Pith. sign in

REVIEW 3 major objections 5 minor 47 references

SerialGen: Personalized Image Generation by First Standardization Then Personalization

T0 review · 3 major / 5 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 Standardizing the reference image before personalization lifts whole-body appearance consistency by 5.49 CLIP-I points without sacrificing text controllability.

desk verdict The serial standardization idea is worth taking seriously, but the key ablation in Table 2 is confounded: the paired one-stage baseline trains on targets produced by the proposed two-stage model, so the CLIP-I gap may be target quality rather than standardization. read the letter →

arxiv 2412.01485 v2 pith:PJ5TOOCR submitted 2024-12-02 cs.CV

Cong Xie , Han Zou , Ruiqi Yu , Yan Zhang , Zhenpeng Zhan This is my paper · ORCID
classification cs.CV
keywords personalizedimagegenerationtext-to-imagediffusiontuning-freepersonalizationappearanceconsistencytextcontrollabilityreferencestandardizationserial
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

This paper claims that the usual trade-off between text controllability and appearance consistency in tuning-free personalized image generation can be broken by a simple reordering: standardize the reference image first, then train the personalization model on (standardized reference, target) pairs. The two-stage SerialGen pipeline first turns any reference photo into a canonical pose with white background and neutral expression via a model trained on synthetic 3D renders, then uses those standardized references to train a reference-encoder diffusion model. Experiments report a 5.49-point gain in CLIP-I over the same architecture trained on paired but non-standardized references, while text controllability (CLIP-T) is essentially unchanged, and a 5.43-point gain in pairwise consistency across serial outputs. If true, this is a practical fix for the replication-versus-controllability problem: the standardized reference gives the model a stable appearance anchor without letting it copy the target image.

What carries the argument

The load-bearing object is the standardization model, a frozen human-image-animation network that maps any reference image to a canonical appearance-preserving template: standard pose, white background, neutral expression, and fixed face position. It is trained on synthetically rendered pairs from 3D character models, with two added modules: the foreground-background distinction module injects learnable foreground/background class tokens into self-attention, and the reference pose injection module adds the reference's pose features into the same attention layers. In the second stage, this standardized reference is fed through the reference encoder of a diffusion-based personalization model trained on roughly 300,000 (standardized reference, target) pairs. Standardization is what does the work: it removes non-appearance elements that otherwise give the model an easy replication shortcut, while pre-generating body parts not visible in the original reference (for example, when the reference is head-only).

What would settle it

Train the same two-stage pipeline with a standardization model trained on real paired photos rather than synthetic renders and check whether the CLIP-I advantage over the paired one-stage baseline survives; if it shrinks or disappears, the synthetic style bias, not standardization itself, is doing the work.

Watch

Extended reading notes

Core claim

The central claim is that using a standardized reference—same appearance, but with pose, background, expression, and viewpoint normalized—as the conditioning input to a personalized text-to-image model resolves the replication problem that arises when reference and target images are nearly identical. The authors show that training on (standardized reference, target) pairs produced by a frozen standardization model yields higher whole-body appearance consistency (CLIP-I 85.49) than training on unstandardized but otherwise paired data (CLIP-I 80.00), while text controllability remains comparable (CLIP-T 21.76 vs 21.99). They further report that standardization increases consistency across serial images generated from different prompts (AP-CLIP-I 83.17 vs 77.74). The standardization model itself is an image-animation backbone trained on 10.5 million synthetic (non-standardized, standardized) pairs rendered from 3D characters, augmented with two modules—a foreground-background distinction module and a reference pose injection module—that improve its output on PSNR/SSIM and on downstream generation metrics.

Load-bearing premise

The standardization model is trained entirely on synthetic 3D character renders, and the paper assumes that applying it to real photos preserves the person's appearance well enough that the resulting (standardized reference, target) pairs are valid training data; the paper concedes a 3D style bias in standardized images and asserts, without a dedicated experiment, that this bias does not impede personalization.

Editorial extensions

If this is right

  • The unpaired one-stage model in the paper replicates the reference, scoring CLIP-I 89.62 but CLIP-T only 16.08; the two-stage model keeps CLIP-T at 21.76 while still scoring CLIP-I 85.49, so standardization removes the replication shortcut without costing text control.
  • When the reference shows only a head, the standardized reference pre-generates the body in a canonical pose, so serial images generated from different prompts share a consistent body appearance; pairwise CLIP-I across serial outputs rises from 77.74 to 83.17.
  • The 3D style bias introduced by synthetic-only training is explicitly mitigated in the personalization stage, so the final outputs recover realistic appearance while retaining the standardization benefit.
  • The foreground-background distinction module and the reference pose injection module both improve the standardization model's output, and using them in the full pipeline raises CLIP-I, CLIP-T, and AP-CLIP-I over the baseline without them.

Reading between the lines

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

  • The same standardization-then-personalization recipe could generalize to non-human subjects: the supplementary shows serial images of non-human characters, and the standardization model is appearance-agnostic, so extending it to animals or objects would test whether the appearance anchor is the general mechanism.
  • Standardization effectively manufactures paired training data from any unpaired corpus of reference images, so it could serve as a general data-augmentation strategy for reference-conditioned generation beyond human characters.
  • The reported 5.49-point CLIP-I gain may understate the benefit in cases where the reference is head-only, because the body-consistency improvement in those cases is qualitative and not fully reflected in whole-image CLIP-I after background removal.
  • A focused test of the mechanism would be to feed the standardized reference directly into a zero-shot image animation pipeline (skipping the personalization stage) and compare whether the appearance consistency across prompts is already comparable; this would isolate the contribution of standardization from the contribution of the second-stage training.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. SerialGen is a tuning-free personalized image generation framework that operates in two stages. A standardization model, built on a human-image-animation architecture with two new modules (FBDM and RPIM), is trained on ~10.5M synthetic pairs rendered from 3D characters to map a reference photo to a standardized view with fixed pose, white background, neutral expression, and centered face. A personalization model (IP-Adapter with SDXL and a CLIP image encoder) is then trained on ~300k pairs (standardized reference, original real image), and at inference the reference is standardized before personalization. The paper reports CLIP-I/CLIP-T/Face Sim improvements over IP-Adapter, FastComposer, and StoryMaker, a user study, ablations against one-stage unpaired and paired training, and ablations for the two standardization modules.

Significance. If the central comparison were clean, the paper would offer a practical resolution of the replication-versus-controllability trade-off in tuning-free personalized generation, and the serial training scheme would be easy to build on. The paper has strengths: it tests against several baselines, includes a user study, gives detailed training and inference settings, and explicitly analyzes the failure mode of same-image unpaired training as a function of encoder capacity. However, the key evidence for 'standardize first'—the comparison to paired one-stage training—is confounded by a difference in target supervision, and the synthetic-to-real behavior of the standardization model is not directly measured. The contribution is therefore promising but not yet established.

major comments (3)
  1. [Sec. 4.3.1, Table 2] The comparison that supports the central claim is confounded. The paired one-stage baseline is trained on pairs 'generated by applying our two-stage model to dataset D,' meaning its target images are outputs of the proposed model, while the two-stage model is trained on targets that are the original real images. The 5.49 CLIP-I gap (80.00 vs 85.49) may therefore reflect differences in target-image quality and distribution rather than the benefit of standardized references; the same confound applies to the AP-CLIP-I comparison (77.74 vs 83.17). Please retrain the paired one-stage baseline on pairs (x_r, x_t) in which both images come from the same source distribution (e.g., real multi-view or video frames of the same character), or use identical target images for both conditions, and report the prompts and filtering used to create the current baseline's targets. Without this, the paper's main claim is not supported.
  2. [Sec. 3.4 / Sec. 4.3.2, Table 3] The standardization model's ability to preserve whole-body appearance on real photographs is asserted rather than measured. Section 3.4 concedes the output is 'biased towards the 3D style used during training,' while the quantitative standardization evaluation in Table 3 is performed on a synthetic test set (a character-ID split), not on real images. Because the entire two-stage pipeline depends on phi(x) being a faithful appearance-preserving normalization of real photos, please report a direct metric on real data (e.g., CLIP-I and Face Sim. between original and standardized real references, ideally with a human evaluation) and show representative success and failure cases. This is especially important because the main comparison in Table 2 currently cannot disentangle standardization quality from target supervision quality.
  3. [Sec. 4.3.2, Tables 3-4] The contribution of the two proposed modules is not strongly established. Table 4 shows improvements of only 0.25 CLIP-I, 0.14 CLIP-T, and 0.29 AP-CLIP-I when adding FBDM and RPIM, with no variance or significance testing across random seeds, and Table 3 evaluates on synthetic data only. Given the small effect sizes, the modules should be validated with multiple seeds and, ideally, on the real-image test set used in Table 2.
minor comments (5)
  1. [Sec. 4.3.1] The label 'unpaired one-stage' is confusing: the setup described uses identical images as reference and target, which is a same-image regression setup rather than unpaired data in the usual sense; please rename or clarify.
  2. [Table 4] The 'with' row reports CLIP-I 85.16, while Tables 1 and 2 report Ours as 85.49 for what appears to be the same configuration; please reconcile the numbers or state that they come from different runs or test subsets.
  3. [All quantitative tables] The paper reports no confidence intervals or variance across random seeds; for the small module ablations this is particularly limiting, so please add error bars or at least a multi-seed analysis.
  4. [Sec. 4.2] Please clarify whether the Face Sim. metric is computed on face crops or on full images after background removal, since the description of background removal is given only for CLIP-I.
  5. [References] Reference [28] is cited both for the DisCo method and for the TikTok benchmark dataset; consider adding a separate citation for the dataset itself.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SerialGen's central claims are empirical comparisons against external baselines; the only self-referential element is a confounded ablation, not a derivation.

full rationale

SerialGen does not derive its central results from its own assumptions by construction. The two-stage personalization model is trained on (standardized reference, target) pairs with loss Eq. (5), and the standardization model is trained on held-out synthetic renders with objective PSNR/SSIM (Table 3); these are ordinary supervised objectives, not tautologies. The headline claims are supported by comparisons to external methods IP-Adapter, FastComposer, StoryMaker (Table 1) and to LCM-Lookahead, plus a TikTok benchmark (Table 6), so the core evidence is self-contained against external systems. The manuscript itself flags in Section 3.4 that standardization introduces a 3D-style bias, and Section 4.3.1 states that the paired one-stage baseline's pairs 'are generated by applying our two-stage model to dataset D.' That means the Table 2 ablation changes both the reference standardization and the target supervision source, so the 5.49 CLIP-I gap is not a clean isolation of standardization; this is a validity/confound concern, not a circular one, because the comparison is an empirical result rather than a quantity forced by definition or by a self-citation. No load-bearing self-citations, imported uniqueness theorems, or ansatz-by-citation steps were found. Accordingly, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The method rests on standard pretrained diffusion components, a hand-chosen definition of the standardized target, and a synthetic-to-real transfer assumption that the paper itself flags as biased. The invented components are trained modules whose evidence is internal to the paper's experiments.

free parameters (3)
  • standardization target definition = fixed pose, white background, neutral expression, face scaled to fixed position
    Defines what 'standardized NAE' means; chosen by hand in Section 3.3 Synthetic Data. All claimed consistency benefits are relative to this specific target.
  • IP-Adapter token feature count = 257
    Reference encoder capacity for the personalization stage (Section 4.1). Supplementary token scaling study shows a consistency/controllability tradeoff, so this value affects the reported balance.
  • training/inference resolution = 512 x 768
    Used for both stages; a hand-set hyperparameter (Section 4.1).
assumptions (4)
  • domain assumption The pretrained components (SDXL, IP-Adapter, CLIP, ArcFace, Animate Anyone-style architecture) function as described by their original papers.
    The method is built on these external components with minimal modification (Sections 3.1, 4.1).
  • ad hoc to paper Synthetic 3D rendered pairs teach the standardizer to preserve appearance while changing pose, background, expression, and viewpoint.
    Stage I is trained only on 10.5M rendered pairs from 2,924 character models (Section 4.1). Generalization to real photos is assumed.
  • domain assumption CLIP-I, CLIP-T, Face Sim, PSNR, SSIM, FVD, and FID-VID are valid proxies for the stated qualities.
    All quantitative conclusions rely on these metrics without calibration against human judgment, although a user study is reported for the main comparison.
  • ad hoc to paper The 3D style bias introduced by standardization is an integral part of standardization and is mitigated by Stage II training.
    Section 3.4 and Supp. Figure 11 acknowledge the bias but assert it does not impede training; this is load-bearing for the realism of outputs.
invented entities (1)
  • Foreground/background learnable class tokens (vf and vb)
    purpose: Inject background mask information into each self-attention module of the ReferenceNet (Eq. 3) to improve standardization.
    Validated only on the paper's own synthetic and final metrics (Tables 3 and 4); no independent external benchmark demonstrates their specific contribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SerialGen: Personalized Image Generation by First Standardization Then Personalization." pith.science (2026). https://pith.science/paper/PJ5TOOCR

@misc{pith2026241201485,
  author       = {Pith},
  title        = {Pith review of: SerialGen: Personalized Image Generation by First Standardization Then Personalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PJ5TOOCR}},
  note         = {Machine review of arXiv:2412.01485}
}
read the original abstract

In this work, we are interested in achieving both high text controllability and whole-body appearance consistency in the generation of personalized human characters. We propose a novel framework, named SerialGen, which is a serial generation method consisting of two stages: first, a standardization stage that standardizes reference images, and then a personalized generation stage based on the standardized reference. Furthermore, we introduce two modules aimed at enhancing the standardization process. Our experimental results validate the proposed framework's ability to produce personalized images that faithfully recover the reference image's whole-body appearance while accurately responding to a wide range of text prompts. Through thorough analysis, we highlight the critical contribution of the proposed serial generation method and standardization model, evidencing enhancements in appearance consistency between reference and output images and across serial outputs generated from diverse text prompts. The term "Serial" in this work carries a double meaning: it refers to the two-stage method and also underlines our ability to generate serial images with consistent appearance throughout.

Figures

Figures reproduced from arXiv: 2412.01485 by the authors.

Figure 1
Figure 1. Serial images generated by SerialGen. Our method can produce personalized images that faithfully recover the reference image’s [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed SerialGen with two stages: (1) [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the standardization model. The pose and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Comparison with other methods. Our method is capable of generating images with high text controllability and appearance [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visual comparison of different training strategies. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Consistency between two prompts. The inconsistent part [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: More serial images generated by SerialGen, showcasing its outstanding ability to maintain whole-body appearance consistency [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Extension of Figure [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Details of the Reference Pose Injection Module. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: The standardization introduces a slight 3D style bias, particularly evident in head-only inputs (last row), resulting in clothing [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: More comparison with other methods. 10. Limitations of Unpaired Training In these experiments, we train models on unpaired image data, using identical images as both reference and target. For the reference encoder, we employ IP-Adapter [33], while SDXL is utilized as …
Figure 13
Figure 13. Figure 13: Visual comparison of different numbers of token features. Leftmost is the reference image. Token- [PITH_FULL_IMAGE:figures/full_fig_p015_13.png]
Figure 14
Figure 14. Figure 14: Comparison of different numbers of token features . [PITH_FULL_IMAGE:figures/full_fig_p015_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 23 canonical work pages

  1. [1]

    Moab Arar, Rinon Gal, Yuval Atzmon, Gal Chechik, Daniel Cohen-Or, Ariel Shamir, and Amit H. Bermano. Domain- agnostic tuning-encoder for fast personalization of text-to- image models. In SIGGRAPH Asia 2023 Conference Papers, pages 1–10, 2023. 2

    work page 2023

  2. [2]

    Magicpose: Realistic human poses and facial expressions retargeting with identity-aware diffusion

    Di Chang, Yichun Shi, Quankai Gao, Hongyi Xu, Jessica Fu, Guoxian Song, Qing Yan, Yizhe Zhu, Xiao Yang, and Mo- hammad Soleymani. Magicpose: Realistic human poses and facial expressions retargeting with identity-aware diffusion. In Forty-first International Conference on Machine Learn- ing, 2023. 3, 5

    work page 2023

  3. [3]

    Subject-driven text-to-image generation via apprenticeship learning

    Wenhu Chen, Hexiang Hu, Yandong Li, Nataniel Ruiz, Xuhui Jia, Ming-Wei Chang, and William W Cohen. Subject-driven text-to-image generation via apprenticeship learning. Advances in Neural Information Processing Sys- tems, 36:30286–30305, 2023. 2

    work page 2023

  4. [4]

    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. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 4690–4699, 2019. 3, 6

    work page 2019

  5. [5]

    Sub-center arcface: Boosting face recognition by large-scale noisy web faces

    Jiankang Deng, Jia Guo, Tongliang Liu, Mingming Gong, and Stefanos Zafeiriou. Sub-center arcface: Boosting face recognition by large-scale noisy web faces. InProceedings of the IEEE Conference on European Conference on Computer Vision, 2020. 3

    work page 2020

  6. [6]

    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. 3

    2021

  7. [7]

    An image is worth one word: Personalizing text-to- image generation using textual inversion

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patash- nik, Amit H Bermano, Gal Chechik, and Daniel Cohen- Or. An image is worth one word: Personalizing text-to- image generation using textual inversion. arXiv preprint arXiv:2208.01618, 2022. 2

    arXiv 2022

  8. [8]

    Encoder-based domain tuning for fast personalization of text-to-image models.ACM Transactions on Graphics (TOG), 42(4):1–13, 2023

    Rinon Gal, Moab Arar, Yuval Atzmon, Amit H Bermano, Gal Chechik, and Daniel Cohen-Or. Encoder-based domain tuning for fast personalization of text-to-image models.ACM Transactions on Graphics (TOG), 42(4):1–13, 2023. 2

    2023

Show all 47 references
  1. [9]

    Lcm- lookahead for encoder-based text-to-image personalization

    Rinon Gal, Or Lichter, Elad Richardson, Or Patashnik, Amit H Bermano, Gal Chechik, and Daniel Cohen-Or. Lcm- lookahead for encoder-based text-to-image personalization. In European Conference on Computer Vision , pages 322–

  2. [10]

    Pulid: Pure and lightning id customization via contrastive alignment

    Zinan Guo, Yanze Wu, Zhuowei Chen, Lang Chen, Peng Zhang, and Qian He. Pulid: Pure and lightning id customization via contrastive alignment. arXiv preprint arXiv:2404.16022, 2024. 2

    2024 arXiv

  3. [11]

    Face-adapter for pre-trained diffusion models with fine- grained id and attribute control

    Yue Han, Junwei Zhu, Keke He, Xu Chen, Yanhao Ge, Wei Li, Xiangtai Li, Jiangning Zhang, Chengjie Wang, and Yong Liu. Face-adapter for pre-trained diffusion models with fine- grained id and attribute control. In European Conference on Computer Vision, pages 20–36. Springer, 2025

    2025

  4. [12]

    Imagine yourself: Tuning-free personalized image generation

    Zecheng He, Bo Sun, Felix Juefei-Xu, Haoyu Ma, Ankit Ramchandani, Vincent Cheung, Siddharth Shah, Anmol Kalia, Harihar Subramanyam, Alireza Zareian, et al. Imagine yourself: Tuning-free personalized image generation. arXiv preprint arXiv:2409.13346, 2024. 3

    2024 arXiv

  5. [13]

    Image quality metrics: Psnr vs

    Alain Hore and Djemel Ziou. Image quality metrics: Psnr vs. ssim. In 2010 20th international conference on pattern recognition, pages 2366–2369. IEEE, 2010. 8

    2010

  6. [14]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. InIn- ternational Conference on Learning Representations , 2022. 3

    2022

  7. [15]

    Animate anyone: Consistent and controllable image- to-video synthesis for character animation

    Li Hu. Animate anyone: Consistent and controllable image- to-video synthesis for character animation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8153–8163, 2024. 3, 5

    2024

  8. [16]

    Tcan: Animating human images with temporally consis- tent pose guidance using diffusion models

    Jeongho Kim, Min-Jung Kim, Junsoo Lee, and Jaegul Choo. Tcan: Animating human images with temporally consis- tent pose guidance using diffusion models. arXiv preprint arXiv:2407.09012, 2024. 3, 5

    2024 arXiv

  9. [17]

    Multi-concept customization of text-to-image diffusion

    Nupur Kumari, Bingliang Zhang, Richard Zhang, Eli Shechtman, and Jun-Yan Zhu. Multi-concept customization of text-to-image diffusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1931–1941, 2023. 2

    1931

  10. [18]

    Photomaker: Customizing re- alistic human photos via stacked id embedding

    Zhen Li, Mingdeng Cao, Xintao Wang, Zhongang Qi, Ming- Ming Cheng, and Ying Shan. Photomaker: Customizing re- alistic human photos via stacked id embedding. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8640–8650, 2024. 2, 3

    2024

  11. [19]

    Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M ¨uller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023. 2, 3, 5

    2023 arXiv

  12. [20]

    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

  13. [21]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International confer- ence on machine learning, pages 8821–8831. Pmlr, 2021. 2

    2021

  14. [22]

    Hierarchical text-conditional image gener- ation with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image gener- ation with clip latents. arXiv preprint arXiv:2204.06125, 1 (2):3, 2022

    2022 arXiv

  15. [23]

    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. 2, 3

    2022

  16. [24]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 2250...

    2023

  17. [25]

    Photorealistic text-to-image diffusion models with deep language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep language understanding. Advances in neural information...

    2022

  18. [26]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. arXiv:2010.02502, 2020. 3, 5

    2010 arXiv

  19. [27]

    Instantid: Zero-shot identity-preserving generation in seconds

    Qixun Wang, Xu Bai, Haofan Wang, Zekui Qin, Anthony Chen, Huaxia Li, Xu Tang, and Yao Hu. Instantid: Zero-shot identity-preserving generation in seconds. arXiv preprint arXiv:2401.07519, 2024. 2, 3

    2024 arXiv

  20. [28]

    Disco: Disentangled control for realistic human dance generation

    Tan Wang, Linjie Li, Kevin Lin, Yuanhao Zhai, Chung- Ching Lin, Zhengyuan Yang, Hanwang Zhang, Zicheng Liu, and Lijuan Wang. Disco: Disentangled control for realistic human dance generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition ,...

    2024

  21. [29]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4):600–612, 2004. 8

    2004

  22. [30]

    Fastcomposer: Tuning-free multi- subject image generation with localized attention

    Guangxuan Xiao, Tianwei Yin, William T Freeman, Fr ´edo Durand, and Song Han. Fastcomposer: Tuning-free multi- subject image generation with localized attention. Interna- tional Journal of Computer Vision, pages 1–20, 2024. 2, 3, 6

    2024

  23. [31]

    Magicanimate: Temporally consistent human im- age animation using diffusion model

    Zhongcong Xu, Jianfeng Zhang, Jun Hao Liew, Hanshu Yan, Jia-Wei Liu, Chenxu Zhang, Jiashi Feng, and Mike Zheng Shou. Magicanimate: Temporally consistent human im- age animation using diffusion model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

    2024

  24. [32]

    Effec- tive whole-body pose estimation with two-stages distillation

    Zhendong Yang, Ailing Zeng, Chun Yuan, and Yu Li. Effec- tive whole-body pose estimation with two-stages distillation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4210–4220, 2023. 4

    2023

  25. [33]

    Ip- adapter: Text compatible image prompt adapter for text-to- image diffusion models

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. Ip- adapter: Text compatible image prompt adapter for text-to- image diffusion models. arXiv preprint arXiv:2308.06721,

    arXiv

  26. [34]

    Posecrafter: One-shot per- sonalized video synthesis following flexible pose control

    Yong Zhong, Min Zhao, Zebin You, Xiaofeng Yu, Chang- wang Zhang, and Chongxuan Li. Posecrafter: One-shot per- sonalized video synthesis following flexible pose control. In European Conference on Computer Vision, pages 243–260. Springer, 2025. 3

    2025

  27. [35]

    Storydiffusion: Consistent self- attention for long-range image and video generation

    Yupeng Zhou, Daquan Zhou, Ming-Ming Cheng, Jiashi Feng, and Qibin Hou. Storydiffusion: Consistent self- attention for long-range image and video generation. arXiv preprint arXiv:2405.01434, 2024. 3

    2024 arXiv

  28. [36]

    Storymaker: Towards holistic consistent characters in text-to-image generation

    Zhengguang Zhou, Jing Li, Huaxia Li, Nemo Chen, and Xu Tang. Storymaker: Towards holistic consistent characters in text-to-image generation. arXiv preprint arXiv:2409.12576,

    arXiv

  29. [37]

    Champ: Controllable and consistent human image animation with 3d parametric guidance

    Shenhao Zhu, Junming Leo Chen, Zuozhuo Dai, Yinghui Xu, Xun Cao, Yao Yao, Hao Zhu, and Siyu Zhu. Champ: Controllable and consistent human image animation with 3d parametric guidance. In European Conference on Computer Vision (ECCV), 2024. 3, 5 SerialGen: Personalized Image Gen...

    2024

  30. [39]

    Details of the Reference Pose Injection Mod- ule We utilize a light convolutional network to extract pose fea- ture maps from pose images. The architectural setup is de- picted in Figure 10, where 3 × 3 conv, 32, ↓ 2 indicates a convolutional layer with a kernel size of 3 × 3,...

  31. [40]

    As shown in Figure 11, the standardization stage introduces a slight 3D style bias when standardizing images

    Impact of 3D Style Bias As depicted in the second paragraph of Section 3.4, we demonstrate the impact of 3D style bias introduced by the synthetic data. As shown in Figure 11, the standardization stage introduces a slight 3D style bias when standardizing images. This bias is e...

  32. [41]

    We give more comparison results with Fast- Composer, IP-Adapter and StoryMaker

    More Comparison Results and Analysis This part gives supplementary comparisons and analysis in Section 4.2. We give more comparison results with Fast- Composer, IP-Adapter and StoryMaker. As shown in the Figure 12, we selected four different characters for analy- sis, which in...

  33. [42]

    User Study As shown in Table 5, we design three criteria for com- parison, where each criterion receives 600 valid votes (30 participant × 20 text-image pairs). The detailed questions Method W AC TC V A IP-Adapter [33] 20.00% 4.33% 5.67% FastComposer [30] 4.67% 3.67% 0.67% Sto...

  34. [43]

    Figure 12

    Visual Appeal: Which method produces the most visu- ally appealing image? To ensure objectivity, the names of all methods are anonymized, and the methods are presented in a randomized order for each question. Figure 12. More comparison with other methods

  35. [44]

    For the reference encoder, we employ IP-Adapter [33], while SDXL is utilized as the diffusion model

    Limitations of Unpaired Training In these experiments, we train models on unpaired image data, using identical images as both reference and target. For the reference encoder, we employ IP-Adapter [33], while SDXL is utilized as the diffusion model. The feature size extracted f...

  36. [45]

    Experiments are conducted using the benchmark dataset TikTok [28]

    Comparison to Human Image Animation Models As discussed in Section 4.3.2, we compare the architec- ture of our standardization model with other leading hu- man image animation models, including DisCo [28], Mag- icPose [2], MagicAnimate [31], Animate Anyone [15], Champ [37], an...

  37. [46]

    As shown in Table 7, after the standardization stage, CLIP-I is 89.47, and Face Sim

    Ablation Study on Identity Loss We conduct an ablation study to evaluate the impact of each stage on identity preservation. As shown in Table 7, after the standardization stage, CLIP-I is 89.47, and Face Sim. is 0.69. Following the personalization stage, CLIP-I decreases to 85...

  38. [47]

    score of 0.46, CLIP-I score of 74.56, and CLIP-T score of 24.63 on the test dataset

    More Quantitative Comparisons We also made a quantitative comparison between our method and the recent face-oriented approach LCM- Lookahead [9], which achieved a Face Sim. score of 0.46, CLIP-I score of 74.56, and CLIP-T score of 24.63 on the test dataset. Our method outperfo...

  39. [340]

    Springer, 2024. 2, 3, 6

    2024

Pith tools

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