arxiv: 2605.00883 · v1 · submitted 2026-04-27 · 💻 cs.CV · cs.AI

Recognition: unknown

Towards High Fidelity Face Swapping: A Comprehensive Survey and New Benchmark

Qi Li , Weining Wang , Shuangjun Du , Bo Peng , Jing Dong , Kun Wang , Zhenan Sun , Ming-Hsuan Yang

Authors on Pith no claims yet

Pith reviewed 2026-05-09 19:56 UTC · model grok-4.3

classification 💻 cs.CV cs.AI

keywords face swappingsurveybenchmark datasetGANdiffusion modelsimage manipulationdeepfakescomputer vision

0 comments

The pith

Face swapping methods fall into five paradigms that require a balanced demographic benchmark and fixed protocols for fair comparison.

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

Face swapping replaces one face with another in images or videos using generative models such as GANs and diffusion models, yet approaches remain scattered and evaluations vary widely. The paper groups existing techniques into five major paradigms based on their core design choices for transferring identity and attributes. This organization makes it possible to compare principles, strengths, and weaknesses across groups in a systematic way. To support reliable testing, the authors release a new dataset with even demographic coverage and controlled variations in factors like age, pose, and expression, plus standard evaluation rules. Experiments on sample methods from each paradigm then expose consistent patterns in where current techniques succeed or fail.

Core claim

Existing face swapping methods can be organized into five major paradigms whose design principles, strengths, and limitations become visible through systematic review; a new high-quality benchmark dataset with balanced demographics and explicit attribute variations, together with standardized evaluation protocols, enables consistent robustness assessment and yields new performance insights.

What carries the argument

The five-paradigm taxonomy that classifies face swapping methods by design principles, paired with the CASIA FaceSwapping benchmark dataset and its standardized evaluation protocols.

If this is right

Methods from different paradigms display distinct trade-offs when handling identity preservation versus attribute changes such as lighting or expression.
Standardized protocols allow direct, reproducible comparisons that reduce conflicting claims across papers.
The benchmark reveals performance gaps tied to demographic factors, guiding targeted improvements.
The unified view supports development of face swapping systems that combine strengths from multiple paradigms.

Where Pith is reading between the lines

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

The same paradigm-based organization could be applied to neighboring tasks such as face reenactment or editing to reduce fragmentation.
Extending the benchmark to video sequences would test whether current protocols capture temporal consistency issues.
Insights on demographic biases could feed into detection systems that identify swapped content more reliably across populations.

Load-bearing premise

The five-paradigm classification covers all significant face swapping techniques without major gaps or overlaps, and the new benchmark plus protocols give a representative sample of real-world conditions.

What would settle it

A newly published face swapping technique that fits none of the five paradigms, or benchmark results that reverse the performance ordering seen on other datasets, would show the framework does not hold.

Figures

Figures reproduced from arXiv: 2605.00883 by Bo Peng, Jing Dong, Kun Wang, Ming-Hsuan Yang, Qi Li, Shuangjun Du, Weining Wang, Zhenan Sun.

**Figure 2.** Figure 2: Overview of the proposed FSGAN approach. (a) The recurrent reenactment generator Fig4: The flowchart of FSGAN [37] [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 5.** Figure 5: Illustration of style transfer–based methods. (a) Latent [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗

**Figure 8.** Figure 8: Normalized radar charts of face swapping methods across the proposed protocols: (a) Normal, (b) Cross-ethnicity, [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

**Figure 9.** Figure 9: Line chart of face swapping methods across the proposed protocols for different metrics: (a) ID retrieval, (b) ID similarity, [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗

**Figure 10.** Figure 10: Qualitative results of face swapping methods across different protocols. (a) Normal protocol (top four rows), (b) [PITH_FULL_IMAGE:figures/full_fig_p014_10.png] view at source ↗

read the original abstract

Face swapping has witnessed significant progress in recent years, largely driven by advances in deep generative models such as GANs and diffusion models.Despite these advances, existing methods remain fragmented across different paradigms, and their evaluation is highly inconsistent due to the lack of standardized datasets and protocols. Moreover, prior surveys primarily focus on broader deepfake generation or detection, leaving face swapping insufficiently studied as a standalone problem. In this paper, we present a comprehensive survey and benchmark for face swapping. We provide a structured review of existing methods, organizing them into five major paradigms and systematically analyzing their design principles, strengths, and limitations. To enable fair and controlled evaluation, we introduce CASIA FaceSwapping, a high-quality benchmark with balanced demographic distributions and explicit attribute variations, and establish standardized protocols to assess the robustness of different face swapping methods. Extensive experiments on representative approaches yield new insights into the performance characteristics and limitations of current techniques. Overall, our work provides a unified perspective and a principled evaluation framework to facilitate the development of more robust and controllable face swapping methods. More results can be found at https://github.com/CASIA-NLPRAI/face-swapping-survey.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The new CASIA benchmark with demographic balance and protocols is the useful part; the five-paradigm survey is mostly a reasonable map of existing work.

read the letter

The main things to know are that this paper groups face swapping methods into five paradigms and releases the CASIA FaceSwapping benchmark with balanced demographics, attribute variations, and standardized evaluation protocols. It also runs experiments on representative methods to surface performance patterns and limits. That combination addresses the inconsistent testing that has slowed the area. Prior surveys stayed broader on deepfakes, so narrowing to swapping and adding a concrete dataset is the concrete step forward. The analysis of design principles, strengths, and limitations for each paradigm gives readers a workable overview without forcing new theory. The experiments produce usable insights on where current techniques still struggle, especially under controlled attribute shifts. The benchmark construction looks careful on paper, with explicit balancing that should reduce obvious demographic bias in comparisons. Soft spots are limited. The five-paradigm split is defensible but still somewhat arbitrary; some methods could fit more than one category, and the inclusion rules are not ironclad. The results section depends on the chosen representatives, so the insights are only as broad as that selection. Real-world robustness claims would be stronger with more discussion of how the dataset covers rare edge cases. Overall this is aimed at computer vision and generative-model researchers who need a common testbed for swapping methods or who want to cite a recent structured review. Readers building or evaluating new techniques will get direct value from the protocols and dataset. It deserves peer review. The benchmark and protocols are a practical contribution that others can adopt, so the paper is worth referee time even if the taxonomy needs minor tightening.

Referee Report

0 major / 2 minor

Summary. The paper presents a comprehensive survey of face swapping techniques, organizing existing methods into five major paradigms and analyzing their design principles, strengths, and limitations. It introduces the CASIA FaceSwapping benchmark dataset, which features balanced demographic distributions and explicit attribute variations, along with standardized evaluation protocols. The work includes extensive experiments on representative methods to provide insights into current techniques' performance and limitations, with the goal of establishing a unified perspective and principled framework for future research in high-fidelity face swapping.

Significance. If the taxonomy is comprehensive without major omissions or overlaps and the benchmark construction is representative, this work would provide a valuable unified evaluation framework for face swapping research. The introduction of a demographically balanced dataset with attribute controls and standardized protocols directly addresses fragmentation in prior evaluations, while the reported experiments on representative methods offer concrete insights into robustness. Credit is due for the new benchmark, public GitHub resources, and the focus on real-world variations.

minor comments (2)

Abstract: the five paradigms are referenced but not named or briefly characterized, which would help readers immediately grasp the taxonomy structure.
The manuscript should include an explicit table or section listing the inclusion criteria used to assign methods to each of the five paradigms, to allow independent verification of the taxonomy boundaries.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary of our survey, the recognition of the CASIA FaceSwapping benchmark, and the recommendation for minor revision. We appreciate the emphasis on the value of a unified evaluation framework.

Circularity Check

0 steps flagged

No significant circularity; purely descriptive survey and benchmark

full rationale

The paper is a survey that organizes existing face-swapping methods into five paradigms based on explicit inclusion criteria and reviews their design principles without any derivations, equations, fitted parameters, or predictions. The new CASIA FaceSwapping benchmark is introduced with documented construction details (demographic balancing, attribute annotations, protocols) that stand as independent contributions rather than reductions to prior inputs. No self-citation chains, ansatzes, or uniqueness theorems are invoked to justify core claims; all taxonomy boundaries and evaluation standards are defined directly in the manuscript. This matches the default expectation for non-circular survey papers.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

This is a survey and benchmark paper whose claims rest on the completeness of the literature review and the representativeness of the new dataset. No free parameters, mathematical axioms, or invented physical entities are introduced.

axioms (1)

domain assumption Advances in deep generative models such as GANs and diffusion models have driven recent progress in face swapping.
Stated directly in the opening sentence of the abstract as background.

pith-pipeline@v0.9.0 · 5522 in / 1316 out tokens · 76867 ms · 2026-05-09T19:56:10.523640+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

108 extracted references · 10 canonical work pages · 1 internal anchor

[1]

The creation and detection of deepfakes: A survey,

Y . Mirsky and W. Lee, “The creation and detection of deepfakes: A survey,”ACM Computing Surveys, vol. 54, no. 1, pp. 1–41, 2021. 2 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 15 TABLE VI: Temporal consistency comparison of face swapping methods on video clips. Subject consistency and background consistency are computed to measure frame...

2021
[2]

Deep learning for deepfakes creation and detection: A survey,

T. T. Nguyen, Q. V . H. Nguyen, D. T. Nguyen, D. T. Nguyen, T. Huynh- The, S. Nahavandi, T. T. Nguyen, Q.-V . Pham, and C. M. Nguyen, “Deep learning for deepfakes creation and detection: A survey,”CVIU, vol. 223, p. 103525, 2022. 2

2022
[3]

Countering malicious deepfakes: Survey, battleground, and horizon,

F. Juefei-Xu, R. Wang, Y . Huang, Q. Guo, L. Ma, and Y . Liu, “Countering malicious deepfakes: Survey, battleground, and horizon,” IJCV, vol. 130, no. 7, pp. 1678–1734, 2022. 2

2022
[4]

Deepfakes generation and detection: State-of-the-art, open challenges, countermeasures, and way forward,

M. Masood, M. Nawaz, K. M. Malik, A. Javed, A. Irtaza, and H. Malik, “Deepfakes generation and detection: State-of-the-art, open challenges, countermeasures, and way forward,”Applied Intelligence, vol. 53, no. 4, pp. 3974–4026, 2023. 2

2023
[5]

Deepfake detection: A comprehensive survey from the reliability perspective,

T. Wang, X. Liao, K. P. Chow, X. Lin, and Y . Wang, “Deepfake detection: A comprehensive survey from the reliability perspective,” ACM Computing Surveys, vol. 57, no. 3, pp. 1–35, 2024. 2

2024
[6]

Deepfake generation and detection: A benchmark and survey,

G. Pei, J. Zhang, M. Hu, Z. Zhang, C. Wang, Y . Wu, G. Zhai, J. Yang, and D. Tao, “Deepfake generation and detection: A benchmark and survey,”ACM Computing Surveys, 2026. 2

2026
[7]

One shot face swapping on megapixels,

Y . Zhu, Q. Li, J. Wang, C.-Z. Xu, and Z. Sun, “One shot face swapping on megapixels,” inCVPR, 2021, pp. 4834–4844. 2, 6, 10, 11, 15

2021
[8]

Stargan: Unified generative adversarial networks for multi-domain image-to- image translation,

Y . Choi, M. Choi, M. Kim, J.-W. Ha, S. Kim, and J. Choo, “Stargan: Unified generative adversarial networks for multi-domain image-to- image translation,” inCVPR, 2018, pp. 8789–8797. 2

2018
[9]

A style-based generator architecture for generative adversarial networks,

T. Karras, S. Laine, and T. Aila, “A style-based generator architecture for generative adversarial networks,” inCVPR, 2019, pp. 4401–4410. 2, 3

2019
[10]

Semantic-aware one-shot face re- enactment with dense correspondence estimation,

Y . Liu, Q. Li, Z. Sun, and T. Tan, “Semantic-aware one-shot face re- enactment with dense correspondence estimation,”MIR, 2024. 2

2024
[11]

Analyzing and improving the image quality of stylegan,

T. Karras, S. Laine, M. Aittala, J. Hellsten, J. Lehtinen, and T. Aila, “Analyzing and improving the image quality of stylegan,” inCVPR, 2020, pp. 8110–8119. 3

2020
[12]

Alias-free generative adversarial networks,

T. Karras, M. Aittala, S. Laine, E. H ¨ark¨onen, J. Hellsten, J. Lehtinen, and T. Aila, “Alias-free generative adversarial networks,”NeurIPS, vol. 34, pp. 852–863, 2021. 3

2021
[13]

Image2stylegan: How to embed images into the stylegan latent space?

R. Abdal, Y . Qin, and P. Wonka, “Image2stylegan: How to embed images into the stylegan latent space?” inICCV, 2019, pp. 4432–4441. 3

2019
[14]

Image2stylegan++: How to edit the embedded images?

——, “Image2stylegan++: How to edit the embedded images?” in CVPR, 2020, pp. 8296–8305. 3

2020
[15]

Stylespace analysis: Disen- tangled controls for stylegan image generation,

Z. Wu, D. Lischinski, and E. Shechtman, “Stylespace analysis: Disen- tangled controls for stylegan image generation,” inCVPR, 2021, pp. 12 863–12 872. 3

2021
[16]

Towards spatially disentangled manipulation of face images with pre-trained stylegans,

Y . Liu, Q. Li, Q. Deng, and Z. Sun, “Towards spatially disentangled manipulation of face images with pre-trained stylegans,”TCSVT, vol. 33, no. 4, pp. 1725–1739, 2022. 3

2022
[17]

Denoising diffusion probabilistic models,

J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,”NeurIPS, vol. 33, pp. 6840–6851, 2020. 3, 6

2020
[18]

A morphable model for the synthesis of 3d faces,

V . Blanz and T. Vetter, “A morphable model for the synthesis of 3d faces,” inSeminal Graphics Papers: Pushing the Boundaries, 2023, pp. 157–164. 3

2023
[19]

Exchanging faces in images,

V . Blanz, K. Scherbaum, T. Vetter, and H.-P. Seidel, “Exchanging faces in images,” inComputer Graphics Forum, vol. 23, no. 3, 2004, pp. 669–676. 3

2004
[20]

Reanimating faces in images and video,

V . Blanz, C. Basso, T. Poggio, and T. Vetter, “Reanimating faces in images and video,” inComputer Graphics Forum, vol. 22, no. 3, 2003, pp. 641–650. 3

2003
[21]

3d-model-based face replacement in video,

Y .-T. Cheng, V . Tzeng, Y . Liang, C.-C. Wang, B.-Y . Chen, Y .-Y . IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 16 Chuang, and M. Ouhyoung, “3d-model-based face replacement in video,” inSIGGRAPH, 2009, pp. 1–1. 3

2009
[22]

Pose-free face swapping based on a deformable 3d shape morphable model,

Y . Lin and S. Wang, “Pose-free face swapping based on a deformable 3d shape morphable model,”IEICE Transactions on Information and Systems, vol. 97, no. 2, pp. 305–314, 2014. 3

2014
[23]

Face swapping under large pose variations: A 3d model based approach,

Y . Lin, S. Wang, Q. Lin, and F. Tang, “Face swapping under large pose variations: A 3d model based approach,” inICME, 2012, pp. 333–338. 3

2012
[24]

A unified framework for high fidelity face swap and expression reenactment,

B. Peng, H. Fan, W. Wang, J. Dong, and S. Lyu, “A unified framework for high fidelity face swap and expression reenactment,”TCSVT, vol. 32, no. 6, pp. 3673–3684, 2021. 3

2021
[25]

On face segmentation, face swapping, and face perception,

Y . Nirkin, I. Masi, A. T. Tuan, T. Hassner, and G. Medioni, “On face segmentation, face swapping, and face perception,” inFG, 2018, pp. 98–105. 3

2018
[26]

Hififace: 3d shape and semantic prior guided high fidelity face swapping,

Y . Wang, X. Chen, J. Zhu, W. Chu, Y . Tai, C. Wang, J. Li, Y . Wu, F. Huang, and R. Ji, “Hififace: 3d shape and semantic prior guided high fidelity face swapping,” inIJCAI, 2021, pp. 1136–1142. 3, 10, 11, 15

2021
[27]

Arbitrary style transfer in real-time with adaptive instance normalization,

X. Huang and S. Belongie, “Arbitrary style transfer in real-time with adaptive instance normalization,” inICCV, 2017, pp. 1501–1510. 4, 5

2017
[28]

Learning dynamic 3d geometry and texture for video face swapping,

C. Otto, J. Naruniec, L. Helminger, T. Etterlin, G. Mignone, P. Chan- dran, G. Zoss, C. Schroers, M. Gross, P. Gotardoet al., “Learning dynamic 3d geometry and texture for video face swapping,” inCom- puter Graphics Forum, vol. 41, no. 7, 2022, pp. 611–622. 4

2022
[29]

3d-aware face swapping,

Y . Li, C. Ma, Y . Yan, W. Zhu, and X. Yang, “3d-aware face swapping,” inCVPR, 2023, pp. 12 705–12 714. 4

2023
[30]

Efficient geometry-aware 3d generative adversarial networks,

E. R. Chan, C. Z. Lin, M. A. Chan, K. Nagano, B. Pan, S. De Mello, O. Gallo, L. J. Guibas, J. Tremblay, S. Khamiset al., “Efficient geometry-aware 3d generative adversarial networks,” inCVPR, 2022, pp. 16 123–16 133. 4

2022
[31]

DeepFakes, https://github.com/ondyari/FaceForensics/tree/master/datas et/DeepFakes, accessed: 2026-04-15. 4, 8

2026
[32]

Fast face-swap using convolutional neural networks,

I. Korshunova, W. Shi, J. Dambre, and L. Theis, “Fast face-swap using convolutional neural networks,” inICCV, 2017, pp. 3677–3685. 4

2017
[33]

Introducing explicit gaze constraints to face swapping,

E. Wilson, F. Shic, and E. Jain, “Introducing explicit gaze constraints to face swapping,” inProceedings of the Symposium on Eye Tracking Research and Applications, 2023, pp. 1–7. 4

2023
[34]

High fidelity face swapping via semantics disentanglement and structure enhancement,

F. Liu, L. Yu, H. Xie, C. Liu, Z. Ding, Q. Yang, and Y . Zhang, “High fidelity face swapping via semantics disentanglement and structure enhancement,” inMM, 2023, pp. 6907–6917. 4

2023
[35]

Stableswap: stable face swapping in a shared and controllable latent space,

Y . Zhu, W. Zhao, Y . Tang, Y . Rao, J. Zhou, and J. Lu, “Stableswap: stable face swapping in a shared and controllable latent space,”TMM,
[36]

FaceSwap-GAN, https://github.com/shaoanlu/faceswap- GAN, accessed: 2026-04-15. 4, 8

2026
[37]

Fsgan: Subject agnostic face swapping and reenactment,

Y . Nirkin, Y . Keller, and T. Hassner, “Fsgan: Subject agnostic face swapping and reenactment,” inICCV, 2019, pp. 7184–7193. 4, 10, 11, 15

2019
[38]

Fsganv2: Improved subject agnostic face swapping and reen- actment,

——, “Fsganv2: Improved subject agnostic face swapping and reen- actment,”TPAMI, vol. 45, no. 1, pp. 560–575, 2022. 4, 10

2022
[39]

Few-shot head swapping in the wild,

C. Shu, H. Wu, H. Zhou, J. Liu, Z. Hong, C. Ding, J. Han, J. Liu, E. Ding, and J. Wang, “Few-shot head swapping in the wild,” inCVPR, 2022, pp. 10 789–10 798. 4

2022
[40]

Rsgan: face swapping and editing using face and hair representation in latent spaces,

R. Natsume, T. Yatagawa, and S. Morishima, “Rsgan: face swapping and editing using face and hair representation in latent spaces,” in SIGGRAPH, 2018, pp. 1–2. 4

2018
[41]

An image is worth 16x16 words: transformers for image recognition at scale,

A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: transformers for image recognition at scale,” inICLR, 2021. 4

2021
[42]

Transfs: Face swapping using transformer,

W. Cao, T. Wang, A. Dong, and M. Shu, “Transfs: Face swapping using transformer,” inFG, 2023, pp. 1–8. 4

2023
[43]

Face transformer: Towards high fidelity and accurate face swapping,

K. Cui, R. Wu, F. Zhan, and S. Lu, “Face transformer: Towards high fidelity and accurate face swapping,” inCVPR Workshop, 2023, pp. 668–677. 4

2023
[44]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” NeurIPS, vol. 30, 2017. 5

2017
[45]

Simswap: An efficient framework for high fidelity face swapping,

R. Chen, X. Chen, B. Ni, and Y . Ge, “Simswap: An efficient framework for high fidelity face swapping,” inMM, 2020, pp. 2003–2011. 5, 10, 11, 15

2020
[46]

Simswap++: Towards faster and high-quality identity swapping,

X. Chen, B. Ni, Y . Liu, N. Liu, Z. Zeng, and H. Wang, “Simswap++: Towards faster and high-quality identity swapping,”TPAMI, vol. 46, no. 1, pp. 576–592, 2024. 5, 10

2024
[47]

Designing one unified framework for high-fidelity face reenactment and swapping,

C. Xu, J. Zhang, Y . Han, G. Tian, X. Zeng, Y . Tai, Y . Wang, C. Wang, and Y . Liu, “Designing one unified framework for high-fidelity face reenactment and swapping,” inECCV, 2022, pp. 54–71. 5

2022
[48]

Flowface: Semantic flow-guided shape-aware face swapping,

H. Zeng, W. Zhang, C. Fan, T. Lv, S. Wang, Z. Zhang, B. Ma, L. Li, Y . Ding, and X. Yu, “Flowface: Semantic flow-guided shape-aware face swapping,” inAAAI, vol. 37, no. 3, 2023, pp. 3367–3375. 5

2023
[49]

Flowface++: Explicit semantic flow-supervised end-to- end face swapping,

Y . Zhang, H. Zeng, B. Ma, W. Zhang, Z. Zhang, Y . Ding, T. Lv, and C. Fan, “Flowface++: Explicit semantic flow-supervised end-to- end face swapping,”arXiv preprint arXiv:2306.12686, 2023. 5

work page arXiv 2023
[50]

Selfswapper: Self-supervised face swapping via shape agnostic masked autoencoder,

J. Lee, J. Hyung, S. Jung, and J. Choo, “Selfswapper: Self-supervised face swapping via shape agnostic masked autoencoder,” inECCV, 2024, pp. 383–400. 5

2024
[51]

Advancing high fidelity identity swapping for forgery detection,

L. Li, J. Bao, H. Yang, D. Chen, and F. Wen, “Advancing high fidelity identity swapping for forgery detection,” inCVPR, 2020, pp. 5074–

2020
[52]

Rsface: subject agnostic face swapping with expression high fidelity,

G. Yang, T. Wang, X. Fang, and j. Zhang, “Rsface: subject agnostic face swapping with expression high fidelity,”The Visual Computer, vol. 39, no. 11, pp. 5497–5511, 2023. 5

2023
[53]

Paste you into game: Towards expression and identity consistency face swapping,

H. Zeng, W. Zhang, K. Chen, Z. Zhang, L. Li, and Y . Ding, “Paste you into game: Towards expression and identity consistency face swapping,” inIEEE Conference on Games, 2022, pp. 1–8. 5

2022
[54]

Learning disentan- gled representation for one-shot progressive face swapping,

Q. Li, W. Wang, C. Xu, Z. Sun, and M.-H. Yang, “Learning disentan- gled representation for one-shot progressive face swapping,”TPAMI, vol. 46, no. 12, pp. 8348–8364, 2024. 5

2024
[55]

Blendface: Re-designing identity encoders for face-swapping,

K. Shiohara, X. Yang, and T. Taketomi, “Blendface: Re-designing identity encoders for face-swapping,” inICCV, 2023, pp. 7634–7644. 5, 10, 11, 15

2023
[56]

Information bottleneck disentanglement for identity swapping,

G. Gao, H. Huang, C. Fu, Z. Li, and R. He, “Information bottleneck disentanglement for identity swapping,” inCVPR, 2021, pp. 3404–

2021
[57]

Identity-preserving face swapping via dual surrogate generative models,

Z. Huang, F. Tang, Y . Zhang, J. Cao, C. Li, S. Tang, J. Li, and T.-Y . Lee, “Identity-preserving face swapping via dual surrogate generative models,”TOG, vol. 43, no. 5, pp. 1–19, 2024. 5, 10, 11, 15

2024
[58]

Facedancer: Pose-and occlusion-aware high fidelity face swapping,

F. Rosberg, E. E. Aksoy, F. Alonso-Fernandez, and C. Englund, “Facedancer: Pose-and occlusion-aware high fidelity face swapping,” inWACV, 2023, pp. 3454–3463. 5, 10, 11, 15

2023
[59]

Reinforced disentanglement for face swapping without skip connection,

X. Ren, X. Chen, P. Yao, H.-Y . Shum, and B. Wang, “Reinforced disentanglement for face swapping without skip connection,” inICCV, 2023, pp. 20 665–20 675. 5

2023
[60]

Smooth-swap: A simple enhancement for face-swapping with smoothness,

J. Kim, J. Lee, and B.-T. Zhang, “Smooth-swap: A simple enhancement for face-swapping with smoothness,” inCVPR, 2022, pp. 10 779– 10 788. 5

2022
[61]

Fastswap: A lightweight one-stage framework for real-time face swapping,

S.-M. Yoo, T.-M. Choi, J.-W. Choi, and J.-H. Kim, “Fastswap: A lightweight one-stage framework for real-time face swapping,” in WACV, 2023, pp. 3558–3567. 5

2023
[62]

Mobilefaceswap: A lightweight framework for video face swapping,

Z. Xu, Z. Hong, C. Ding, Z. Zhu, J. Han, J. Liu, and E. Ding, “Mobilefaceswap: A lightweight framework for video face swapping,” inAAAI, vol. 36, no. 3, 2022, pp. 2973–2981. 5

2022
[63]

Phinet-gan: Bringing real-time face swapping to embedded devices,

A. Ancilotto, F. Paissan, and E. Farella, “Phinet-gan: Bringing real-time face swapping to embedded devices,” inIEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events, 2023, pp. 677–682. 5

2023
[64]

La- tentswap: An efficient latent code mapping framework for face swap- ping,

C. Choi, M. Kim, J. Lee, H.-K. Song, Y . Kim, and S. Kim, “La- tentswap: An efficient latent code mapping framework for face swap- ping,”arXiv preprint arXiv:2402.18351, 2024. 6

work page arXiv 2024
[65]

High fidelity face- swapping with style convtransformer and latent space selection,

J. Yang, C. Cheng, S. Xiao, G. Lan, and J. Wen, “High fidelity face- swapping with style convtransformer and latent space selection,”TMM,
[66]

End-to-end face-swapping via adap- tive latent representation learning,

C. Lin, P. Hu, C. Shen, and Q. Li, “End-to-end face-swapping via adap- tive latent representation learning,”arXiv preprint arXiv:2303.04186,

work page arXiv
[67]

High-resolution face swapping via latent semantics disentanglement,

Y . Xu, B. Deng, J. Wang, Y . Jing, J. Pan, and S. He, “High-resolution face swapping via latent semantics disentanglement,” inCVPR, 2022, pp. 7642–7651. 6, 10, 11, 15

2022
[68]

Style- face: Towards identity-disentangled face generation on megapixels,

Y . Luo, J. Zhu, K. He, W. Chu, Y . Tai, C. Wang, and J. Yan, “Style- face: Towards identity-disentangled face generation on megapixels,” in ECCV, 2022, pp. 297–312. 6

2022
[69]

Fine-grained face swapping via regional gan inversion,

Z. Liu, M. Li, Y . Zhang, C. Wang, Q. Zhang, J. Wang, and Y . Nie, “Fine-grained face swapping via regional gan inversion,” inCVPR, 2023, pp. 8578–8587. 6, 10, 11, 15

2023
[70]

Region-aware face swapping,

C. Xu, J. Zhang, M. Hua, Q. He, Z. Yi, and Y . Liu, “Region-aware face swapping,” inCVPR, 2022, pp. 7632–7641. 6, 10, 11, 15

2022
[71]

Styleswap: Style-based generator empowers robust face swapping,

Z. Xu, H. Zhou, Z. Hong, Z. Liu, J. Liu, Z. Guo, J. Han, J. Liu, E. Ding, and J. Wang, “Styleswap: Style-based generator empowers robust face swapping,” inECCV, 2022, pp. 661–677. 6

2022
[72]

Controlface: Feature disentangling for controllable face swapping,

X. Zhang, W. Zhou, K. Liu, H. Tang, Z. Zhang, W. Zhang, and N. Yu, “Controlface: Feature disentangling for controllable face swapping,” Journal of Imaging, vol. 10, no. 1, pp. 1–16, 2024. 6 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 17

2024
[73]

Styleipsb: Identity- preserving semantic basis of stylegan for high fidelity face swapping,

D. Jiang, D. Song, R. Tong, and M. Tang, “Styleipsb: Identity- preserving semantic basis of stylegan for high fidelity face swapping,” inCVPR, 2023, pp. 352–361. 6

2023
[74]

Mfim: Megapixel facial identity manipulation,

S. Na, “Mfim: Megapixel facial identity manipulation,” inECCV, 2022, pp. 143–159. 6

2022
[75]

Diffface: Diffusion-based face swapping with facial guidance,

K. Kim, Y . Kim, S. Cho, J. Seo, J. Nam, K. Lee, S. Kim, and K. Lee, “Diffface: Diffusion-based face swapping with facial guidance,”arXiv preprint arXiv:2212.13344, 2022. 7

work page arXiv 2022
[76]

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

Y . Han, J. Zhu, K. He, X. Chen, Y . Ge, W. Li, X. Li, J. Zhang, C. Wang, and Y . Liu, “Face-adapter for pre-trained diffusion models with fine- grained id and attribute control,” inECCV, 2024, pp. 20–36. 7, 10, 11, 15

2024
[77]

Diffswap: High- fidelity and controllable face swapping via 3d-aware masked diffusion,

W. Zhao, Y . Rao, W. Shi, Z. Liu, J. Zhou, and J. Lu, “Diffswap: High- fidelity and controllable face swapping via 3d-aware masked diffusion,” inCVPR, 2023, pp. 8568–8577. 7, 10, 11, 15

2023
[78]

Realistic and efficient face swapping: A unified approach with diffusion models,

S. Baliah, Q. Lin, S. Liao, X. Liang, and M. H. Khan, “Realistic and efficient face swapping: A unified approach with diffusion models,” in WACV, 2025, pp. 1062–1071. 7

2025
[79]

Uniface++: Revisiting a unified framework for face reenactment and swapping via 3d priors,

C. Xu, Y . Qian, S. Zhu, B. Sun, J. Zhao, Y . Liu, and X. Li, “Uniface++: Revisiting a unified framework for face reenactment and swapping via 3d priors,”IJCV, vol. 133, no. 7, pp. 4538–4554, 2025. 7

2025
[80]

Hs-diffusion: Semantic-mixing diffusion for head swapping,

Q. Wang, L. Liu, M. Hua, P. Zhu, W. Zuo, Q. Hu, H. Lu, and B. Cao, “Hs-diffusion: Semantic-mixing diffusion for head swapping,”arXiv preprint arXiv:2212.06458, 2022. 7

work page arXiv 2022

Showing first 80 references.