pith. sign in

arxiv: 2606.21030 · v1 · pith:UYTPCZHMnew · submitted 2026-06-19 · 📡 eess.IV · cs.CV

FlowCodec: One-Step Flow Prior for Generative Image Compression

Yinhuan Huang , Hao Cao , Pu chen , Wenqi Guo , Zhijin Qin This is my paper

Pith reviewed 2026-06-26 13:26 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords generative image compressionflow priorone-step transportlatent compressionultra-low bitratetext-to-image priorsperceptual quality
0
0 comments X

The pith

FlowCodec integrates pretrained text-to-image priors into compression via one-step latent transport without extra conditioning or auxiliary networks.

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

The paper sets out to demonstrate that large generative models can be adapted for ultra-low-bitrate image compression through a simpler two-stage design than current approaches require. It separates the process into first mapping clean latents to noisy constrained versions for bitrate control, then applying the pretrained prior to move those noisy latents back toward clean ones in a single step. This matters to a sympathetic reader because scaling generative models typically demands more engineering and retraining, so a lightweight plug-in method could make stronger priors usable in practice with far less overhead. The work reports that the resulting codecs maintain visual quality below 0.05 bits per pixel while improving perceptual scores and encoding speed over existing one-step methods.

Core claim

FlowCodec shows that pretrained text-to-image priors can be inserted directly into an image codec by decoupling latent compression from a single-step latent transport, with no additional conditioning signals or auxiliary networks needed, and that lightweight adaptation alone supports multiple bitrates while keeping trainable parameters below 0.54 percent of the backbone.

What carries the argument

The one-step latent transport stage, which uses the pretrained flow prior to refine bitrate-constrained noisy latents toward clean ones.

If this is right

  • High visual quality is preserved at bitrates below 0.05 bits per pixel.
  • The Qwen-image variant outperforms existing methods on LPIPS and DISTS.
  • Both variants achieve higher PSNR and faster encoding than existing one-step diffusion-based methods.
  • Multiple bitrates can be supported while training under 0.54 percent of the generative backbone parameters.

Where Pith is reading between the lines

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

  • The same decoupled compression-plus-transport pattern could be tested on video or audio using corresponding pretrained priors.
  • Larger backbones might yield further gains at the same low adaptation cost.
  • Direct prior reuse may reduce task-specific retraining needs in other low-rate generative tasks.

Load-bearing premise

That pretrained text-to-image models can directly refine noisy latents into high-quality clean images in one step with only lightweight adaptation and no extra conditioning signals.

What would settle it

A side-by-side test of whether removing the lightweight adaptation or forcing multi-step transport produces a measurable drop in LPIPS, DISTS, or PSNR at the same bitrate.

Figures

Figures reproduced from arXiv: 2606.21030 by Hao Cao, Pu Chen, Wenqi Guo, Yinhuan Huang, Zhijin Qin.

Figure 1
Figure 1. Figure 1: Visual comparison of different image compression methods. DCVC-RT [ [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Motivation of FlowCodec. Existing diffusion-based image compression methods often rely on task-specific components, such as [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The framework of FlowCodec. We perform both Latent Compression and Latent Transport in the latent space of a frozen VAE. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of Latent Transport. The clean latent [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Rate-distortion and rate-perception curves on Kodak, Tecnick, DIV2K, and CLIC 2020. We report distortion fidelity using PSNR [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visual comparison of diffusion-based methods. Zoom in for a better view. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visual ablation of generative-prior components. Variants are obtained by removing different modules. Zoom in for a better view. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Additional semantic and perceptual metrics, including [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visual comparison (1800×1200) of diffusion codecs. Zoom in for a better view. 4 [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Visual comparison (2040×1356) of diffusion codecs. Zoom in for a better view. 5 [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Visual comparison (1228×1840) of diffusion codecs. Zoom in for a better view. 6 [PITH_FULL_IMAGE:figures/full_fig_p016_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Visual comparison (512×768) of diffusion codecs. Zoom in for a better view. 7 [PITH_FULL_IMAGE:figures/full_fig_p017_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Visual comparison (2048×2048) of diffusion codecs. Zoom in for a better view. 8 [PITH_FULL_IMAGE:figures/full_fig_p018_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Visual ablation (2040×1464) of generative-prior components. Zoom in for a better view. 9 [PITH_FULL_IMAGE:figures/full_fig_p019_14.png] view at source ↗
read the original abstract

Diffusion-based image compression methods, leveraging powerful generative priors, have demonstrated remarkable perceptual quality at ultra-low bitrates. However, adapting modern generative models to image compression often relies on carefully engineered conditioning or auxiliary branches, together with substantial retraining, and these costs grow as the models scale. This motivates an open question: Can stronger generative priors be integrated into compression through a simpler, more extensible design? To answer this, we propose FlowCodec, a streamlined framework that plugs pretrained large-scale text-to-image priors (e.g., Qwen-image-2512 and FLUX.1-dev) into ultra-low-bitrate codecs. FlowCodec decomposes the pipeline into two decoupled stages: (1) Latent Compression, which maps clean latents to bitrate-constrained noisy latents; and (2) Latent Transport, which leverages the pretrained prior to refine the noisy latents toward the clean ones in a single step. Notably, FlowCodec requires neither additional conditioning signals nor auxiliary networks. Furthermore, with lightweight adaptation, it can flexibly support multiple bitrates while keeping the number of trainable parameters below 0.54% of the generative backbone. Experiments show that FlowCodec preserves high visual quality at bitrates below 0.05 bits per pixel. The Qwen-image variant significantly outperforms existing methods in terms of LPIPS and DISTS, while both variants deliver higher PSNR and clearly faster encoding than existing one-step diffusion-based methods, with the FLUX variant also maintaining competitive decoding speed.

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

0 major / 2 minor

Summary. The paper proposes FlowCodec, a two-stage framework for generative image compression at ultra-low bitrates that decouples latent compression (mapping clean latents to bitrate-constrained noisy latents) from single-step latent transport (using a pretrained text-to-image prior such as Qwen-image-2512 or FLUX.1-dev to refine noisy latents to clean ones). The design requires no additional conditioning signals or auxiliary networks, supports multiple bitrates via lightweight adaptation with <0.54% trainable parameters relative to the generative backbone, and claims to deliver high visual quality below 0.05 bpp while outperforming prior one-step diffusion methods in LPIPS, DISTS, PSNR, and encoding speed.

Significance. If the central claims are substantiated by the experiments, the work would offer a meaningful simplification for incorporating large-scale generative priors into compression codecs, reducing the engineering overhead of conditioning design and retraining while preserving extensibility and efficiency through the single-step transport mechanism. This could facilitate broader adoption of powerful pretrained models for perceptual compression tasks.

minor comments (2)
  1. The abstract references specific performance metrics (LPIPS, DISTS, PSNR) and bitrate thresholds but does not indicate the evaluation datasets or number of test images; adding this information would strengthen reproducibility claims.
  2. The description of 'lightweight adaptation' for multi-bitrate support would benefit from a brief statement on the adaptation mechanism (e.g., whether it involves LoRA-style updates or other parameter-efficient methods) to clarify the <0.54% parameter claim.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their careful reading and for recognizing the potential of FlowCodec to simplify the integration of large-scale generative priors into compression codecs. The report raises no specific major comments or requests for clarification, so we provide no point-by-point rebuttals below. We remain available to supply additional experimental details or address any unlisted concerns the referee may have.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The provided abstract and description contain no equations, fitted parameters presented as predictions, or self-citations that bear the central claims. The method is described as plugging in external pretrained models (Qwen-image, FLUX) with lightweight adaptation, and performance is evaluated against external benchmarks. No load-bearing step reduces by construction to an input defined within the paper itself. This is the common case of a self-contained empirical proposal relying on independent external priors and comparative experiments.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available, so no explicit free parameters, axioms, or invented entities can be extracted.

pith-pipeline@v0.9.1-grok · 5804 in / 1053 out tokens · 16334 ms · 2026-06-26T13:26:48.995630+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

67 extracted references · 12 canonical work pages · 9 internal anchors

  1. [1]

    Ntire 2017 challenge on single image super-resolution: Dataset and study

    Eirikur Agustsson and Radu Timofte. Ntire 2017 challenge on single image super-resolution: Dataset and study. InIEEE Conf. Comput. Vis. Pattern Recog. Worksh., pages 126–135,

    2017

  2. [2]

    Testimages: A large- scale archive for testing visual devices and basic image pro- cessing algorithms

    Nicola Asuni, Andrea Giachetti, et al. Testimages: A large- scale archive for testing visual devices and basic image pro- cessing algorithms. InSTAG: Smart Tools and Applications in Computer Graphics, pages 63–70, 2014. 7, 1, 3

    2014

  3. [3]

    Qwen3-VL Technical Report

    Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xionghui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Chang Gao, Chunjiang Ge, et al. Qwen3-VL technical report.arXiv preprint arXiv:2511.21631, 2025. 10

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  4. [4]

    Variational image compression with a scale hyperprior

    Johannes Ball ´e, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston. Variational image compression with a scale hyperprior. InInt. Conf. Learn. Represent., 2018. 3

    2018

  5. [5]

    CompressAI: A pytorch library and evalua- tion platform for end-to-end compression research.arXiv preprint arXiv:2011.03029, 2020

    Jean B ´egaint, Fabien Racap ´e, Simon Feltman, and Akshay Pushparaja. CompressAI: A pytorch library and evalua- tion platform for end-to-end compression research.arXiv preprint arXiv:2011.03029, 2020. 6

    work page arXiv 2011

  6. [6]

    Calculation of average PSNR differences between RD-curves.ITU-T SG16, Doc

    Gisle Bjontegaard. Calculation of average PSNR differences between RD-curves.ITU-T SG16, Doc. VCEG-M33, 2001. 7, 1

    2001

  7. [7]

    The perception-distortion tradeoff

    Yochai Blau and Tomer Michaeli. The perception-distortion tradeoff. InIEEE Conf. Comput. Vis. Pattern Recog., pages 6228–6237, 2018. 1

    2018

  8. [8]

    Rethinking lossy compres- sion: The rate-distortion-perception tradeoff

    Yochai Blau and Tomer Michaeli. Rethinking lossy compres- sion: The rate-distortion-perception tradeoff. InInt. Conf. Mach. Learn., pages 675–685, 2019. 1

    2019

  9. [9]

    HunyuanImage 3.0 Technical Report

    Siyu Cao, Hangting Chen, Peng Chen, Yiji Cheng, Yutao Cui, Xinchi Deng, Ying Dong, Kipper Gong, Tianpeng Gu, Xiusen Gu, et al. HunyuanImage 3.0 technical report.arXiv preprint arXiv:2509.23951, 2025. 2, 4

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  10. [10]

    Towards image compression with per- fect realism at ultra-low bitrates

    Marlene Careil, Matthew J Muckley, Jakob Verbeek, and St´ephane Lathuili`ere. Towards image compression with per- fect realism at ultra-low bitrates. InInt. Conf. Learn. Repre- sent., 2023. 2, 3

    2023

  11. [11]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. InIEEE Conf. Comput. Vis. Pattern Recog., pages 248–255. IEEE, 2009. 6

    2009

  12. [12]

    Arcface: Additive angular margin loss for deep face recognition

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. InCVPR, 2019. 1

    2019

  13. [13]

    Image quality assessment: Unifying structure and texture similarity.IEEE Trans

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simoncelli. Image quality assessment: Unifying structure and texture similarity.IEEE Trans. Pattern Anal. Mach. Intell., 44(5): 2567–2581, 2020. 2, 7

    2020

  14. [14]

    Scaling rec- tified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, and Robin Rombach. Scaling rec- tified flow transformers for high-resolution image synthesis. InInt. Conf. Mach. Learn., pages 12606–12633, 2024. 4

    2024

  15. [15]

    Linear attention model- ing for learned image compression

    Donghui Feng, Zhengxue Cheng, Shen Wang, Ronghua Wu, Hongwei Hu, Guo Lu, and Li Song. Linear attention model- ing for learned image compression. InIEEE Conf. Comput. Vis. Pattern Recog., pages 7623–7632, 2025. 3

    2025

  16. [16]

    Seedream 3.0 Technical Report

    Yu Gao, Lixue Gong, Qiushan Guo, Xiaoxia Hou, Zhichao Lai, Fanshi Li, Liang Li, Xiaochen Lian, Chao Liao, Liyang Liu, et al. Seedream 3.0 technical report.arXiv preprint arXiv:2504.11346, 2025. 2, 4

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  17. [17]

    Generative adversarial nets

    Ian J Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. InAdv. Neural Inform. Process. Syst.Curran Associates, Inc., 2014. 1, 2, 5

    2014

  18. [18]

    OSCAR: One- step diffusion codec across multiple bit-rates

    Jinpei Guo, Yifei Ji, Zheng Chen, Kai Liu, Min Liu, Wang Rao, Wenbo Li, Yong Guo, and Yulun Zhang. OSCAR: One- step diffusion codec across multiple bit-rates. InAdv. Neural Inform. Process. Syst.Curran Associates, Inc., 2025. 1, 2, 3, 4, 5, 7

    2025

  19. [19]

    ELIC: Efficient learned image compres- sion with unevenly grouped space-channel contextual adap- tive coding

    Dailan He, Ziming Yang, Weikun Peng, Rui Ma, Hongwei Qin, and Yan Wang. ELIC: Efficient learned image compres- sion with unevenly grouped space-channel contextual adap- tive coding. InIEEE Conf. Comput. Vis. Pattern Recog., pages 5718–5727, 2022. 3, 6

    2022

  20. [20]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. InAdv. Neural Inform. Process. Syst., pages 6840–6851. Curran Associates, Inc., 2020. 4

    2020

  21. [21]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. Int. Conf. Learn. Represent., 1(2):3, 2022. 3

    2022

  22. [22]

    Generative latent coding for ultra-low bitrate image com- pression

    Zhaoyang Jia, Jiahao Li, Bin Li, Houqiang Li, and Yan Lu. Generative latent coding for ultra-low bitrate image com- pression. InIEEE Conf. Comput. Vis. Pattern Recog., pages 26088–26098, 2024. 1, 3, 7

    2024

  23. [23]

    Towards practical real-time neu- ral video compression

    Zhaoyang Jia, Bin Li, Jiahao Li, Wenxuan Xie, Linfeng Qi, Houqiang Li, and Yan Lu. Towards practical real-time neu- ral video compression. InIEEE Conf. Comput. Vis. Pattern Recog., pages 12543–12552, 2025. 1, 3, 5, 7

    2025

  24. [24]

    Ultra lowrate image compression with semantic residual coding and compression-aware diffu- sion

    Anle Ke, Xu Zhang, Tong Chen, Ming Lu, Chao Zhou, Ji- awen Gu, and Zhan Ma. Ultra lowrate image compression with semantic residual coding and compression-aware diffu- sion. InInt. Conf. Mach. Learn., pages 29626–29650, 2025. 2, 3, 7, 1

    2025

  25. [25]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. InICCV, 2021. 1

    2021

  26. [26]

    Adam: A Method for Stochastic Optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980,

    work page internal anchor Pith review Pith/arXiv arXiv

  27. [27]

    Auto-Encoding Variational Bayes

    Diederik P Kingma and Max Welling. Auto-encoding varia- tional bayes.arXiv preprint arXiv:1312.6114, 2013. 3

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  28. [28]

    http://r0k.us/graphics/kodak/, 1993

    Kodak lossless true color image suite. http://r0k.us/graphics/kodak/, 1993. 7, 1, 3

    1993

  29. [29]

    Alina Kuznetsova, Hassan Rom, Neil Alldrin, Jasper Ui- jlings, Ivan Krasin, Jordi Pont-Tuset, Shahab Kamali, Stefan Popov, Matteo Malloci, Alexander Kolesnikov, et al. The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale.Interna- tional journal of computer vision, 128(7):1956–1981, 2020. 6 10

    1956

  30. [30]

    Black Forest Labs. Flux. https://github.com/black-forest- labs/flux, 2024. 2, 5, 6

    2024

  31. [31]

    Flux.1 kontext: Flow matching for in-context image generation and editing in latent space,

    Black Forest Labs, Stephen Batifol, Andreas Blattmann, Frederic Boesel, Saksham Consul, Cyril Diagne, Tim Dock- horn, Jack English, Zion English, Patrick Esser, Sumith Ku- lal, Kyle Lacey, Yam Levi, Cheng Li, Dominik Lorenz, Jonas M¨uller, Dustin Podell, Robin Rombach, Harry Saini, Axel Sauer, and Luke Smith. Flux.1 kontext: Flow matching for in-context i...

  32. [32]

    Text+ sketch: Image compression at ultra low rates

    Eric Lei, Yigit Berkay Uslu, Hamed Hassani, and Shirin Saeedi Bidokhti. Text+ sketch: Image compression at ultra low rates. InInt. Conf. Mach. Learn. Worksh., 2023. 2, 3

    2023

  33. [33]

    Once-for-All: Controllable generative image compression with dynamic granularity adaptation

    Anqi Li, Feng Li, Yuxi Liu, Runmin Cong, Yao Zhao, and Huihui Bai. Once-for-All: Controllable generative image compression with dynamic granularity adaptation. InInt. Conf. Learn. Represent., 2025. 2, 3, 7, 1

    2025

  34. [34]

    Neural video compression with diverse contexts

    Jiahao Li, Bin Li, and Yan Lu. Neural video compression with diverse contexts. InIEEE Conf. Comput. Vis. Pattern Recog., pages 22616–22626, 2023. 5

    2023

  35. [35]

    Neural video compression with feature modulation

    Jiahao Li, Bin Li, and Yan Lu. Neural video compression with feature modulation. InIEEE Conf. Comput. Vis. Pattern Recog., pages 26099–26108, 2024. 5, 6

    2024

  36. [36]

    Towards extreme image compression with latent feature guidance and diffusion prior.IEEE Trans

    Zhiyuan Li, Yanhui Zhou, Hao Wei, Chenyang Ge, and Jing- wen Jiang. Towards extreme image compression with latent feature guidance and diffusion prior.IEEE Trans. Circuit Syst. Video Technol., 35(1):888–899, 2025. 2, 3, 7, 1

    2025

  37. [37]

    RDEIC: Accelerating diffusion-based extreme image compression with relay residual diffusion.IEEE Trans

    Zhiyuan Li, Yanhui Zhou, Hao Wei, Chenyang Ge, and Aj- mal Mian. RDEIC: Accelerating diffusion-based extreme image compression with relay residual diffusion.IEEE Trans. Circuit Syst. Video Technol., 35(11):11540–11552,

  38. [38]

    Flow matching for generative modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximil- ian Nickel, and Matthew Le. Flow matching for generative modeling. InInt. Conf. Learn. Represent., 2022. 4

    2022

  39. [39]

    Learned im- age compression with mixed transformer-cnn architectures

    Jinming Liu, Heming Sun, and Jiro Katto. Learned im- age compression with mixed transformer-cnn architectures. InIEEE Conf. Comput. Vis. Pattern Recog., pages 14388– 14397, 2023. 3

    2023

  40. [40]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    Xingchao Liu, Chengyue Gong, et al. Flow straight and fast: Learning to generate and transfer data with rectified flow. In Int. Conf. Learn. Represent., 2023. 4

    2023

  41. [41]

    High-fidelity generative image compres- sion

    Fabian Mentzer, George D Toderici, Michael Tschannen, and Eirikur Agustsson. High-fidelity generative image compres- sion. InAdv. Neural Inform. Process. Syst., pages 11913– 11924. Curran Associates, Inc., 2020. 2, 3

    2020

  42. [42]

    completely blind

    Anish Mittal, Rajiv Soundararajan, and Alan C. Bovik. Mak- ing a “completely blind” image quality analyzer.SPL, 2013. 1

    2013

  43. [43]

    Improving statistical fi- delity for neural image compression with implicit local like- lihood models

    Matthew J Muckley, Alaaeldin El-Nouby, Karen Ullrich, Herv´e J ´egou, and Jakob Verbeek. Improving statistical fi- delity for neural image compression with implicit local like- lihood models. InInt. Conf. Mach. Learn., pages 25426– 25443. PMLR, 2023. 2, 3, 6, 7, 1

    2023

  44. [44]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models. InIEEE Conf. Comput. Vis. Pattern Recog., pages 10684–10695, 2022. 2, 3, 4, 9

    2022

  45. [45]

    Adversarial diffusion distillation

    Axel Sauer, Dominik Lorenz, Andreas Blattmann, and Robin Rombach. Adversarial diffusion distillation. InEur. Conf. Comput. Vis., pages 87–103, 2024. 3

    2024

  46. [46]

    Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model

    Team Seedance, Heyi Chen, Siyan Chen, Xin Chen, Yan- fei Chen, Ying Chen, Zhuo Chen, Feng Cheng, Tianheng Cheng, Xinqi Cheng, et al. Seedance 1.5 pro: A native audio- visual joint generation foundation model.arXiv preprint arXiv:2512.13507, 2025. 2, 4

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  47. [47]

    Very Deep Convolutional Networks for Large-Scale Image Recognition

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition.arXiv preprint arXiv:1409.1556, 2014. 7

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  48. [48]

    A hybrid scheme for face video compres- sion.ACM Trans

    Anni Tang, Zhiyu Zhang, Chen Zhu, Jun Ling, Rong Xie, and Li Song. A hybrid scheme for face video compres- sion.ACM Trans. Multimedia Comput. Commun. Appl., 22 (1), 2026. 3

    2026

  49. [49]

    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. 2, 4

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  50. [50]

    StableCodec: Taming one-step diffusion for extreme image compression

    Zhang Tianyu, Luo Xin, Li Li, and Liu Dong. StableCodec: Taming one-step diffusion for extreme image compression. InInt. Conf. Comput. Vis., 2025. 1, 2, 3, 4, 7

    2025

  51. [51]

    Workshop and challenge on learned image compression (clic2020)

    George Toderici, Wenzhe Shi, Radu Timofte, Lucas Theis, Johannes Balle, Eirikur Agustsson, Nick Johnston, and Fabian Mentzer. Workshop and challenge on learned image compression (clic2020). InIEEE Conf. Comput. Vis. Pattern Recog., 2020. 7, 1, 3

    2020

  52. [52]

    https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware VTM/

    VTM-23.10. https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware VTM/. Accessed: 2026-03-02. 3, 7, 1

    2026

  53. [53]

    G.K. Wallace. The jpeg still picture compression standard. IEEE Transactions on Consumer Electronics, 38(1), 1992. 3

    1992

  54. [54]

    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

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  55. [55]

    Ex- ploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Ex- ploring clip for assessing the look and feel of images. In AAAI, 2023. 1

    2023

  56. [56]

    One- shot free-view neural talking-head synthesis for video con- ferencing

    Ting-Chun Wang, Arun Mallya, and Ming-Yu Liu. One- shot free-view neural talking-head synthesis for video con- ferencing. InIEEE Conf. Comput. Vis. Pattern Recog., pages 10039–10049, 2021. 3

    2021

  57. [57]

    Multi- scale structural similarity for image quality assessment

    Zhou Wang, Eero P Simoncelli, and Alan C Bovik. Multi- scale structural similarity for image quality assessment. In The thirty-seventh Asilomar Conference on Signals, Systems & Computers, 2003, pages 1398–1402. IEEE, 2003. 3

    2003

  58. [58]

    Hunyuanvideo 1.5 technical report, 2025

    Bing Wu, Chang Zou, Changlin Li, Duojun Huang, Fang Yang, Hao Tan, Jack Peng, Jianbing Wu, Jiangfeng Xiong, Jie Jiang, et al. Hunyuanvideo 1.5 technical report, 2025. 2, 4

    2025

  59. [59]

    Qwen-image technical report, 2025

    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, 2025. 2, 4, 5, 6 11

    2025

  60. [60]

    Sana 1.5: Efficient scaling of training-time and inference-time compute in linear diffusion transformer.arXiv preprint arXiv:2501.18427,

    Enze Xie, Junsong Chen, Yuyang Zhao, Jincheng Yu, Ligeng Zhu, Chengyue Wu, Yujun Lin, Zhekai Zhang, Muyang Li, Junyu Chen, et al. Sana 1.5: Efficient scaling of training-time and inference-time compute in linear diffusion transformer. arXiv:2501.18427, 2025. 9

    work page arXiv 2025

  61. [61]

    One-step diffusion-based image compression with semantic distillation

    Naifu Xue, Zhaoyang Jia, Jiahao Li, Bin Li, Yuan Zhang, and Yan Lu. One-step diffusion-based image compression with semantic distillation. InAdv. Neural Inform. Process. Syst.Curran Associates, Inc., 2025. 1, 2, 3, 4, 7

    2025

  62. [62]

    DLF: Extreme image compression with dual- generative latent fusion

    Naifu Xue, Zhaoyang Jia, Jiahao Li, Bin Li, Yuan Zhang, and Yan Lu. DLF: Extreme image compression with dual- generative latent fusion. InInt. Conf. Comput. Vis., 2025. 3, 7, 1

    2025

  63. [63]

    On perceptual lossy compression: The cost of perceptual re- construction and an optimal training framework

    Zeyu Yan, Fei Wen, Rendong Ying, Chao Ma, and Peilin Liu. On perceptual lossy compression: The cost of perceptual re- construction and an optimal training framework. InInt. Conf. Mach. Learn., pages 11682–11692, 2021. 2, 3

    2021

  64. [64]

    Lossy image compression with conditional diffusion models

    Ruihan Yang and Stephan Mandt. Lossy image compression with conditional diffusion models. InAdv. Neural Inform. Process. Syst., pages 64971–64995. Curran Associates, Inc.,

  65. [65]

    Mambaic: State space models for high- performance learned image compression

    Fanhu Zeng, Hao Tang, Yihua Shao, Siyu Chen, Ling Shao, and Yan Wang. Mambaic: State space models for high- performance learned image compression. InIEEE Conf. Comput. Vis. Pattern Recog., pages 18041–18050, 2025. 3

    2025

  66. [66]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InIEEE Conf. Comput. Vis. Pattern Recog., pages 586–595, 2018. 1, 2, 5, 7

    2018

  67. [67]

    Is nano banana pro a low-level vision all-rounder? a comprehensive evaluation on 14 tasks and 40 datasets.arXiv preprint arXiv:2512.15110,

    Jialong Zuo, Haoyou Deng, Hanyu Zhou, Jiaxin Zhu, Yicheng Zhang, Yiwei Zhang, Yongxin Yan, Kaixing Huang, Weisen Chen, Yongtai Deng, et al. Is nano banana pro a low-level vision all-rounder? a comprehensive evaluation on 14 tasks and 40 datasets.arXiv preprint arXiv:2512.15110,

    work page arXiv