CoChat
Sign Up
Cited by 15Generative Adversarial Networks and Image SynthesisMultimodal Machine Learning ApplicationsAdvanced Neural Network Applications

The DOI matches its Crossref record; 2 of the 2 references it cites were found in a registry or an index.

Realgeneral: Unifying Visual Generation Via Temporal in-Context Learning With Video Models

Open access
Lin, Yijing, Huang, Mengqi, Zhuang, Shuhan, Mao, Zhendong·2025·2025 IEEE/CVF International Conference on Computer Vision (ICCV)
DOI 10.1109/iccv51701.2025.01391
Open this paper in CoChat — free accountOpen PDF

Abstract

Unifying diverse image generation tasks within a single framework remains a fundamental challenge in visual generation. While large language models (LLMs) achieve unification through task-agnostic data and generation, existing visual generation models fail to meet these principles. Current approaches either rely on per-task datasets and large-scale training or adapt pre-trained image models with task-specific modifications, limiting their generalizability. In this work, we explore video models as a foundation for unified image generation, leveraging their inherent ability to model temporal correlations. We introduce RealGeneral, a novel framework that reformulates image generation as a conditional frame prediction task, analogous to in-context learning in LLMs. To bridge the gap between video models and condition-image pairs, we propose (1) a Unified Conditional Embedding module for multi-modal alignment and (2) a Unified Stream DiT Block with decoupled adaptive LayerNorm and attention mask to mitigate cross-modal interference. RealGeneral demonstrates effectiveness in multiple important visual generation tasks, e.g., it achieves a 14.5% improvement in subject similarity for customized generation and a 10% enhancement in image quality for canny-to-image task. Project page: https://lyne1.github.io/RealGeneral/

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/iccv51701.2025.01391

Cites

0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Image-based customization methods such as RealGeneral, UNO, and DreamO cannot utilize layout conditions, which creates confusion in placing characters.

    “Image-based customiza- tion methods such as RealGeneral [28], UNO [63], and DreamO [35] cannot utilize layout conditions, which creates confusion in placing characters.”

    From DreamingComics: A Story Visualization Pipeline via Subject and Layout Customized Generation using Video Models (2025) · cites this paper as [28]

    The claim says RealGeneral cannot use layout conditions, but the paper evaluates it with spatial conditions: canny edge and depth maps (4.1, p. 6). It reports that RealGeneral follows those maps and generates more controllable images (4.2, p. 7). The paper does not report confusion in placing characters. (LLM, conf=0.86)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

The publisher hasn’t deposited this paper’s reference list.

Research works printed but not linked: 57

  1. Achiam, 2023, Gpt-4 technical report, arXiv preprint, 2
  2. 10.1109/CVPR52733.2024.02157 · DOI
  3. 10.1109/CVPR52729.2023.02161 · DOI · Paper page
  4. Brown, 2020, Language models are few-shot learners, Advances in neural information processing systems, 33, 1877
  5. Chang, 2023, Muse: Text-to-image generation via masked generative transformers, arXiv preprint, 2
  6. Chen, 2024, Omnicreator: Self-supervised unified generation with universal editing, arXiv preprint, 2
  7. Chen, 2024, Customcontrast: A multilevel contrastive perspective for subject-driven text-to-image customization, arXiv preprint, 6
  8. Dhariwal, 2021, Diffusion models beat gans on image synthesis, Advances in neural information processing systems, 34, 8780
  9. 10.1109/CVPR46437.2021.01268 · DOI · Paper page
  10. Gal, 2022, An image is worth one word: Personalizing text-toimage generation using textual inversion, arXiv preprint, 6
  11. Guo, 2023, Animatediff: Animate your personalized text-to-image diffusion models without specific tuning, arXiv preprint, 3
  12. Han, 2024, Ace: Allround creator and editor following instructions via diffusion transformer, arXiv preprint, 3
  13. Heusel, 2017, Gans trained by a two time-scale update rule converge to a local nash equilibrium, Advances in neural information processing systems, 30, 6
  14. Ho, 2020, Denoising diffusion probabilistic models, Advances in neural information processing systems, 33, 6840
  15. Ho, 2022, Cascaded diffusion models for high fidelity image generation, Journal of Machine Learning Research, 23, 1
  16. 10.52202/068431-0628 · DOI
  17. Hong, 2022, Cogvideo: Large-scale pretraining for text-to-video generation via transformers, arXiv preprint, 3
  18. Hu, 2021, Lora: Low-rank adaptation of large language models, arXiv preprint, 6
  19. 10.1145/3503161.3547881 · DOI
  20. 10.1109/CVPR52729.2023.02164 · DOI
  21. 10.1109/CVPR52733.2024.00714 · DOI
  22. 10.1109/CVPR52729.2023.00192 · DOI · Paper page
  23. 2024, Flux: Official inference repository for flux. 1 models
  24. 10.52202/075280-1312 · DOI
  25. 10.52202/068431-2092 · DOI
  26. Lin, 2024, Pixwizard: Versatile image-to-image visual assistant with open-language instructions, arXiv preprint
  27. Liu, 2025, One-prompt-one-story: Free-lunch consistent text-to-image generation using a single prompt, arXiv preprint, 2
  28. Mao, 2024, Realcustom++: Representing images as real-word for real-time customization, arXiv preprint, 3
  29. 10.52202/075280-2544 · DOI
  30. 10.1609/aaai.v38i5.28226 · DOI
  31. 2024, Sora, 3
  32. Oquab, 2023, Dinov2: Learning robust visual features without supervision, arXiv preprint, 6
  33. Podell, 2023, Sdxl: Improving latent diffusion models for high-resolution image synthesis, arXiv preprint
  34. Qin, 2023, Unicontrol: A unified diffusion model for controllable visual generation in the wild, arXiv preprint
  35. Raffel, 2020, Exploring the limits of transfer learning with a unified text-to-text transformer, Journal of machine learning research, 21, 1
  36. Ramesh, Zero-shot text-to-image generation, 8821
  37. Ren, 2024, Grounded sam: Assembling open-world models for diverse visual tasks, arXiv preprint, 6
  38. 10.1109/CVPR52688.2022.01042 · DOI · Paper page
  39. 10.1109/CVPR52729.2023.02155 · DOI · Paper page
  40. Singer, 2022, Make-a-video: Text-to-video generation without text-video data, arXiv preprint, 3
  41. Sun, 2023, Emu: Generative pretraining in multimodality, arXiv preprint, 3
  42. 10.1109/CVPR52733.2024.01365 · DOI · Paper page
  43. Sun, 2024, X-prompt: Towards universal in-context image generation in auto-regressive vision language foundation models, arXiv preprint
  44. Tan, 2024, Ominicontrol: Minimal and universal control for diffusion transformer, arXiv preprint
  45. Villegas, 2022, Phenaki: Variable length video generation from open domain textual description, arXiv preprint, 3
  46. 10.1109/CVPR52729.2023.00660 · DOI
  47. Wang, 2025, MS-diffusion: Multi-subject zero-shot image personalization with layout guidance, 7
  48. Wang, 2024, Lavindit: Large vision diffusion transformer, arXiv preprint, 2
  49. 10.52202/068431-1800 · DOI · Paper page
  50. 10.1109/ICCV51070.2023.01461 · DOI
  51. 10.1109/cvpr52734.2025.01241 · DOI · Paper page
  52. 10.1109/CVPR52733.2024.00829 · DOI
  53. Yang, 2024, Cogvideox: Text-to-video diffusion models with an expert transformer, arXiv preprint
  54. Ye, 2023, Ipadapter: Text compatible image prompt adapter for text-toimage diffusion models, arXiv preprint
  55. Yu, 2024, Anyedit: Mastering unified high-quality image editing for any idea, arXiv preprint, 2
  56. 10.1109/ICCV51070.2023.00355 · DOI · Paper page
  57. 10.1109/CVPR52733.2024.00771 · DOI

Cited by

Sorted by citations

Related

Metadata

Authors
Lin, Y., Huang, M., Zhuang, S., & Mao, Z.
Affiliations
University of Science and Technology of China (CN)
ORCID
Venue
2025 IEEE/CVF International Conference on Computer Vision (ICCV)
Publisher
IEEE
Published
2025-10-19
Pages
14994-15004
DOI
10.1109/iccv51701.2025.01391
OpenAlex
W4416039564
Cited by
1
References
0
Citations by year
2026: 1
FWCI
0.76
Type
conference paper
Language
en
Open access
Licence
Locations
  • IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Funders
National Natural Science Foundation of China (62222212,623B2094)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
Keywords
video models, cross-modal alignment, diffusion models, attention masking

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review