wang, and Yi Jin. 2024. MotionClone: Training-Free Motion Cloning for Controllable Video Generation. arXiv preprint arXiv:2406.05338 (2024)

    2024 arXiv

  34. [43]

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matthew Le. 2023. Flow Matching for Generative Modeling. In Proc. ICLR

    2023

  35. [44]

    Simian Luo, Yiqin Tan, Longbo Huang, Jian Li, and Hang Zhao. 2023. Latent consistency models: Synthesizing high-resolution images with few-step inference. arXiv preprint arXiv:2310.04378 (2023)

    2023 arXiv

  36. [45]

    Xiaofeng Mao, Zhengkai Jiang, Fu-Yun Wang, Wenbing Zhu, Jiangning Zhang, Hao Chen, Mingmin Chi, and Yabiao Wang. 2024. Osv: One step is enough for high-quality image to video generation. In Proc. IEEE Conf. CVPR

    2024

  37. [46]

    Willi Menapace, Aliaksandr Siarohin, Ivan Skorokhodov, Ekaterina Deyneka, Tsai-Shien Chen, Anil Kag, Yuwei Fang, Aleksei Stoliar, Elisa Ricci, Jian Ren, et al

  38. [47]

    Chenlin Meng, Ruiqi Gao, Diederik P Kingma, Stefano Ermon, Jonathan Ho, and Tim Salimans. 2023. On distillation of guided diffusion models. InProc. IEEE Conf. CVPR

    2023

  39. [48]

    Pavlo Molchanov, Stephen Tyree, Tero Karras, Timo Aila, and Jan Kautz. 2017. Pruning Convolutional Neural Networks for Resource Efficient Inference. InProc. ICLR

    2017

  40. [49]

    Michael C Mozer and Paul Smolensky. 1988. Skeletonization: A technique for trimming the fat from a network via relevance assessment. In Proc. NeurIPS

    1988

  41. [50]

    Kepan Nan, Rui Xie, Penghao Zhou, Tiehan Fan, Zhenheng Yang, Zhijie Chen, Xiang Li, Jian Yang, and Ying Tai. 2024. Openvid-1m: A large-scale high-quality dataset for text-to-video generation. arXiv preprint arXiv:2407.02371 (2024)

    2024 arXiv

  42. [51]

    OpenAI. 2024. Sora. https://openai.com/index/video-generation-models-as- world-simulators/

    2024

  43. [52]

    William Peebles and Saining Xie. 2023. Scalable Diffusion Models with Trans- formers. Proc. IEEE Conf. ICCV

    2023

  44. [53]

    Russell Reed. 1993. Pruning algorithms - a survey. IEEE transactions on Neural Networks 4, 5 (1993), 740–747

    1993

  45. [54]

    Tim Salimans and Jonathan Ho. 2022. Progressive Distillation for Fast Sampling of Diffusion Models. In Proc. ICLR. MM ’25, October 27–31, 2025, Dublin, Ireland Wu et al

    2022

  46. [55]

    Axel Sauer, Frederic Boesel, Tim Dockhorn, Andreas Blattmann, Patrick Esser, and Robin Rombach. 2024. Fast high-resolution image synthesis with latent adversarial diffusion distillation. 1–11

    2024

  47. [56]

    Ivan Skorokhodov, Sergey Tulyakov, and Mohamed Elhoseiny. 2022. Stylegan-v: A continuous video generator with the price, image quality and perks of stylegan2. In Proc. IEEE Conf. CVPR . 3626–3636

    2022

  48. [57]

    Jiaming Song, Chenlin Meng, and Stefano Ermon. 2021. Denoising diffusion implicit models. Proc. ICLR

    2021

  49. [58]

    Yang Song and Prafulla Dhariwal. 2024. Improved Techniques for Training Consistency Models. In Proc. ICLR

    2024

  50. [59]

    Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. 2023. Consistency Models. In Proc. ICML. 127–144. doi:10.1007/978-1-4842-1329-2_9

    2023 doi

  51. [60]

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. 2021. Score-Based Generative Modeling through Sto- chastic Differential Equations. In Proc. ICLR. https://openreview.net/forum?id= PxTIG12RRHS

    2021

  52. [61]

    Zhiyu Tan, Xiaomeng Yang, Luozheng Qin, and Hao Li. 2024. VidGen-1M: A Large-Scale Dataset for Text-to-video Generation.arXiv preprint arXiv:2408.02629 (2024)

    2024 arXiv

  53. [62]

    Genmo Team. 2024. Mochi 1. https://github.com/genmoai/models

    2024

  54. [63]

    Fu-Yun Wang, Zhaoyang Huang, Xiaoyu Shi, Weikang Bian, Guanglu Song, Yu Liu, and Hongsheng Li. 2024. AnimateLCM: Computation-Efficient Personalized Style Video Generation without Personalized Video Data. SIGGRAPH ASIA TCS

    2024

  55. [64]

    Huan Wang, Can Qin, Yulun Zhang, and Yun Fu. 2021. Neural Pruning via Growing Regularization. In Proc. ICLR

    2021

  56. [65]

    Zhendong Wang, Huangjie Zheng, Pengcheng He, Weizhu Chen, and Mingyuan Zhou. 2023. Diffusion-GAN: Training GANs with Diffusion. In Proc. ICLR

    2023

  57. [66]

    WanTeam, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, Jianyuan Zeng, Jiayu Wang, Jingfeng Zhang, Jingren Zhou, Jinkai Wang, Jixuan Chen, Kai Zhu, Kang Zhao, Keyu Yan, Lianghua Huang, Mengyang Feng, Ningyi Zhang, Panden...

    2025 arXiv

  58. [67]

    Yushu Wu, Zhixing Zhang, Yanyu Li, Yanwu Xu, Anil Kag, Yang Sui, Huseyin Coskun, Ke Ma, Aleksei Lebedev, Ju Hu, et al. 2025. Snapgen-v: Generating a five-second video within five seconds on a mobile device. In Proc. IEEE Conf. CVPR. 2479–2490

    2025

  59. [68]

    Yanwu Xu, Yang Zhao, Zhisheng Xiao, and Tingbo Hou. 2024. Ufogen: You forward once large scale text-to-image generation via diffusion gans. In Proc. IEEE Conf. CVPR

    2024

  60. [69]

    Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, Yuanming Yang, Wenyi Hong, Xiaohan Zhang, Guanyu Feng, et al . 2024. CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer. arXiv preprint arXiv:2408.06072 (2024)

    2024 arXiv

  61. [70]

    Yuanhao Zhai, Kevin Lin, Zhengyuan Yang, Linjie Li, Jianfeng Wang, Chung- Ching Lin, David Doermann, Junsong Yuan, and Lijuan Wang. [n. d.]. Motion Consistency Model: Accelerating Video Diffusion with Disentangled Motion- Appearance Distillation. In Proc. NeurIPS

  62. [71]

    Yabo Zhang, Yuxiang Wei, Dongsheng Jiang, XIAOPENG ZHANG, Wangmeng Zuo, and Qi Tian. 2024. ControlVideo: Training-free Controllable Text-to-video Generation. In Proc. ICLR

    2024

  63. [72]

    Zhixing Zhang, Yanyu Li, Yushu Wu, Yanwu Xu, Anil Kag, Ivan Skorokhodov, Willi Menapace, Aliaksandr Siarohin, Junli Cao, Dimitris Metaxas, et al . 2024. SF-V: Single Forward Video Generation Model. In Proc. NeurIPS

    2024

  64. [73]

    Yang Zhao, Yanwu Xu, Zhisheng Xiao, and Tingbo Hou. 2023. Mobilediffu- sion: Subsecond text-to-image generation on mobile devices. arXiv preprint arXiv:2311.16567 (2023)

    2023 arXiv

  65. [74]

    MidBlock

    Zangwei Zheng, Xiangyu Peng, Tianji Yang, Chenhui Shen, Shenggui Li, Hongxin Liu, Yukun Zhou, Tianyi Li, and Yang You. 2024. Open-Sora: Democratizing Efficient Video Production for All . https://github.com/hpcaitech/Open-Sora This appendix provides details about Section A, the...

    2024

  66. [2021]

    JMLR 22, 241 (2021), 1–124

    Sparsity in Deep Learning: Pruning and growth for efficient inference and training in neural networks. JMLR 22, 241 (2021), 1–124

    2021

  67. [2024]

    Snap video: Scaled spatiotemporal transformers for text-to-video synthesis. In Proc. IEEE Conf. CVPR . 7038–7048

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.