Pith. sign in

REVIEW 5 major objections 6 minor 52 references

PersonaMagic: Stage-Regulated High-Fidelity Face Customization with Tandem Equilibrium

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read PersonaMagic customizes faces from a single image by learning timestep-restricted embeddings and a text-encoder attention-balancing loss, preserving both identity and prompt semantics.

desk verdict A plausible and well-ablated face personalization method whose quantitative claims outrun the evidence because the stage thresholds are tuned and evaluated on the same split. read the letter →

arxiv 2412.15674 v1 pith:NKGBCY5E submitted 2024-12-20 cs.CV

Xinzhe Li , Jiahui Zhan , Shengfeng He , Yangyang Xu , Junyu Dong , Huaidong Zhang , Yong Du This is my paper · ORCID
classification cs.CV
keywords facecustomizationpersonalizedimagegenerationdiffusionmodelstext-to-imagestage-regulatedembeddingstandemequilibriumidentitypreservationself-attention
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to solve the core tension in personalized face generation: keeping a person's identity while following a text prompt. It proposes PersonaMagic, which learns a series of word embeddings that are active only in a middle slice of the diffusion timesteps, where cross-attention maps lock onto the face rather than the whole image. A companion loss, called Tandem Equilibrium, rebalances self-attention weights inside the final text-encoder layer so the learned placeholder token does not crowd out other prompt words. If correct, the method preserves facial identity from a single portrait while staying faithful to the requested scene or style, and it can be bolted onto existing personalization models to improve them.

What carries the argument

The mechanism is a pair of interventions on attention: (1) a stage-regulated embedding network that only learns in the timestep interval where the placeholder token's cross-attention maps overlap the face mask (IoU above λ1 and below λ2), and (2) the Tandem Equilibrium loss Lte = -φ(Att_S*) × φ(Att_max), applied to the final self-attention layer of the frozen text encoder, which drives the summed attention on the placeholder and on the strongest original token toward equality. A cosine-scheduled identity loss λid(t)Lid is added to the masked denoising loss to preserve facial features during training.

What would settle it

Run PersonaMagic on a face dataset with systematically different pose or occlusion statistics than CelebA-HQ using the paper's fixed λ1=0.7 and λ2=0.8; if identity preservation or text similarity drops substantially and retuning thresholds recovers the gap, the stage partition is overfitted to the tuning set. Alternatively, an ablation that replaces Lte with a simple L2 penalty on the placeholder self-attention sum and observes equal or better metrics would undercut the claim that the product form is needed.

Watch

Extended reading notes

Core claim

The paper's central claim is that the fidelity-editability trade-off in face customization can be resolved by partitioning the denoising schedule into three stages determined by the intersection-over-union between the placeholder token's cross-attention map and a facial mask. In the early and late stages, fixed supercategory embeddings are used; only in the intermediate dynamic stage does a lightweight MLP output timestep-dependent embeddings that carry identity. The Tandem Equilibrium loss maximizes the product of the attention sums of the learned token and the strongest remaining prompt token in the text encoder's last self-attention layer, which makes the model attend to both identity and description. The paper reports that this yields state-of-the-art results on CelebA-HQ, FFHQ, and LFW, and that injecting the stage-regulated embeddings into PhotoMaker and IP-Adapter improves both text similarity and identity preservation.

Load-bearing premise

The approach assumes that rebalancing self-attention in the final text-encoder layer will shift U-Net cross-attention enough to restore prompt semantics without eroding identity, and that the fixed IoU thresholds λ1=0.7 and λ2=0.8 tuned on CelebA-HQ transfer to other face datasets and web images.

Editorial extensions

If this is right

  • Single-image face customization can rival few-shot personalization methods in identity preservation.
  • The stage partition removes background overfitting caused by learning embeddings across all timesteps.
  • Tandem Equilibrium improves text similarity without eroding identity, resolving the usual fidelity-editability trade-off.
  • PersonaMagic acts as a model-agnostic plug-in for pretrained personalization models like PhotoMaker and IP-Adapter, raising both metrics.
  • The method transfers to non-face domains, such as animals and objects, with similar identity retention.

Reading between the lines

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

  • One implication not drawn by the paper is that the IoU-based stage partition could be made adaptive per image rather than using fixed thresholds, potentially improving robustness on faces with unusual poses or occlusions.
  • The tandem equilibrium loss could be applied to other personalized generation tasks beyond faces, anywhere a learned token tends to dominate the text encoder's self-attention.
  • The claim implies that text-encoder self-attention is a lever for editability in personalization, suggesting future methods might tune it directly rather than only through U-Net cross-attention.
  • Because the thresholds were selected on CelebA-HQ, a testable extension is to measure performance degradation on systematically different face distributions.
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

5 major / 6 minor

Summary. The paper proposes PersonaMagic, a single-image face customization method for text-to-image diffusion models. The method partitions the denoising timestep schedule into static and dynamic stages based on cross-attention IoU with a facial mask, learns timestep-dependent embeddings in the dynamic stage with a lightweight MLP, and introduces a Tandem Equilibrium (TE) loss that rebalances self-attention in the frozen text encoder between the learned concept token and the strongest original token. The diffusion model is kept frozen. Experiments are reported on CelebA-HQ, FFHQ, and LFW for one-shot and few-shot settings, with comparisons to Textual Inversion, DreamBooth, Custom Diffusion, NeTI, and Perfusion, plus plug-in experiments with PhotoMaker and IP-Adapter and applications to non-facial domains.

Significance. If the quantitative claims survive re-evaluation, PersonaMagic is a practically attractive contribution: freezing the diffusion model and learning dynamic embeddings only in an intermediate timestep interval is a sensible middle ground between textual inversion and full fine-tuning, and the TE loss targets a real failure mode in which the learned token dominates prompt semantics. The plug-in experiments with PhotoMaker and IP-Adapter (Table 3) and the cross-model robustness check on Dreamlike Photoreal v2.0 are useful extensions. The temporal IoU analysis of cross-attention maps in Fig. 2 gives an intuitive basis for stage partitioning. However, the current evidence is not yet strong enough for the headline 'superiority' and 'Pareto front' claims: the main hyperparameters are tuned and evaluated on the same dataset, the baselines are not fully controlled, and no uncertainty quantification is provided for the key quantitative comparisons.

major comments (5)
  1. [Quantitative Evaluation / Table 2 / Table 6 / Fig. 8] The stage thresholds λ1 and λ2 are selected by a grid search on CelebA-HQ (Table 2), and the same dataset is then used as the headline one-shot evaluation in Fig. 8 and Table 6. This creates a selection loop: the reported CelebA-HQ advantage over Custom Diffusion (T-Sim 0.747 vs 0.738; I-Pre 0.345 vs 0.283) reflects hyperparameters chosen on that same data. Table 2 shows that neighboring configurations differ by small amounts (e.g., λ1=0.7, λ2=0.9 gives 0.741/0.328 versus the chosen 0.744/0.358), so without a held-out validation split or confidence intervals one cannot rule out selection noise. The FFHQ/LFW results with fixed constants are suggestive, but they do not validate the selection procedure for CelebA-HQ. Please either tune on a separate validation split and report the chosen configuration, or report performance across the full Table 2 grid on all datasets together with uncertainty estimates.
  2. [Appendix: Implementation Details / Table 6] In the main comparison, Textual Inversion is evaluated on the official LDM model while all other methods, including PersonaMagic, use Stable Diffusion v1.4, as stated in the Appendix and in the Table 6 footnote. Even if the choice of LDM is intended to favor Textual Inversion, the reported numbers are not directly comparable across different backbones, and the main-text comparison does not disclose this limitation. The same issue applies to the use of unofficial implementations for DreamBooth and Perfusion. Please re-run Textual Inversion on the same Stable Diffusion v1.4 pipeline used for the other baselines, or clearly mark the comparison as cross-model and provide a same-backbone result for at least the key competitors.
  3. [Quantitative Evaluation / Fig. 8 / Table 6] No error bars, confidence intervals, or significance tests are reported for any of the headline quantitative metrics, despite the text saying the evaluation was repeated five times. The reported margins are small: on CelebA-HQ, PersonaMagic is ahead of Custom Diffusion by 0.009 in T-Sim and by 0.062 in I-Pre (Table 6). Without variance information it is impossible to determine whether these differences are stable. Please report per-run standard deviations or confidence intervals and, where appropriate, paired significance tests over the sampled subjects.
  4. [Face Customization With Tandem Equilibrium / Eq. (1)] The definition of the Tandem Equilibrium loss is ambiguous and may be vacuous as written. The text says a softmax function is applied to the self-attention map of the highest-attention original token 'with that of S*', and then Lte = -φ(AttS*) × φ(Attmax). If the softmax is applied separately to each map, then each map sums to 1, so φ(AttS*) = φ(Attmax) = 1 for all inputs and the loss is constant, which cannot train anything. If instead the softmax is applied to the concatenation of the two maps, this should be stated explicitly and the total over which the 'total sum remains constant' is computed should be defined. Please clarify the normalization and demonstrate that the loss is not constant by construction.
  5. [Appendix: Additional Ablation Study / Eq. (4)] The appendix states that replacing λid(t) with 'its mathematical expectation 1/T over the interval [0,T]' results in a slight decline in identity preservation. This is incorrect: the average of λid(t) = cos(π t/(2T)) over [0,T] is 2/π ≈ 0.637, not 1/T. If the authors actually used 1/T as the constant, the identity-loss weight would be nearly zero for typical T (e.g., T=1000), so the ablation would essentially remove L_id rather than test the temporal schedule. Please correct the calculation and re-run the ablation with the correct constant.
minor comments (6)
  1. [Stage-regulated Textual Conditioning] Please clarify the ordering of the static and dynamic stages with respect to the timestep variable t: the text says 'IoU gradually increases as timestep progresses', but 'progresses' can mean either increasing noise (larger t) or increasing denoising progress (smaller t).
  2. [Stage-regulated Textual Conditioning / Fig. 2] In Fig. 2 and the stage-partition description, specify whether the mask used for IoU is the CLIPSeg-predicted mask or a ground-truth facial mask; the method section says CLIPSeg, while the ablation text refers to 'real masks'.
  3. [Face Customization With Tandem Equilibrium] There is a typo in the sentence introducing Eq. (1): 'We then applying a softmax function' should read 'We then apply a softmax function'.
  4. [Quantitative Evaluation] The description 'we randomly selected 100 distinct images from CelebA-HQ and FFHQ respectively and repeated five times, gotten 500 images of each dataset' should clarify whether the 100 images are newly sampled in each repetition and should report how many individuals are used for the LFW few-shot evaluation.
  5. [Ablation Study / Table 1] The '+Tandem Equilibrium' row in Table 1 is not explained in the same detail as the '+Stage Regulation' row; please state explicitly which components are present in this row and discuss why adding TE to the vanilla model reduces I-Pre from 0.334 to 0.284 while the full model improves it.
  6. [User Study] In the user-study paragraph, 'Prefusion' should be 'Perfusion'; please also provide details of the 2AFC procedure, including how the 120 subjects and 15 prompts per subject were distributed across the 50 participants.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild evaluation circularity: the stage-partition thresholds are tuned and reported on the same CelebA-HQ T-Sim/I-Pre metrics, but the core stage-regulation and Tandem Equilibrium mechanism is otherwise self-contained.

  1. fitted input called prediction [Experiments - Ablation Study (Table 2) and Quantitative Evaluation (Fig. 8/Table 6)]
    "Based on the results, we chose λ1 = 0.7. Further refinement of λ2 divided the process into three stages, with a third static stage emerging as λ2 decreased, using fixed embeddings. ... Therefore, we set λ1 = 0.7 and λ2 = 0.8. ... For quantitative evaluation of one-shot setting, we randomly selected 100 distinct images from CelebA-HQ and FFHQ datasets respectively ... The average score presented in Fig. 8. Our method outperforms competitors, sitting on the pareto front."

    The stage thresholds λ1=0.7 and λ2=0.8 are chosen by grid search in Table 2, which reports T-Sim and I-Pre on CelebA-HQ and states the chosen configuration 'achieves the optimal balance between fidelity and editability.' The same dataset with the same two metrics is then the basis of the headline quantitative comparison (Fig. 8 and Table 6) where PersonaMagic is claimed to 'sit on the Pareto front.' Consequently the reported CelebA-HQ advantage is an in-sample selected optimum of the threshold sweep, not an independent prediction: the configuration was selected because it maximized these metrics on these images, so the CelebA-HQ Pareto-front position is partially forced by the selection procedure.

full rationale

The method's core derivation is not circular. The Tandem Equilibrium loss Lte = -φ(Att_S*)φ(Att_max) is a self-attention regularizer in the text encoder, not equivalent to the reported CLIPScore or CosFace metrics; the identity loss Lid uses Arcface features while evaluation uses CosFace; and the ablation in Table 5 shows identity loss alone does not produce the gains. No load-bearing argument rests on a self-citation chain: citations to prior work (NeTI, IP-Adapter, PhotoMaker) are external and architectural. The one defensible circularity is the evaluation loop around the stage-partition thresholds: λ1/λ2 are tuned on CelebA-HQ using the same T-Sim/I-Pre metrics and then CelebA-HQ appears in the headline Pareto-front comparison. This makes the CelebA-HQ quantitative superiority partly in-sample. Because the rest of the derivation is self-contained and FFHQ/LFW serve as held-out datasets for the fixed thresholds, the overall circularity is mild.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on four fitted or hand-chosen schedules, several domain assumptions about diffusion attention and face masks, and no new physical entities. The dynamic embeddings are learned parameters and the Tandem Equilibrium term is a loss function.

free parameters (4)
  • Stage thresholds λ1, λ2 = 0.7, 0.8
    Selected by sweeping values on CelebA-HQ in Table 2 to maximize reported text similarity and identity preservation; the same dataset is used for the headline comparison.
  • Identity loss schedule λid(t) = cos(t π / (2T))
    Hand-designed cosine weight that down-weights identity loss at large timesteps; the appendix says replacing it with its expectation lowers identity preservation, but the form is heuristic.
  • Classifier-free guidance scale = 8.0
    Set in implementation details for all sampling; no sensitivity analysis is reported.
  • Training budget hyperparameters = 1000 iterations, batch size 2, learning rate 5e-5, weight decay 0.01
    Fixed across experiments; no evidence that the reported Pareto front is robust to these choices.
assumptions (6)
  • domain assumption A pretrained Stable Diffusion text encoder and U-Net provide a valid generative prior for faces.
    The entire method freezes SD 1.4 and assumes learned embeddings can carry identity through its text conditioning (Method section).
  • domain assumption CLIPSeg face masks and cross-attention IoU correctly identify when the learned token focuses on the face.
    Stage partition is computed from IoU between CLIPSeg masks and cross-attention maps (Figure 2 and Method section).
  • ad hoc to paper The middle interval of the denoising schedule is the best place to learn identity embeddings.
    The paper cites weak text control at late timesteps and noisy attention at early timesteps, but the exact interval is set by fitted thresholds rather than derived.
  • ad hoc to paper Rebalancing final-layer self-attention in the text encoder transfers to U-Net cross-attention semantics.
    The Tandem Equilibrium loss is built on this causal link; only the ablation in Figure 9 supports it.
  • standard math Softmax normalization makes the two attention masses complementary, so maximizing their product balances them.
    This is the math behind Eq. (1), used to justify the Tandem Equilibrium loss.
  • domain assumption ArcFace features of the predicted clean image are a usable training signal for identity.
    Identity loss Lid is defined in Eq. (3); Table 5 shows it alone does not improve Custom Diffusion, so its contribution is limited.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PersonaMagic: Stage-Regulated High-Fidelity Face Customization with Tandem Equilibrium." pith.science (2026). https://pith.science/paper/NKGBCY5E

@misc{pith2026241215674,
  author       = {Pith},
  title        = {Pith review of: PersonaMagic: Stage-Regulated High-Fidelity Face Customization with Tandem Equilibrium},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NKGBCY5E}},
  note         = {Machine review of arXiv:2412.15674}
}
read the original abstract

Personalized image generation has made significant strides in adapting content to novel concepts. However, a persistent challenge remains: balancing the accurate reconstruction of unseen concepts with the need for editability according to the prompt, especially when dealing with the complex nuances of facial features. In this study, we delve into the temporal dynamics of the text-to-image conditioning process, emphasizing the crucial role of stage partitioning in introducing new concepts. We present PersonaMagic, a stage-regulated generative technique designed for high-fidelity face customization. Using a simple MLP network, our method learns a series of embeddings within a specific timestep interval to capture face concepts. Additionally, we develop a Tandem Equilibrium mechanism that adjusts self-attention responses in the text encoder, balancing text description and identity preservation, improving both areas. Extensive experiments confirm the superiority of PersonaMagic over state-of-the-art methods in both qualitative and quantitative evaluations. Moreover, its robustness and flexibility are validated in non-facial domains, and it can also serve as a valuable plug-in for enhancing the performance of pretrained personalization models.

Figures

Figures reproduced from arXiv: 2412.15674 by the authors.

Figure 1
Figure 1. PersonaMagic seamlessly generates images of new roles, styles, or scenes based on a user-provided portrait. By learning stage-regulated embeddings through a Tandem Equilibrium strategy, our method accurately captures and represents unseen concepts, faithfully creating personas aligned with the provided prompts while minimizing identity distortion. the static stage employs fixed word embeddings correspond￾ing to supe… view at source ↗
Figure 2
Figure 2. Cross-attention maps of S∗ at each timestep. We calculate IoU with facial mask for stage partition. recently excelled in generating images from text (Ramesh et al. 2021; Gu et al. 2022; Yu et al. 2022; Saharia et al. 2022). Noteworthy examples include GLIDE (Nichol et al. 2022), which crafts high-resolution images using diverse diffusion models, and DALL·E 2 (Ramesh et al. 2022), generating CLIP (Radford et al. 2021… view at source ↗
Figure 3
Figure 3. Overview of our pipeline. Given a single image, we learn a series of embeddings during dynamic stage to capture [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: The overlooked semantic yield suboptimal atten [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Illustration of the proposed Tandem Equilibrium. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison with state-of-the-art methods on celebrities. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison with state-of-the-art methods on non-celebrities. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Quantitative evaluation on CelebA-HQ, FFHQ, [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 9
Figure 9. Figure 9: Customized results with and without Lte during training. Attention weights are annotated in the lower left corner of cross-attention maps. a series of embeddings for new concepts, which outperforms optimizing a single embedding. However, text similarity de￾creased slig…
Figure 10
Figure 10. Figure 10: Qualitative ablation study of different model vari [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 12
Figure 12. Figure 12: Our method can be applied to various down [PITH_FULL_IMAGE:figures/full_fig_p009_12.png]
Figure 11
Figure 11. Figure 11: Quantitative evaluation based on Dreamlike Pho [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 14
Figure 14. Figure 14: Failure cases of PersonaMagic [PITH_FULL_IMAGE:figures/full_fig_p010_14.png]
Figure 15
Figure 15. Figure 15: Integrating PersonaMagic into pre-trained per [PITH_FULL_IMAGE:figures/full_fig_p010_15.png]
Figure 16
Figure 16. Figure 16: More visual results of PersonaMagic on celebrities and non-celebrities, along with evaluation prompts, demonstrate [PITH_FULL_IMAGE:figures/full_fig_p011_16.png]
Figure 17
Figure 17. Figure 17: Additional visual results of PersonaMagic integrated as a plug-in for pre-trained personalization model PhotoMaker. [PITH_FULL_IMAGE:figures/full_fig_p012_17.png]
Figure 18
Figure 18. Figure 18: Additional visual results of PersonaMagic integrated as a plug-in for pre-trained personalization model IP-Adapter. [PITH_FULL_IMAGE:figures/full_fig_p013_18.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 33 canonical work pages

  1. [1]

    Alaluf, Y.; Richardson, E.; Metzer, G.; and Cohen-Or, D. 2023. A neural space-time representation for text-to-image personalization. ACM TOG, 42(6): 1--10

    work page 2023

  2. [2]

    Alaluf, Y.; Tov, O.; Mokady, R.; Gal, R.; and Bermano, A. 2022. Hyperstyle: Stylegan inversion with hypernetworks for real image editing. In CVPR, 18511--18521

    work page 2022

  3. [3]

    Avrahami, O.; Fried, O.; and Lischinski, D. 2023. Blended latent diffusion. ACM TOG, 42(4): 1--11

    work page 2023

  4. [4]

    L.; Kiros, J

    Ba, J. L.; Kiros, J. R.; and Hinton, G. E. 2016. Layer normalization. arXiv preprint arXiv:1607.06450

    arXiv 2016

  5. [5]

    Balaji, Y.; Nah, S.; Huang, X.; Vahdat, A.; Song, J.; Kreis, K.; Aittala, M.; Aila, T.; Laine, S.; Catanzaro, B.; et al. 2022. ediffi: Text-to-image diffusion models with an ensemble of expert denoisers. arXiv preprint arXiv:2211.01324

    arXiv 2022

  6. [6]

    ChenDarYen. 2023. Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization. https://github.com/ChenDarYen/Key-Locked-Rank-One-Editing-for-Text-to-Image-Personalization

    work page 2023

  7. [7]

    CompVis. 2022. Stable Diffusion. https://huggingface.co/CompVis/stable-diffusion-v-1-4-original

    work page 2022

  8. [8]

    Deng, J.; Guo, J.; Xue, N.; and Zafeiriou, S. 2019. Arcface: Additive angular margin loss for deep face recognition. In CVPR, 4690--4699

    2019

Show all 52 references
  1. [9]

    Dreamlike.art. 2023. Dreamlike Photoreal. https://huggingface.co/dreamlike-art/dreamlike-photoreal-2.0

    2023

  2. [10]

    Esser, P.; Rombach, R.; and Ommer, B. 2021. Taming transformers for high-resolution image synthesis. In CVPR, 12873--12883

    2021

  3. [11]

    H.; Chechik, G.; and Cohen-Or, D

    Gal, R.; Alaluf, Y.; Atzmon, Y.; Patashnik, O.; Bermano, A. H.; Chechik, G.; and Cohen-Or, D. 2023 a . An image is worth one word: Personalizing text-to-image generation using textual inversion. In ICLR

    2023

  4. [12]

    H.; Chechik, G.; and Cohen-Or, D

    Gal, R.; Arar, M.; Atzmon, Y.; Bermano, A. H.; Chechik, G.; and Cohen-Or, D. 2023 b . Encoder-based domain tuning for fast personalization of text-to-image models. ACM TOG, 42(4): 1--13

    2023

  5. [13]

    Goodfellow, I.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Courville, A.; and Bengio, Y. 2020. Generative adversarial networks. Communications of the ACM, 63(11): 139--144

    2020

  6. [14]

    Gu, S.; Chen, D.; Bao, J.; Wen, F.; Zhang, B.; Chen, D.; Yuan, L.; and Guo, B. 2022. Vector quantized diffusion model for text-to-image synthesis. In CVPR, 10696--10706

    2022

  7. [15]

    L.; and Choi, Y

    Hessel, J.; Holtzman, A.; Forbes, M.; Bras, R. L.; and Choi, Y. 2021. CLIPScore: A Reference-free Evaluation Metric for Image Captioning. In EMNLP

    2021

  8. [16]

    Ho, J.; and Salimans, T. 2022. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598

    2022 arXiv

  9. [17]

    B.; Mattar, M.; Berg, T.; and Learned-Miller, E

    Huang, G. B.; Mattar, M.; Berg, T.; and Learned-Miller, E. 2008. Labeled faces in the wild: A database forstudying face recognition in unconstrained environments. In Workshop on faces in'Real-Life'Images: detection, alignment, and recognition

    2008

  10. [18]

    Karras, T.; Aila, T.; Laine, S.; and Lehtinen, J. 2018. Progressive growing of gans for improved quality, stability, and variation. In ICLR

    2018

  11. [19]

    Karras, T.; Laine, S.; and Aila, T. 2019. A style-based generator architecture for generative adversarial networks. In CVPR, 4401--4410

    2019

  12. [20]

    Karras, T.; Laine, S.; Aittala, M.; Hellsten, J.; Lehtinen, J.; and Aila, T. 2020. Analyzing and improving the image quality of stylegan. In CVPR, 8110--8119

    2020

  13. [21]

    P.; and Ba, J

    Kingma, D. P.; and Ba, J. 2015. Adam: A method for stochastic optimization. In ICLR

    2015

  14. [22]

    Kumari, N.; Zhang, B.; Zhang, R.; Shechtman, E.; and Zhu, J.-Y. 2023. Multi-concept customization of text-to-image diffusion. In CVPR, 1931--1941

    2023

  15. [23]

    Lee, J.; Cho, K.; and Kiela, D. 2019. Countering language drift via visual grounding. In EMNLP

    2019

  16. [24]

    Li, Z.; Cao, M.; Wang, X.; Qi, Z.; Cheng, M.-M.; and Shan, Y. 2024. Photomaker: Customizing realistic human photos via stacked id embedding. In CVPR, 8640--8650

    2024

  17. [25]

    Lu, Y.; Singhal, S.; Strub, F.; Courville, A.; and Pietquin, O. 2020. Countering language drift with seeded iterated learning. In ICML, 6437--6447. PMLR

    2020

  18. [26]

    L\"uddecke, T.; and Ecker, A. 2022. Image Segmentation Using Text and Image Prompts. In CVPR, 7086--7096

    2022

  19. [27]

    Nichol, A.; Dhariwal, P.; Ramesh, A.; Shyam, P.; Mishkin, P.; McGrew, B.; Sutskever, I.; and Chen, M. 2022. Glide: Towards photorealistic image generation and editing with text-guided diffusion models. In ICML, 16784--16804. PMLR

    2022

  20. [28]

    Nitzan, Y.; Aberman, K.; He, Q.; Liba, O.; Yarom, M.; Gandelsman, Y.; Mosseri, I.; Pritch, Y.; and Cohen-Or, D. 2022. Mystyle: A personalized generative prior. ACM TOG, 41(6): 1--10

    2022

  21. [29]

    OpenAI. 2022. ChatGPT. https://chat.openai.com/chat

    2022

  22. [30]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In ICML, 8748--8763. PMLR

    2021

  23. [31]

    Raffel, C.; Shazeer, N.; Roberts, A.; Lee, K.; Narang, S.; Matena, M.; Zhou, Y.; Li, W.; and Liu, P. J. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. JMLR, 21(1): 5485--5551

    2020

  24. [32]

    Ramesh, A.; Dhariwal, P.; Nichol, A.; Chu, C.; and Chen, M. 2022. Hierarchical text-conditional image generation with clip latents. arXiv preprint arXiv:2204.06125

    2022 arXiv

  25. [33]

    Ramesh, A.; Pavlov, M.; Goh, G.; Gray, S.; Voss, C.; Radford, A.; Chen, M.; and Sutskever, I. 2021. Zero-shot text-to-image generation. In ICML, 8821--8831. PMLR

    2021

  26. [34]

    H.; and Cohen-Or, D

    Roich, D.; Mokady, R.; Bermano, A. H.; and Cohen-Or, D. 2022. Pivotal tuning for latent-based editing of real images. ACM TOG, 42(1): 1--13

    2022

  27. [35]

    Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. In CVPR, 10684--10695

    2022

  28. [36]

    Ruiz, N.; Li, Y.; Jampani, V.; Pritch, Y.; Rubinstein, M.; and Aberman, K. 2023. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In CVPR, 22500--22510

    2023

  29. [37]

    L.; Ghasemipour, K.; Gontijo Lopes, R.; Karagol Ayan, B.; Salimans, T.; et al

    Saharia, C.; Chan, W.; Saxena, S.; Li, L.; Whang, J.; Denton, E. L.; Ghasemipour, K.; Gontijo Lopes, R.; Karagol Ayan, B.; Salimans, T.; et al. 2022. Photorealistic text-to-image diffusion models with deep language understanding. In NeurIPS, 36479--36494

    2022

  30. [38]

    Song, H.; Du, Y.; Xiang, T.; Dong, J.; Qin, J.; and He, S. 2022. Editing out-of-domain gan inversion via differential activations. In ECCV, 1--17. Springer

    2022

  31. [39]

    Song, J.; Meng, C.; and Ermon, S. 2021. Denoising diffusion implicit models. In ICLR

    2021

  32. [40]

    Song, Y.; and Ermon, S. 2019. Generative modeling by estimating gradients of the data distribution. In NeurIPS, 11895--11907

    2019

  33. [41]

    P.; Kumar, A.; Ermon, S.; and Poole, B

    Song, Y.; Sohl-Dickstein, J.; Kingma, D. P.; Kumar, A.; Ermon, S.; and Poole, B. 2021. Score-based generative modeling through stochastic differential equations. In ICLR

    2021

  34. [42]

    Tang, Z.; Gu, S.; Bao, J.; Chen, D.; and Wen, F. 2022. Improved vector quantized diffusion models. arXiv preprint arXiv:2205.16007

    2022 arXiv

  35. [43]

    Tewel, Y.; Gal, R.; Chechik, G.; and Atzmon, Y. 2023. Key-locked rank one editing for text-to-image personalization. In ACM SIGGRAPH, 1--11

    2023

  36. [44]

    von Platen, P.; Patil, S.; Lozhkov, A.; Cuenca, P.; Lambert, N.; Rasul, K.; Davaadorj, M.; and Wolf, T. 2022. Diffusers: State-of-the-art diffusion models. https://github.com/huggingface/diffusers

    2022

  37. [45]

    Wang, H.; Wang, Y.; Zhou, Z.; Ji, X.; Gong, D.; Zhou, J.; Li, Z.; and Liu, W. 2018. Cosface: Large margin cosine loss for deep face recognition. In CVPR, 5265--5274

    2018

  38. [46]

    Xu, Y.; Du, Y.; Xiao, W.; Xu, X.; and He, S. 2021. From continuity to editability: Inverting gans with consecutive images. In ICCV, 13910--13918

    2021

  39. [47]

    Yang, H.; Chai, L.; Wen, Q.; Zhao, S.; Sun, Z.; and He, S. 2021. Discovering interpretable latent space directions of gans beyond binary attributes. In CVPR, 12177--12185

    2021

  40. [48]

    Ye, H.; Zhang, J.; Liu, S.; Han, X.; and Yang, W. 2023. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint arXiv:2308.06721

    2023 arXiv

  41. [49]

    Y.; Luong, T.; Baid, G.; Wang, Z.; Vasudevan, V.; Ku, A.; Yang, Y.; Ayan, B

    Yu, J.; Xu, Y.; Koh, J. Y.; Luong, T.; Baid, G.; Wang, Z.; Vasudevan, V.; Ku, A.; Yang, Y.; Ayan, B. K.; Hutchinson, B.; Han, W.; Parekh, Z.; Li, X.; Zhang, H.; Baldridge, J.; and Wu, Y. 2022. Scaling Autoregressive Models for Content-Rich Text-to-Image Generation. TMLR

    2022

  42. [50]

    Zhang, K.; Zhang, Z.; Li, Z.; and Qiao, Y. 2016. Joint face detection and alignment using multitask cascaded convolutional networks. IEEE signal processing letters, 23(10): 1499--1503

    2016

  43. [51]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  44. [52]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.