Pith. sign in

REVIEW 4 major objections 4 minor 58 references

Why Compress What You Can Generate? When GPT-4o Generation Ushers in Image Compression Fields

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read GPT-4o generation reconstructs images at 0.001 bpp without training

desk verdict A timely prompt-engineering demo for GPT-4o-based compression whose 'competitive' claim outruns the evidence: 10 tuned images, no error bars, closed web API. read the letter →

arxiv 2504.21814 v1 pith:LZPBNAD7 submitted 2025-04-30 cs.CV

Yixin Gao , Xiaohan Pan , Xin Li , Zhibo Chen This is my paper · ORCID
classification cs.CV
keywords imagecompressionGPT-4ogenerationmultimodaltext-to-imagepromptengineeringultra-lowbitrateraster-scangenerative
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 asks why compress pixels that a powerful generative model could recreate from compact descriptors, and proposes an answer: use GPT-4o's image generation as a training-free decoder. It transmits either a text-only description or a text plus an extremely low-resolution image, and introduces a structural raster-scan prompt that forces the caption to list objects in top-to-bottom, left-to-right order with attention to six visual-consistency dimensions. The authors report that this prompt yields reconstructions that are competitive with or better than trained ultra-low-bitrate codecs on perceptual metrics such as CLIP-IQA and MUSIQ, while keeping semantic consistency measured by CLIPSIM, at bitrates near 0.001-0.006 bpp. The upshot is that the bottleneck of image compression may shift from pixel transform coding to prompt design and generative-model capability.

What carries the argument

The central object is the structural raster-scan prompt, a captioning instruction that asks the multimodal language model to enumerate the main items of the image from top to bottom and left to right, then to refine the description along six named dimensions: feature correspondence, geometric consistency, photometric consistency, style consistency, semantic consistency, and structural integrity, while keeping the response within a fixed word limit. Its role is to convert spatial arrangement into textual form so that GPT-4o's decoder can reconstruct an image whose layout matches the original, countering the consistency failure that afflicts generic captions. The pipeline also leans on a visual-condition branch—a factor-8 downsampled image compressed by MS-ILLM—and Lempel-Ziv lossless coding of the text; together these carry the rate while GPT-4o supplies all pixel-level generation.

What would settle it

Run the pipeline on the full DIV2K validation set or Kodak, with several repeated decodes and a fixed GPT-4o version, and compare at matched bitrates; if the raster-scan prompt no longer improves CLIPSIM/DISTS over a generic caption of equal length, or the metric advantages over Text+Sketch and PerCo disappear, the central claim is falsified.

Watch

Extended reading notes

Core claim

In its own terms, the paper claims that a large multimodal model used for generation can replace most of the reconstruction machinery of a codec. The authors test two paradigms: textual coding, where only a losslessly compressed caption is sent, and multimodal coding, where the caption is joined by an image downsampled by a factor of eight and compressed with MS-ILLM. They claim the structural raster-scan prompt, which describes main items in spatial scan order and elaborates feature, geometric, photometric, style, semantic, and structural-integrity dimensions within a strict word budget, is what preserves layout and consistency. Their experiments on the first ten DIV2K validation images report higher CLIP-IQA and MUSIQ scores than MS-ILLM, Text+Sketch, and PerCo at comparable or lower bitrates, and best CLIPSIM among the compared methods, all with no training of the generative model. The paper's conclusion is that GPT-4o's generation capability is sufficient to bring competitive ultra-low-bitrate compression, so the field should consider generation-based coding as a viable paradigm.

Load-bearing premise

The paper's competitive-performance claim rests on three-run averages over the first ten DIV2K validation images, with the prompt variants tuned on that same set and GPT-4o accessed through a closed web interface whose version and behavior the authors do not control.

Editorial extensions

If this is right

  • If the claim holds, ultra-low-bitrate compression can be achieved without training or fine-tuning any pixel codec; the transmitted rate consists of a compressed caption plus an optional tiny image.
  • The structural raster-scan prompt is a transferable prompt-engineering mechanism: any text-to-image model that follows spatial ordering can use it to improve layout consistency.
  • Multimodal coding (short text plus low-resolution image) gives the best balance of perceptual quality and consistency, and increasing prompt length past a point degrades structural consistency.
  • Because GPT-4o decoding is stochastic, the same transmitted bitstream can produce several plausible reconstructions, and adding the visual condition reduces this randomness.

Reading between the lines

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

  • A likely next test is whether the same prompting recipe transfers to open-weight models; if an open model matches GPT-4o, the paradigm becomes deployable without a closed API.
  • A fair rate-distortion comparison would count the compute or service cost of generation, not just transmitted bits; including that could change which operating points are actually advantageous.
  • The raster-scan ordering is a form of serialization, so the same idea may extend to diagrams, medical images, or video keyframes where spatial layout matters as much as semantics.
  • Re-running on full validation sets with model versions frozen is the natural stress test; the reported 10-image, 3-run averages are the paper's evidence but not a bound on the method.
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

4 major / 4 minor

Summary. This paper proposes an ultra-low-bitrate image compression framework based on GPT-4o image generation. It introduces a structural raster-scan prompt that describes image content in a top-to-bottom, left-to-right order with six consistency dimensions, and evaluates two paradigms: text-only and text-plus-low-resolution-image conditioning. The method is compared against MS-ILLM, Text+Sketch, and PerCo on the first ten DIV2K validation images using CLIP-IQA, MUSIQ, CLIPSIM, and DISTS. The authors report that their method achieves competitive performance at ultra-low bitrates without any additional training.

Significance. If the claims hold, the paper could open a new training-free direction for extreme image compression based on large multimodal generation models, and the raster-scan prompt mechanism is a simple, transferable design idea. The paper also provides ablation studies on prompt structure and length and qualitative results that illustrate consistency improvements. However, the evidence base is currently too small and uncontrolled to support the stated claims of 'impressive' and 'competitive' performance, and the use of a closed web interface raises reproducibility concerns.

major comments (4)
  1. [§4.1 Datasets and Evaluation; Table 1] The quantitative evaluation is based on only the first 10 images of the DIV2K validation set, with each result averaged over three GPT-4o generations. No standard deviations, confidence intervals, or per-image results are reported. Given that the central claim of 'competitive performance' is read directly from the curves in Fig. 3 and the numbers in Table 1, the reported differences, e.g., CLIPSIM 0.901 vs. 0.888 in the ablation, could easily lie within run-to-run and image-to-image variance. I request a larger evaluation set, more independent runs, and at least per-image breakdowns or error bars before the headline claim can be supported.
  2. [§4.3 Ablations; §4.1 Datasets and Evaluation] The prompt length limits (30 words for text-only, 15 words for text+image in the final configuration) and the design of the raster-scan prompt's six dimensions appear to have been selected after inspecting metrics on the same 10 DIV2K images used for the final comparisons (Figs. 3, 4, and Table 1). This selection-on-the-test-set procedure risks overfitting the reported results to this particular set. The authors should either report results on a held-out set or explicitly describe the selection procedure and show that the chosen hyperparameters are not the result of cherry-picking.
  3. [§4.1 Implementation details] GPT-4o is accessed through a closed web interface using automation scripts, so the model version, sampling parameters, and any server-side updates are not under the authors' control. Three-run averages over this stochastic, moving target are not reproducible, and the reported numbers may change if the service is updated. The manuscript should at minimum pin the exact access date, model version, and interface, and ideally evaluate with a locally runnable generative model (e.g., an open-weight model) to verify that the main claims do not depend on the specific proprietary service.
  4. [§4.2 Main Results; Fig. 3] The bitrate comparison is not tightly controlled: the points for 'Ours' and the baselines in Fig. 3 are at different rate points, and the ablation in Table 1 compares Ours (bpp 0.0008) with Ours w/o structure prompt (bpp 0.0007), which are not at identical rates. To fairly claim superior rate-distortion performance, the methods should be evaluated at matched anchor bitrates or compared via interpolation, and the rate variation across runs should be reported.
minor comments (4)
  1. [§2.1 Related Works] References [36] and [24] are cited as 'Mao' and 'Jia' without 'et al.'; please correct the citation format.
  2. [Figure 5 caption] The caption uses 'PICS [29]' and 'PIC [29]' inconsistently; these should be 'Text+Sketch [29]' for clarity.
  3. [Table 1] Given the small sample size and stochastic generation, reporting bitrates to four decimal places (e.g., 0.0008 vs 0.0007) implies a precision the experiment does not support; report fewer decimal places or include standard deviations.
  4. [§4.2 Main Results] The statement 'our method consistently achieves higher scores than existing approaches at similar or even lower bitrates' is too strong in the absence of error bars; it should be softened to 'tends to achieve' or accompanied by statistical significance testing.

Circularity Check

1 steps flagged · score 4.0 of 10

The reported 'Ours' performance is partly constructed by selecting prompt-length limits on the same 10 DIV2K images used for the final comparison; no other circular derivation is present.

  1. fitted input called prediction [Section 4.3 'Prompt Length' (Figs. 3, 4); Section 4.1 'Datasets and Evaluation']
    "we use the first 10 images from the DIV2K [2] validation set... all experiments are repeated three times, and the average is reported... For only textual coding, as shown in Fig. 3, we test on 15, 30 and 120 words. The word number represents the upper limit of caption length. We observe that from 15 to 30 words, all metrics improve significantly. However, further increasing the number of words does not lead to noticeable gains, suggesting that there is a saturation point beyond which additional textual information brings limited benefit."

    The word-count limit is a cost-controlling hyperparameter of the proposed method. The paper selects it by inspecting CLIPSIM/DISTS/CLIP-IQA/MUSIQ on the same first-10 DIV2K images on which the 'Ours' points in Fig. 3, Fig. 5, and Table 1 are then reported. The reported 'competitive performance' is thus the value of the best-performing variant on the evaluation set, i.e., an argmax over prompt lengths on the same data, not a prediction of a fixed method on unseen images. With 10 images and 3 random generations, this selection is statistically unstable. The structure-vs-no-structure comparison retains independent content, so the circularity is partial.

full rationale

The paper contains no analytic derivation; its claims are empirical evaluations of a prompt-engineering scheme on GPT-4o. The central novelty, the structural raster-scan prompt, is tested by a direct ablation (Table 1, Fig. 8) that is not circular: the prompt is not defined in terms of the evaluation metrics, and the comparison against a no-structure prompt is meaningful. Self-citations (e.g., [17] for the compressed-image visual prior) are not load-bearing because the design choice is also supported by external references [18,23,30]. The only circular element is the prompt-length selection: 15/30/120 and 0/15/60 word variants are tested and the best-performing configuration is reported as 'Ours' on the same 10-image set, so the headline quantitative conclusion is partly built from selection on the evaluation data rather than from an independent prediction. This warrants a moderate circularity score, not a higher one, because the structure prompt itself and the comparison to external baselines still carry independent evidence.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper introduces no mathematical free-fitting, but it does select prompt word limits and visual downsampling on the evaluation set, and assumes evaluation metrics and GPT-4o stability are valid. No invented entities are introduced.

free parameters (2)
  • Prompt word limit (text-only: 30, text+image: 15) = 30 / 15 words
    Chosen based on ablation on the same 10-image evaluation set (Figs. 3 and 4); reported as the method's setting, not held-out.
  • Downsample factor for visual condition = 8
    Adopted from prior works (following [17,18,23,30]) without dedicated tuning; affects the bitrate/quality tradeoff of the visual branch.
assumptions (5)
  • domain assumption The four selected metrics (CLIP-IQA, MUSIQ, CLIPSIM, DISTS) are valid proxies for perceptual quality and consistency.
    Used to rank methods; the paper does not validate that these metrics agree with human judgment for generated images.
  • domain assumption The first 10 DIV2K validation images are representative of natural images for ultra-low bitrate compression.
    All experiments, ablations, and comparisons use these 10 images; no statistical analysis across diverse content is provided.
  • domain assumption GPT-4o generation through the web interface is stable enough that 3-run averages yield meaningful comparisons.
    The paper notes randomness and repeats runs (Section 4.1) but treats averaged scores as deterministic without variance.
  • standard math Lempel-Ziv (zlib) is an appropriate lossless entropy coder for the text side information, fairly capturing its bitrate.
    Text is compressed with zlib, consistent with prior works [29, 9]; no arithmetic coding or custom entropy model is used.
  • domain assumption The MS-ILLM codec at the chosen quality level provides a fair visual condition for the text+image branch.
    MS-ILLM is one design choice; other visual priors (sketches, VQ latents) could yield different tradeoffs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Why Compress What You Can Generate? When GPT-4o Generation Ushers in Image Compression Fields." pith.science (2026). https://pith.science/paper/LZPBNAD7

@misc{pith2026250421814,
  author       = {Pith},
  title        = {Pith review of: Why Compress What You Can Generate? When GPT-4o Generation Ushers in Image Compression Fields},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LZPBNAD7}},
  note         = {Machine review of arXiv:2504.21814}
}
read the original abstract

The rapid development of AIGC foundation models has revolutionized the paradigm of image compression, which paves the way for the abandonment of most pixel-level transform and coding, compelling us to ask: why compress what you can generate if the AIGC foundation model is powerful enough to faithfully generate intricate structure and fine-grained details from nothing more than some compact descriptors, i.e., texts, or cues. Fortunately, recent GPT-4o image generation of OpenAI has achieved impressive cross-modality generation, editing, and design capabilities, which motivates us to answer the above question by exploring its potential in image compression fields. In this work, we investigate two typical compression paradigms: textual coding and multimodal coding (i.e., text + extremely low-resolution image), where all/most pixel-level information is generated instead of compressing via the advanced GPT-4o image generation function. The essential challenge lies in how to maintain semantic and structure consistency during the decoding process. To overcome this, we propose a structure raster-scan prompt engineering mechanism to transform the image into textual space, which is compressed as the condition of GPT-4o image generation. Extensive experiments have shown that the combination of our designed structural raster-scan prompts and GPT-4o's image generation function achieved the impressive performance compared with recent multimodal/generative image compression at ultra-low bitrate, further indicating the potential of AIGC generation in image compression fields.

Figures

Figures reproduced from arXiv: 2504.21814 by the authors.

Figure 1
Figure 1. Overall pipeline of our multimodal image compression framework based on GPT-4o image generation. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Our proposed structural raster-scan prompt. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Tradeoffs between bitrate and different metrics on DIV2K. The quality is evaluated by both perceptual ( [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Ablation on prompt length. 4.2. Main Results Quantitative Results In [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: The comparison of qualitative results between the baselines and ours (text+image) method is provided, with the bits per pixel [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Comparison of results generated using only text. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: Image Comparison for Ablation Study of Structure [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: We present results generated multiple times at the de [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 27 canonical work pages

  1. [1]

    https:// docs.python.org/3/library/zlib.html

    zlib — compression compatible with gzip. https:// docs.python.org/3/library/zlib.html. 4

    work page

  2. [2]

    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. In Pro- ceedings of the IEEE conference on computer vision and pat- tern recognition workshops, pages 126–135, 2017. 4

    work page 2017

  3. [3]

    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. In ICLR, 2018. 1

    2018

  4. [4]

    Nonlinear transform coding

    Johannes Ball ´e, Philip A Chou, David Minnen, Saurabh Singh, Nick Johnston, Eirikur Agustsson, Sung Jin Hwang, and George Toderici. Nonlinear transform coding. IEEE Journal of Selected Topics in Signal Processing, 15(2):339– 353, 2020. 1

    work page 2020

  5. [5]

    Better portable graphics (bpg) image for- 7 mat

    Fabrice Bellard. Better portable graphics (bpg) image for- 7 mat. https://bellard.org/bpg/. Accessed: 2024- 05-22. 1

    work page 2024

  6. [6]

    Improving image generation with better captions

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, Long Ouyang, Juntang Zhuang, Joyce Lee, Yufei Guo, et al. Improving image generation with better captions. Computer Science. https://cdn. openai. com/papers/dall-e-3. pdf, 2(3):8, 2023. 2

    2023

  7. [7]

    Improving image generation with better captions

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, Long Ouyang, Juntang Zhuang, Joyce Lee, Yufei Guo, et al. Improving image generation with better captions. Computer Science. https://cdn. openai. com/papers/dall-e-3. pdf, 2(3):8, 2023. 3

    2023

  8. [8]

    Overview of the versatile video coding (vvc) standard and its applications

    Benjamin Bross, Ye-Kui Wang, Yan Ye, Shan Liu, Jianle Chen, Gary J Sullivan, and Jens-Rainer Ohm. Overview of the versatile video coding (vvc) standard and its applications. TCSVT, 2021. 1

    work page 2021

Show all 58 references
  1. [9]

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

    Marl `ene Careil, Matthew J Muckley, Jakob Verbeek, and St´ephane Lathuili`ere. Towards image compression with per- fect realism at ultra-low bitrates. InThe Twelfth International Conference on Learning Representations, 2023. 1, 2, 4, 6, 10

    2023

  2. [10]

    Learned image compression with discretized gaussian mixture likelihoods and attention modules

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto. Learned image compression with discretized gaussian mixture likelihoods and attention modules. In CVPR, pages 7939–7948, 2020. 1

    2020

  3. [11]

    Image quality assessment: Unifying structure and texture similarity

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simoncelli. Image quality assessment: Unifying structure and texture similarity. IEEE transactions on pattern analysis and ma- chine intelligence, 44(5):2567–2581, 2020. 4

    2020

  4. [12]

    Diffusion self-guidance for control- lable image generation

    Dave Epstein, Allan Jabri, Ben Poole, Alexei Efros, and Aleksander Holynski. Diffusion self-guidance for control- lable image generation. Advances in Neural Information Processing Systems, 36:16222–16239, 2023. 2

    2023

  5. [13]

    Taming transformers for high-resolution image synthesis

    Patrick Esser, Robin Rombach, and Bjorn Ommer. Taming transformers for high-resolution image synthesis. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12873–12883, 2021. 1, 2

    2021

  6. [14]

    Scaling recti- fied 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, et al. Scaling recti- fied flow transformers for high-resolution image synthesis. In Forty-first international conference on mach...

  7. [15]

    Dit4edit: Dif- fusion transformer for image editing

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

    2025

  8. [16]

    Nvtc: Nonlinear vector transform coding

    Runsen Feng, Zongyu Guo, Weiping Li, and Zhibo Chen. Nvtc: Nonlinear vector transform coding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6101–6110, 2023. 1

    2023

  9. [17]

    Unimic: To- wards universal multi-modality perceptual image compres- sion

    Yixin Gao, Xin Li, Xiaohan Pan, Runsen Feng, Zongyu Guo, Yiting Lu, Yulin Ren, and Zhibo Chen. Unimic: To- wards universal multi-modality perceptual image compres- sion. arXiv preprint arXiv:2412.04912, 2024. 3

    2024 arXiv

  10. [18]

    A residual diffusion model for high perceptual quality codec augmentation

    Noor Fathima Ghouse, Jens Petersen, Auke Wiggers, Tianlin Xu, and Guillaume Sautiere. A residual diffusion model for high perceptual quality codec augmentation. arXiv preprint arXiv:2301.05489, 2023. 3

    2023 arXiv

  11. [19]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. NeurIPS, 27,

  12. [20]

    Causal contextual prediction for learned image com- pression

    Zongyu Guo, Zhizheng Zhang, Runsen Feng, and Zhibo Chen. Causal contextual prediction for learned image com- pression. TCSVT, 32(4):2329–2341, 2021. 1

    2021

  13. [21]

    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. In CVPR, pages 5718–5727, 2022. 1

    2022

  14. [22]

    Denoising dif- fusion probabilistic models

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

    2020

  15. [23]

    High- fidelity image compression with score-based generative models

    Emiel Hoogeboom, Eirikur Agustsson, Fabian Mentzer, Luca Versari, George Toderici, and Lucas Theis. High- fidelity image compression with score-based generative models. arXiv preprint arXiv:2305.18231, 2023. 3

    2023 arXiv

  16. [24]

    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. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 26088– 26098, 2024. 1, 2

    2024

  17. [25]

    Imagic: Text-based real image editing with diffusion models

    Bahjat Kawar, Shiran Zada, Oran Lang, Omer Tov, Huiwen Chang, Tali Dekel, Inbar Mosseri, and Michal Irani. Imagic: Text-based real image editing with diffusion models. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6007–6017, 2023. 2

    2023

  18. [26]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pages 5148–5157, 2021. 4

    2021

  19. [27]

    Perco (SD): Open perceptual compression

    Nikolai K ¨orber, Eduard Kromer, Andreas Siebert, Sascha Hauke, Daniel Mueller-Gritschneder, and Bj ¨orn Schuller. Perco (SD): Open perceptual compression. In Workshop on Machine Learning and Compression, NeurIPS 2024 , 2024. 10

    2024

  20. [28]

    Black Forest Labs. Flux. https://github.com/ black-forest-labs/flux, 2024. 2, 3

    2024

  21. [29]

    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. In ICML 2023 Workshop Neural Compres- sion: From Information Theory to Applications , 2023. 1, 2, 3, 4, 5, 6, 10

    2023

  22. [30]

    Misc: Ultra-low bitrate image semantic com- pression driven by large multimodal model

    Chunyi Li, Guo Lu, Donghui Feng, Haoning Wu, Zicheng Zhang, Xiaohong Liu, Guangtao Zhai, Weisi Lin, and Wen- jun Zhang. Misc: Ultra-low bitrate image semantic com- pression driven by large multimodal model. arXiv preprint arXiv:2402.16749, 2024. 1, 2, 3

    2024 arXiv

  23. [31]

    Task-driven semantic cod- ing via reinforcement learning

    Xin Li, Jun Shi, and Zhibo Chen. Task-driven semantic cod- ing via reinforcement learning. TIP, 2021. 1

    2021

  24. [32]

    Diffusion models for image restoration and enhancement–a compre- hensive survey

    Xin Li, Yulin Ren, Xin Jin, Cuiling Lan, Xingrui Wang, Wenjun Zeng, Xinchao Wang, and Zhibo Chen. Diffusion models for image restoration and enhancement–a compre- hensive survey. arXiv preprint arXiv:2308.09388, 2023. 2 8

    2023

  25. [33]

    Towards extreme image compression with latent feature guidance and diffusion prior

    Zhiyuan Li, Yanhui Zhou, Hao Wei, Chenyang Ge, and Jing- wen Jiang. Towards extreme image compression with latent feature guidance and diffusion prior. IEEE Transactions on Circuits and Systems for Video Technology, 2024. 1, 2

    2024

  26. [34]

    Diffusion-based extreme image compres- sion with compressed feature initialization

    Zhiyuan Li, Yanhui Zhou, Hao Wei, Chenyang Ge, and Ajmal Mian. Diffusion-based extreme image compres- sion with compressed feature initialization. arXiv preprint arXiv:2410.02640, 2024. 2

    2024 arXiv

  27. [35]

    Learned image compression with mixed transformer-cnn architectures

    Jinming Liu, Heming Sun, and Jiro Katto. Learned image compression with mixed transformer-cnn architectures. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 14388–14397, 2023. 1

    2023

  28. [36]

    Extreme im- age compression using fine-tuned vqgans

    Qi Mao, Tinghan Yang, Yinuo Zhang, Zijian Wang, Meng Wang, Shiqi Wang, Libiao Jin, and Siwei Ma. Extreme im- age compression using fine-tuned vqgans. In 2024 Data Compression Conference (DCC) , pages 203–212. IEEE,

    2024

  29. [37]

    Channel-wise autoregres- sive entropy models for learned image compression

    David Minnen and Saurabh Singh. Channel-wise autoregres- sive entropy models for learned image compression. In 2020 IEEE International Conference on Image Processing (ICIP), pages 3339–3343. IEEE, 2020. 1

    2020

  30. [38]

    Joint autoregressive and hierarchical priors for learned image compression

    David Minnen, Johannes Ball ´e, and George Toderici. Joint autoregressive and hierarchical priors for learned image compression. In NeurIPS, 2018. 1

    2018

  31. [39]

    Null-text inversion for editing real im- ages using guided diffusion models

    Ron Mokady, Amir Hertz, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Null-text inversion for editing real im- ages using guided diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 6038–6047, 2023. 2

    2023

  32. [40]

    T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In Proceedings of the AAAI conference on artificial intelligence, pages 4296–4304, 2024. 2

    2024

  33. [41]

    Neuralcompres- sion

    Matthew Muckley, Jordan Juravsky, Daniel Severo, Mannat Singh, Quentin Duval, and Karen Ullrich. Neuralcompres- sion. https://github.com/facebookresearch/ NeuralCompression, 2021. 10

    2021

  34. [42]

    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. In International Conference on Machine Learning, pages 25426–25443. PMLR, 2023. 3, 4, 6, 10

    2023

  35. [43]

    Addendum to gpt-4o system card: 4o image gener- ation, 2025

    OpenAI. Addendum to gpt-4o system card: 4o image gener- ation, 2025. Accessed: 2025-04-22. 1, 3

    2025

  36. [44]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  37. [45]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1, 2, 4

    2022

  38. [46]

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

    Yujun Shi, Chuhui Xue, Jun Hao Liew, Jiachun Pan, Han- shu Yan, Wenqing Zhang, Vincent YF Tan, and Song Bai. Dragdiffusion: Harnessing diffusion models for interactive point-based image editing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogniti...

    2024

  39. [47]

    Score-based generative modeling through stochastic differential equa- tions

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions. arXiv preprint arXiv:2011.13456, 2020. 2

    2011 arXiv

  40. [48]

    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. InPro- ceedings of the AAAI conference on artificial intelligence , pages 2555–2563, 2023. 4

    2023

  41. [49]

    Stylediffusion: Controllable disentangled style transfer via diffusion models

    Zhizhong Wang, Lei Zhao, and Wei Xing. Stylediffusion: Controllable disentangled style transfer via diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 7677–7689, 2023. 2

    2023

  42. [50]

    Learned block-based hybrid image compression

    Yaojun Wu, Xin Li, Zhizheng Zhang, Xin Jin, and Zhibo Chen. Learned block-based hybrid image compression. IEEE Transactions on Circuits and Systems for Video Tech- nology, 32(6):3978–3990, 2021. 1

    2021

  43. [51]

    Sana 1.5: Efficient scaling of training-time and inference-time compute in linear diffusion transformer

    Enze Xie, Junsong Chen, Yuyang Zhao, Jincheng Yu, Ligeng Zhu, Yujun Lin, Zhekai Zhang, Muyang Li, Junyu Chen, Han Cai, et al. Sana 1.5: Efficient scaling of training-time and inference-time compute in linear diffusion transformer. arXiv preprint arXiv:2501.18427, 2025. 2

    2025 arXiv

  44. [52]

    Unifying generation and compression: Ultra-low bi- trate image coding via multi-stage transformer

    Naifu Xue, Qi Mao, Zijian Wang, Yuan Zhang, and Siwei Ma. Unifying generation and compression: Ultra-low bi- trate image coding via multi-stage transformer. In2024 IEEE International Conference on Multimedia and Expo (ICME) , pages 1–6. IEEE, 2024. 1, 2

    2024

  45. [53]

    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. arXiv preprint arXiv:2503.01428 ,

  46. [54]

    Gpt-imgeval: A comprehensive benchmark for diagnosing gpt4o in image generation

    Zhiyuan Yan, Junyan Ye, Weijia Li, Zilong Huang, Sheng- hai Yuan, Xiangyang He, Kaiqing Lin, Jun He, Conghui He, and Li Yuan. Gpt-imgeval: A comprehensive benchmark for diagnosing gpt4o in image generation. arXiv preprint arXiv:2504.02782, 2025. 4

    2025 arXiv

  47. [55]

    Lossy image compression with conditional diffusion models

    Ruihan Yang and Stephan Mandt. Lossy image compression with conditional diffusion models. Advances in Neural In- formation Processing Systems, 36, 2024. 4

    2024

  48. [56]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 3836–3847, 2023. 2

    2023

  49. [57]

    Inversion-based style transfer with diffusion models

    Yuxin Zhang, Nisha Huang, Fan Tang, Haibin Huang, Chongyang Ma, Weiming Dong, and Changsheng Xu. Inversion-based style transfer with diffusion models. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10146–10156, 2023. 2

    2023

  50. [58]

    Artbank: Artistic style transfer with pre-trained diffusion model and implicit style prompt bank

    Zhanjie Zhang, Quanwei Zhang, Wei Xing, Guangyuan Li, Lei Zhao, Jiakai Sun, Zehua Lan, Junsheng Luan, Yiling Huang, and Huaizhong Lin. Artbank: Artistic style transfer with pre-trained diffusion model and implicit style prompt bank. In Proceedings of the AAAI Conference on Art...

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.