Pith. sign in

REVIEW 3 major objections 4 minor 68 references

Deepfakes on Demand: the rise of accessible non-consensual deepfake image generators

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that over 34,000 publicly downloadable deepfake model variants are available, with nearly 15 million cumulative downloads, and that 96% of a labelled sample target women.

desk verdict First large-scale count of the public deepfake-model supply chain; headline precision rests on an unvalidated tag, but the core finding is solid and deserves serious review. read the letter →

arxiv 2505.03859 v1 pith:6G4QB6DJ submitted 2025-05-06 cs.CY cs.AIcs.CV

Will Hawkins , Chris Russell , Brent Mittelstadt This is my paper · ORCID
classification cs.CYcs.AIcs.CV
keywords deepfakesnon-consensualintimateimagerytext-to-imagemodelsLoRAfine-tuningmodel-hostingplatformsStableDiffusionFluxcontentmoderation
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 public supply of ready-made deepfake image generators is no longer a niche or marginal phenomenon. By reading metadata attached to model variants on two large public model-sharing platforms, it identifies more than 34,000 downloadable variants explicitly intended to produce images of identifiable people, with nearly 15 million cumulative downloads since late 2022. The paper further argues that the overwhelming majority of the carefully examined variants target women, that many signal intent to generate non-consensual intimate imagery, and that the cost of creating such a model has fallen to about 20 training images, 24GB of VRAM, and 15 minutes of fine-tuning time. If these numbers hold, the bottleneck for this kind of abuse is no longer technical skill or compute, but platform enforcement and legal intervention.

What carries the argument

The load-bearing object is the model variant itself: a fine-tuned text-to-image model adapted to reproduce a specific person's likeness. The paper treats creator-assigned metadata tags as a lens onto this population, with one platform's 'Celebrity' tag marking variants intended to depict identifiable individuals and additional tags and description terms serving as red flags for sexualised intent. The named mechanism that makes the phenomenon scalable is low-rank adaptation (LoRA), a parameter-efficient fine-tuning technique that updates only small adapter matrices while keeping the base model frozen, letting a user create a likeness-specific model with as few as 20 images, 24GB of VRAM, and roughly 15 minutes of compute. The analysis is carried by comparing these metadata signals across two model families and two platforms, then manually labelling names and descriptions in a 15,349-model sample to separate deepfake variants from other fine-tunes.

What would settle it

Take a random sample of, say, 400 of the 34,439 models carrying the creator-assigned 'Celebrity' tag and have independent annotators, working without the study's labelling rubric, check whether each model's name, description, and example images actually target an identifiable real person and whether any consent statement appears. If more than about 10% of the sample turns out to target fictional characters, non-photorealistic subjects, or public figures with documented consent, the 34,000 count and the 'non-consensual' framing would need to be revised downward.

Watch

Extended reading notes

Core claim

The paper's central discovery is that deepfake image generators have become a commodity available for direct download. Across the full platform census, 34,439 model variants carry a creator-assigned tag indicating they generate identifiable individuals, and those variants have been downloaded 14,908,183 times since November 2022. In a manually labelled sample of 2,083 deepfake variants from the Stable Diffusion and Flux model families, 96% target women, and 97 of the top 100 most downloaded variants target women. The authors interpret the absence of any consent statement in the examined model cards, together with tags and descriptions referencing sexualised or adult content, as evidence that many of these models are intended for non-consensual intimate imagery. The paper also finds that 80% of the tagged variants are LoRA adapters, that the release of Flux in August 2024 coincides with a sharp acceleration in uploads, and that by December 2024 deepfake-oriented LoRAs made up 44.3% of all Flux LoRA variants on the main platform examined.

Load-bearing premise

The count depends on models being tagged 'Celebrity' by their uploaders, and the 'non-consensual' label depends on consent being absent from model cards; if the tag is used loosely or consent is simply not documented, the headline numbers overstate the problem.

Editorial extensions

If this is right

  • If the counts are right, the supply side of non-consensual deepfake imagery is already industrialised: each of the 34,000-plus downloadable variants can generate an effectively unlimited number of images.
  • Creation barriers are low enough that removing any individual model is unlikely to stop production, because the same small image set and consumer GPU can be reused locally without any public upload.
  • The concentration of uploads among a small number of prolific creators means platform-level user bans could reduce public availability more effectively than per-model takedowns.
  • Because 96% of the manually labelled deepfake models target women, the measurement implies that the abuse burden of this technology falls almost entirely on women, from celebrities to low-follower social media users.
  • The jump in uploads after the release of the Flux model family suggests that future high-quality open text-to-image base models are likely to produce another step-change in deepfake model creation unless access or fine-tuning is constrained.

Reading between the lines

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

  • Editorial inference: the public platform census almost certainly understates total deepfake model production, because locally trained models and models shared through private channels are invisible to metadata analysis; the paper itself says as much.
  • Editorial inference: the same metadata method could be rerun quarterly as an early-warning indicator for new base models, and it could be extended to video-generation models if comparable creator tags emerge.
  • Editorial inference: if platform terms were revised to require a consent statement before hosting any model depicting a real person, the paper's finding that no examined model card contains one suggests the public stock would shrink sharply, though production might move into less visible spaces.
  • Editorial inference: the 96% figure describes the manually labelled sample, not the full 34,000, so extrapolating it to the whole population would require validating the creator-assigned tag on the full set.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. This paper reports an empirical, three-part study of publicly downloadable text-to-image model variants that can generate images of identifiable people. Using metadata from Civitai and Hugging Face, the authors identify 34,439 Civitai models tagged 'Celebrity' (treated as deepfake model variants), with roughly 14.9 million cumulative downloads. They manually label 15,349 Stable Diffusion and Flux model variants across both platforms, finding 2,083 deepfake models, of which 96.4% target women. The paper also analyzes temporal trends, Terms of Service, and accessibility, arguing that LoRA fine-tuning enables the creation of deepfake models with as few as 20 images, 24GB VRAM, and about 15 minutes, and that platform enforcement and regulation lag behind the growth of the phenomenon.

Significance. If the headline estimates are correct, this is an important empirical contribution to the deepfake and NCII literature: it quantifies the supply side of non-consensual deepfake generators, documents a sharp temporal increase coinciding with Flux's release, and provides reproducible code and API-based data collection. The study's strengths include the independent measurement against platform APIs, the decision not to generate images, the large manually labeled sample in Part B, and the policy-relevant framing. However, the paper's central count rests on a creator-assigned tag whose precision is not systematically established, and the 'non-consensual' label is inferred from the absence of consent references in model cards. These issues affect the abstract's headline claims and need to be fixed before the results can be fully relied upon.

major comments (3)
  1. [Section 3.1, Table 2] The headline count of 34,439 deepfake model variants rests entirely on Civitai's creator-assigned 'Celebrity' tag, but the validation reported in Section 3.1 is only described as an 'initial evaluation' with no sample size, criteria, or error rate. The manual labeling in Part B is not drawn from the Celebrity-tagged population, so it cannot validate the tag's precision. Since the tag is self-assigned by the same creators being studied, it may include stylized or fictional celebrity likenesses, 'famous people' compilations, or models miscategorized for discoverability, all of which would inflate the central count and the 14.9 million download figure. The authors should report a systematic validation: draw a random sample of Celebrity-tagged models, classify them independently against a pre-registered rubric that separates photorealistic identifiable-person models from other categories, and report precision, recall, and inter-rater reliability.
  2. [Section 3.2, Appendix C] The manual labeling of 15,349 model variants appears to be a single-pass procedure without inter-rater reliability or dual coding. The criteria in Appendix C involve judgment calls, such as deciding when a character name implies an identifiable person and when example images are 'photorealistic' rather than cartoon depictions. Especially for the 2,083 models classified as deepfakes, the lack of reported inter-rater reliability makes the 96.4% female-targeting statistic and the Flux deepfake share in Table 7 harder to interpret. The authors should report the number of coders, a random subsample coded independently, and agreement statistics (e.g., Cohen's kappa), or otherwise justify why a single pass is reliable.
  3. [Section 4.3.1] The paper repeatedly refers to these models as 'non-consensual deepfake model variants' and as models 'without consent,' but the support for non-consent is the absence of consent references in model cards. Absence of a consent statement is not equivalent to evidence of non-consent, particularly because model cards on these platforms do not have a structured consent field. The claim would be strengthened by reporting how many model cards were examined, whether any explicit consent statements were found, and by softening the wording from 'non-consensual' to 'no evidence of consent identified' where the data only support the latter. This affects the title, abstract, and policy conclusions, so it should be addressed before publication.
minor comments (4)
  1. [Appendix A, Table 11] Appendix A states 'the 34,440 Celebrity models' while the body and Table 2 consistently report 34,439; the count should be made consistent.
  2. [Section 4.3.2] The phrase 'posing a risk to public and non-public figures alika' appears to contain a typo ('alika' should likely be 'alike') and should be corrected.
  3. [Section 6, Limitations] The limitation that Hugging Face monthly downloads are not directly comparable to Civitai lifetime downloads is acknowledged, but the abstract and discussion still present aggregate download figures without this caveat; one sentence noting the non-comparability when citing the 15 million figure would improve precision.
  4. [Section 3.1] The paper does not state the date on which the Civitai API data were collected, beyond noting December 2024 in Figure 1; specifying the exact collection window would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical metadata measurement with no fitted parameters, no derivation chain, and no load-bearing self-citation; the tag-based operationalization is transparent and acknowledged as a limitation.

full rationale

This paper does not contain a derivation or prediction chain in which an output is constructed from its own inputs. Part A counts Civitai models carrying the creator-assigned 'Celebrity' tag; Part B independently manually labels 2,083 Flux and Stable Diffusion variants by name/description; Part C compares platform Terms of Service and accessibility evidence. None of these are fitted quantities, and none of the headline numbers are produced by a model, equation, or statistical procedure that embeds the target conclusion. The only arguably self-referential element is the operational definition in Section 3.1, where models tagged 'Celebrity' are 'considered to be deepfake model variants,' so the 34,439 count is by construction a count of tagged models. However, the paper states this definition explicitly rather than hiding it, and Section 6 openly acknowledges that reliance on user tagging 'could lead to false positives or negatives.' That is a measurement-validity limitation, not a circular derivation: the manual labeling in Part B does not depend on the tag, and the temporal, gender, download, and policy findings are independent observations of API metadata. Self-citations in the reference list, including the Imagen 3 paper on which one author appears, are not used to justify the central empirical claims. No circular step meeting the evidentiary standard of this review can be identified, so the score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no invented entities and performs no fitted-parameter derivation. The analytic choices that shape the results are the download thresholds, the keyword list for sexual content, and the proxy assumptions described above. The most consequential is the use of the 'Celebrity' tag and the inference from missing consent references to non-consent.

free parameters (3)
  • Hugging Face minimum monthly downloads = 10
    Chosen cutoff in Appendix B to exclude inactive models from Part B; changing it changes the analyzed subset and derived proportions.
  • Civitai minimum lifetime downloads = 250
    Chosen cutoff in Appendix B to exclude inactive models from Part B; does not affect the Part A 'Celebrity' count of 34,439.
  • Sexual model keyword list = nsfw; porn; sexy; babes; hentai; unsafe; xxx
    Hand-constructed keywords in Appendix C used to label 'Sexual Models'; directly influences the NCII-related estimates.
assumptions (4)
  • domain assumption Civitai's creator-assigned 'Celebrity' tag reliably identifies deepfake model variants intended to generate identifiable individuals.
    Section 3.1 states an 'initial evaluation' found the tag almost exclusively marks photorealistic identifiable people, but no quantitative validation of the full 34,439 models is reported.
  • domain assumption The absence of consent references in model cards implies the depicted individuals did not consent.
    Section 4.3.1 reports no consent references were found and concludes the models likely violate policy (b); the paper's 'non-consensual' framing rests on this inference from evidence of absence.
  • domain assumption Manual labeling of model names and descriptions accurately identifies deepfake models and perceived gender of subjects.
    Section 3.2 and Appendix C describe a single manual labeling process with no inter-rater reliability, based on names, character references, and Google searches.
  • domain assumption Repository APIs provide complete and accurate metadata for all relevant model variants.
    Section 6 acknowledges the study is restricted to the limitations of data available via APIs; incomplete or inaccurate API data would change the counts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deepfakes on Demand: the rise of accessible non-consensual deepfake image generators." pith.science (2026). https://pith.science/paper/6G4QB6DJ

@misc{pith2026250503859,
  author       = {Pith},
  title        = {Pith review of: Deepfakes on Demand: the rise of accessible non-consensual deepfake image generators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6G4QB6DJ}},
  note         = {Machine review of arXiv:2505.03859}
}
read the original abstract

Advances in multimodal machine learning have made text-to-image (T2I) models increasingly accessible and popular. However, T2I models introduce risks such as the generation of non-consensual depictions of identifiable individuals, otherwise known as deepfakes. This paper presents an empirical study exploring the accessibility of deepfake model variants online. Through a metadata analysis of thousands of publicly downloadable model variants on two popular repositories, Hugging Face and Civitai, we demonstrate a huge rise in easily accessible deepfake models. Almost 35,000 examples of publicly downloadable deepfake model variants are identified, primarily hosted on Civitai. These deepfake models have been downloaded almost 15 million times since November 2022, with the models targeting a range of individuals from global celebrities to Instagram users with under 10,000 followers. Both Stable Diffusion and Flux models are used for the creation of deepfake models, with 96% of these targeting women and many signalling intent to generate non-consensual intimate imagery (NCII). Deepfake model variants are often created via the parameter-efficient fine-tuning technique known as low rank adaptation (LoRA), requiring as few as 20 images, 24GB VRAM, and 15 minutes of time, making this process widely accessible via consumer-grade computers. Despite these models violating the Terms of Service of hosting platforms, and regulation seeking to prevent dissemination, these results emphasise the pressing need for greater action to be taken against the creation of deepfakes and NCII.

Figures

Figures reproduced from arXiv: 2505.03859 by the authors.

Figure 1
Figure 1. Count of models tagged ‘ Celebrity’ per month on Civitai (up to December 2024). [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Proportion of Flux LoRA models labelled as ‘Deepfake’ between August 2024 and December 2024 on Civitai platform. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 31 canonical work pages

  1. [1]

    4chan. 2025. 4chan. https://www.4chan.org/

    work page 2025

  2. [2]

    ActiveFence. 2023. AI Surge in Non-Consensual Intimate Imagery (NCII): New Insights. https://www.activefence.com/research/ai-surge-ncii/

    work page 2023

  3. [3]

    Ben Bariach, Bernie Hogan, and Keegan McBride. 2024. Towards a Harms Tax- onomy of AI Likeness Generation. arXiv:2407.12030 [cs.CY] https://arxiv.org/ abs/2407.12030

    work page Pith review arXiv 2024

  4. [4]

    Abeba Birhane, Vinay Prabhu, Sang Han, Vishnu Naresh Boddeti, and Alexan- dra Sasha Luccioni. 2023. Into the LAIONs Den: Investigating Hate in Multimodal Datasets. arXiv:2311.03449 [cs.CY] https://arxiv.org/abs/2311.03449

    arXiv 2023

  5. [5]

    Abeba Birhane, Vinay Uday Prabhu, and Emmanuel Kahembwe. 2021. Multimodal datasets: misogyny, pornography, and malignant stereotypes. doi:10.48550/arXiv. 2110.01963 arXiv:2110.01963 [cs]

    arXiv doi:10.48550/arxiv.2110.01963 2021

  6. [7]

    Joel Castaño, Silverio Martínez-Fernández, Xavier Franch, and Justus Bogner

  7. [8]

    Junsong Chen, Jincheng Yu, Chongjian Ge, Lewei Yao, Enze Xie, Yue Wu, Zhong- dao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. 2023. PixArt-𝛼: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis. doi:10.48550/arXiv.2310.00426 arXiv:2310.00426 [cs]

    arXiv doi:10.48550/arxiv.2310.00426 2023

  8. [9]

    ComfyUI Studio. 2024. How to Train a Flux LoRA Locally in 15 Minutes! Create an AI Influencer with ComfyUI Workflows. https://www.youtube.com/watch? v=5af2U7JPeBs

    work page 2024

Show all 68 references
  1. [10]

    Sumanth Dathathri, Abigail See, Sumedh Ghaisas, Po-Sen Huang, Rob McAdam, Johannes Welbl, Vandana Bachani, Alex Kaskasoli, Robert Stanforth, Tatiana Matejovicova, Jamie Hayes, Nidhi Vyas, Majd Al Merey, Jonah Brown-Cohen, Rudy Bunel, Borja Balle, Taylan Cemgil, Zahra Ahmed, Ki...

  2. [11]

    Mostafa Dehghani, Josip Djolonga, Basil Mustafa, Piotr Padlewski, Jonathan Heek, Justin Gilmer, Andreas Steiner, Mathilde Caron, Robert Geirhos, Ibrahim Alabdul- mohsin, Rodolphe Jenatton, Lucas Beyer, Michael Tschannen, Anurag Arnab, Xiao Wang, Carlos Riquelme, Matthias Minde...

    2023 doi arXiv

  3. [12]

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, Kyle Lacey, Alex Goodwin, Yannik Marek, and Robin Rombach. 2024. Scaling Rectified Flow ...

    2024 doi arXiv

  4. [13]

    634, 8035 (2024), 818–823

    Scalable watermarking for identifying large language model outputs. 634, 8035 (2024), 818–823. doi:10.1038/s41586-024-08025-4 Publisher: Nature Publishing Group

    2024 doi

  5. [14]

    Ángel Fernández Gambín, Anis Yazidi, Athanasios Vasilakos, Hårek Haugerud, and Youcef Djenouri. 2024. Deepfakes: current and future trends. (2024). doi:10. 1007/s10462-023-10679-x Accepted: 2024-07-01T09:37:33Z

    2024

  6. [15]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde- Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde- Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative Adversarial Networks. doi:10.48550/arXiv.1406.2661 arXiv:1406.2661 [stat]

    2014 doi arXiv

  7. [16]

    Hugging Face. 2023. Content Policy – Hugging Face . https://huggingface.co/ content-guidelines

    2023

  8. [17]

    Michelle M. Graham. 2024. Deepfakes: Federal and state regulation aims to curb a growing threat. https://www.thomsonreuters.com/en-us/posts/government/ Deepfakes on Demand: the rise of accessible non-consensual deepfake image generators FAccT ’25, June 23-26, 2025, Athens, Gre...

    2024

  9. [18]

    Ligong Han, Yinxiao Li, Han Zhang, Peyman Milanfar, Dimitris Metaxas, and Feng Yang. 2023. SVDiff: Compact Parameter Space for Diffu- sion Fine-Tuning. 7323–7334. https://openaccess.thecvf.com/content/ ICCV2023/html/Han_SVDiff_Compact_Parameter_Space_for_Diffusion_Fine- Tuning...

    2023

  10. [19]

    UK Government. 2023. Online Safety Act 2023 . https://www.legislation.gov.uk/ ukpga/2023/50 Publisher: Statute Law Database

    2023

  11. [20]

    Susan Hao, Renee Shelby, Yuchi Liu, Hansa Srinivasan, Mukul Bhutani, Burcu Karagol Ayan, Ryan Poplin, Shivani Poddar, and Sarah Laszlo. 2024. Harm Amplification in Text-to-Image Models. doi:10.48550/arXiv.2402.01787 arXiv:2402.01787 [cs]

    2024 doi arXiv

  12. [21]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising Diffusion Probabilistic Models. doi:10.48550/arXiv.2006.11239 arXiv:2006.11239 [cs]

    2020 doi arXiv

  13. [22]

    Susan Hao, Piyush Kumar, Sarah Laszlo, Shivani Poddar, Bhaktipriya Radharapu, and Renee Shelby. 2023. Safety and Fairness for Content Moderation in Generative Models. doi:10.48550/arXiv.2306.06135 arXiv:2306.06135 [cs]

    2023 doi arXiv

  14. [23]

    Lianghua Huang, Wei Wang, Zhi-Fan Wu, Yupeng Shi, Huanzhang Dou, Chen Liang, Yutong Feng, Yu Liu, and Jingren Zhou. 2024. In-Context LoRA for Diffusion Transformers. doi:10.48550/arXiv.2410.23775 arXiv:2410.23775 [cs]

    2024 doi arXiv

  15. [24]

    HuggingFace. 2025. The Model Hub. https://huggingface.co/docs/hub/en/models- the-hub

    2025

  16. [25]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. doi:10.48550/arXiv.2106.09685 arXiv:2106.09685 [cs]

    2021 doi arXiv

  17. [26]

    Imagen-Team-Google, Jason Baldridge, Jakob Bauer, Mukul Bhutani, Nicole Brich- tova, Andrew Bunner, Lluis Castrejon, Kelvin Chan, Yichang Chen, Sander Diele- man, Yuqing Du, Zach Eaton-Rosen, Hongliang Fei, Nando de Freitas, Yilin Gao, Evgeny Gladchenko, Sergio Gómez Colmenare...

    2024 doi

  18. [27]

    Yuming Jiang, Shuai Yang, Haonan Qiu, Wayne Wu, Chen Change Loy, and Ziwei Liu. 2022. Text2Human: Text-Driven Controllable Human Image Generation. doi:10.48550/arXiv.2205.15996 arXiv:2205.15996 [cs]

    2022 doi arXiv

  19. [28]

    JaeYoung Hwang and SangHoon Oh. 2023. A Brief Survey of Watermarks in Generative AI. In 2023 14th International Conference on Information and Com- munication Technology Convergence (ICTC) (2023-10). 1157–1160. doi:10.1109/ ICTC58733.2023.10392465 ISSN: 2162-1241

    2023

  20. [29]

    Hyung-Kwon Ko, Gwanmo Park, Hyeon Jeon, Jaemin Jo, Juho Kim, and Jinwook Seo. 2023. Large-scale Text-to-Image Generation Models for Visual Artists’ Cre- ative Works. In Proceedings of the 28th International Conference on Intelligent User Interfaces (2023-03-27). 919–933. doi:1...

    2023

  21. [30]

    Black Forest Labs. 2024. Announcing Black Forest Labs. https://blackforestlabs. ai/announcing-black-forest-labs/

    2024

  22. [31]

    Beatriz Kira. 2024. When non-consensual intimate deepfakes go viral: The insufficiency of the UK Online Safety Act. 54 (2024), 106024. doi:10.1016/j.clsr. 2024.106024

    2024

  23. [32]

    Internet Matters. 2024. Report estimates half a million UK teenagers have encoun- tered AI-generated nude deepfakes . https://www.internetmatters.org/hub/press- release/internet-matters-calls-for-government-to-take-urgent-action-as- new-report-estimates-half-a-million-uk-teena...

    2024

  24. [33]

    Clare McGlynn. 2024. Deepfake porn: why we need to make it a crime to create it, not just share it . https://www.durham.ac.uk/research/current/thought- leadership/2024/04/deepfake-porn-why-we-need-to-make-it-a-crime-to- create-it-not-just-share-it/

    2024

  25. [34]

    Gon- zalez, Zhifeng Chen, Ruslan Salakhutdinov, and Ion Stoica

    Michael Luo, Justin Wong, Brandon Trabucco, Yanping Huang, Joseph E. Gon- zalez, Zhifeng Chen, Ruslan Salakhutdinov, and Ion Stoica. 2024. Stylus: Auto- matic Adapter Selection for Diffusion Models. doi:10.48550/arXiv.2404.18928 arXiv:2404.18928 [cs]

    2024 doi arXiv

  26. [35]

    Ranjita Naik and Besmira Nushi. 2023. Social Biases through the Text-to-Image Generation Lens. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society (New York, NY, USA, 2023-08-29) (AIES ’23). Association for Comput- ing Machinery, 786–808. doi:10.1145/3600...

    2023

  27. [36]

    United Nations. 2024. Draft United Nations convention against cyber- crime. //www.unodc.org/unodc/en/frontpage/2024/August/united-nations_- member-states-finalize-a-new-cybercrime-convention.html

    2024

  28. [37]

    Yisroel Mirsky and Wenke Lee. 2022. The Creation and Detection of Deepfakes: A Survey. 54, 1 (2022), 1–41. doi:10.1145/3425780 arXiv:2004.11138 [cs]

    2022 arXiv

  29. [38]

    Niki Parmar, Ashish Vaswani, Jakob Uszkoreit, Łukasz Kaiser, Noam Shazeer, Alexander Ku, and Dustin Tran. 2018. Image Transformer. doi:10.48550/arXiv. 1802.05751 arXiv:1802.05751 [cs]

    2018 doi arXiv

  30. [39]

    Bender, Emily Denton, and Alex Hanna

    Amandalynne Paullada, Inioluwa Deborah Raji, Emily M. Bender, Emily Denton, and Alex Hanna. 2021. Data and its (dis)contents: A survey of dataset development and use in machine learning research. 2, 11 (2021), 100336. doi:10.1016/j.patter. 2021.100336

    2021

  31. [40]

    UK Ministry of Justice. 2025. Government crackdown on explicit deep- fakes. https://www.gov.uk/government/news/government-crackdown-on- explicit-deepfakes

    2025

  32. [41]

    P.W. 2024. Creating a Flux Dev LORA - Full Guide . https://reticulated.net/dailyai/ creating-a-flux-dev-lora-full-guide/ Section: DailyAI

    2024

  33. [42]

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen

  34. [43]

    William Peebles and Saining Xie. 2023. Scalable Diffusion Models with Trans- formers. doi:10.48550/arXiv.2212.09748 arXiv:2212.09748 [cs]

    2023 doi arXiv

  35. [44]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-Resolution Image Synthesis with Latent Diffusion Models. doi:10.48550/arXiv.2112.10752 arXiv:2112.10752 [cs]

    2022 doi arXiv

  36. [45]

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. 2023. DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation. doi:10.48550/arXiv.2208.12242 arXiv:2208.12242 [cs]

    2023 doi arXiv

  37. [46]

    Sara Mahdavi, Rapha Gontijo Lopes, Tim Salimans, Jonathan Ho, David J

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Den- ton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim Salimans, Jonathan Ho, David J. Fleet, and Moham- mad Norouzi. 2022. Photorealistic Text-to-Image Di...

    2022 doi arXiv

  38. [47]

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Rad- ford, Mark Chen, and Ilya Sutskever. 2021. Zero-Shot Text-to-Image Generation. doi:10.48550/arXiv.2102.12092 arXiv:2102.12092 [cs]

    2021 doi arXiv

  39. [48]

    Weiss, Niru Maheswaranathan, and Surya Ganguli

    Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli

  40. [49]

    Irene Solaiman. 2023. The Gradient of Generative AI Release: Methods and Con- siderations. In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (New York, NY, USA, 2023-06-12) (FAccT ’23). Association for Computing Machinery, 111–122. doi:10....

    2023

  41. [50]

    Midjourney Team. 2025. Midjourney. https://www.midjourney.com/website

    2025

  42. [51]

    Sachith Seneviratne, Damith Senanayake, Sanka Rasnayaka, Rajith Vi- danaarachchi, and Jason Thompson. 2022. DALLE-URBAN: Capturing the ur- ban design expertise of large text to image transformers. In 2022 International Conference on Digital Image Computing: Techniques and Appl...

    2022

  43. [52]

    2023.Found through Google, bought with Visa and Mastercard: Inside the deepfake porn economy

    Kat Tenbarge. 2023.Found through Google, bought with Visa and Mastercard: Inside the deepfake porn economy . https://www.nbcnews.com/tech/internet/deepfake- porn-ai-mr-deep-fake-economy-google-visa-mastercard-download- rcna75071

    2023

  44. [53]

    Thorn. 2024. Thorn’s Safety by Design for Generative AI: 3-Month Progress Report on Civitai and Metaphysic . https://www.thorn.org/blog/safety-by-design-for- generative-ai-3-month-progress-report/

    2024

  45. [54]

    European Union. 2022. The EU’s Digital Services Act . https: //commission.europa.eu/strategy-and-policy/priorities-2019-2024/europe-fit- digital-age/digital-services-act_en

    2022

  46. [55]

    European Union. 2024. Regulation - EU - 2024/1689 - EN - EUR-Lex . https://eur- lex.europa.eu/eli/reg/2024/1689/oj/eng Doc ID: 32024R1689 Doc Sector: 3 Doc Title: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rul...

    2024

  47. [56]

    OpenAI Team. 2023. DALL·E 3 system card . https://openai.com/index/dall-e-3- system-card/

    2023

  48. [57]

    Henriikka Vartiainen and Matti Tedre. 2023. Using artificial intelligence in craft education: crafting with text-to-image generative models. 34, 1 (2023), 1–21. doi:10.1080/14626268.2023.2174557 Publisher: CAA Website

    2023

  49. [58]

    Why Take the Photo if You Didn’t Want It Online?

    Meghan Velez. 2019. “Why Take the Photo if You Didn’t Want It Online?”: Agency, Transformation, and Nonconsensual Pornography. 42, 4 (2019), 452–470. doi:10.1080/07491409.2019.1676350 Publisher: Routledge _eprint: https://doi.org/10.1080/07491409.2019.1676350

    2019 arXiv

  50. [59]

    Marco Viola and Cristina Voto. 2023. Designed to abuse? Deepfakes and the non- consensual diffusion of intimate images. 201, 1 (2023), 30. doi:10.1007/s11229- 022-04012-2

    2023 doi

  51. [60]

    Jionghao Wang, Yuan Liu, Zhiyang Dou, Zhengming Yu, Yongqing Liang, Cheng Lin, Rong Xie, Li Song, Xin Li, and Wenping Wang. 2025. Disentangled Clothed Avatar Generation from Text Descriptions. In Computer Vision – ECCV 2024 (Cham, 2025), Aleš Leonardis, Elisa Ricci, Stefan Rot...

    2025

  52. [61]

    Redacted User. 2024. Detailed Flux Training Guide: Dataset Preparation | Civitai. https://civitai.com/articles/7777/detailed-flux-training-guide-dataset- preparation

    2024

  53. [62]

    Enze Xie, Lewei Yao, Han Shi, Zhili Liu, Daquan Zhou, Zhaoqiang Liu, Jiawei Li, and Zhenguo Li. 2023. DiffFit: Unlocking Transferability of Large Diffusion Models via Simple Parameter-Efficient Fine-Tuning. doi:10.48550/arXiv.2304. 06648 arXiv:2304.06648 [cs]

    2023 doi arXiv

  54. [63]

    2025.Fine-Tuning Flux.1-dev LoRA: A Practical Guide

    Amit Yadav. 2025.Fine-Tuning Flux.1-dev LoRA: A Practical Guide. https://medium. com/@amit25173/fine-tuning-flux-1-dev-lora-a-practical-guide-b6f33af345e0

    2025

  55. [64]

    Celebrity

    Kang Zhao, Xinyu Zhao, Zhipeng Jin, Yi Yang, Wen Tao, Cong Han, Shuanglong Li, and Lin Liu. 2024. Enhancing Baidu Multimodal Advertisement with Chinese Text-to-Image Generation via Bilingual Alignment and Caption Synthesis. In Proceedings of the 47th International ACM SIGIR Co...

    2024

  56. [66]

    Matteo Wong. 2024. High School Is Becoming a Cesspool of Sexually Explicit Deep- fakes. https://www.theatlantic.com/technology/archive/2024/09/ai-generated- csam-crisis/680034/ Section: Technology

    2024

  57. [2015]

    doi:10.48550/arXiv.1503.03585 arXiv:1503.03585 [cs] FAccT ’25, June 23-26, 2025, Athens, Greece Will Hawkins, Chris Russell, and Brent Mittelstadt

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics. doi:10.48550/arXiv.1503.03585 arXiv:1503.03585 [cs] FAccT ’25, June 23-26, 2025, Athens, Greece Will Hawkins, Chris Russell, and Brent Mittelstadt

    2025 doi arXiv

  58. [2022]

    Hierarchical Text-Conditional Image Generation with CLIP Latents. doi:10. 48550/arXiv.2204.06125 arXiv:2204.06125 [cs]

    doi arXiv

  59. [2023]

    In 2023 ACM/IEEE International Symposium on Empirical Software Engineering and Measurement (ESEM) (2023-10-26)

    Exploring the Carbon Footprint of Hugging Face’s ML Models: A Repository Mining Study. In 2023 ACM/IEEE International Symposium on Empirical Software Engineering and Measurement (ESEM) (2023-10-26). 1–12. doi:10.1109/ESEM56168. 2023.10304801 arXiv:2305.11164 [cs]

    2023

  60. [2024]

    In 2024 IEEE/ACM 21st International Conference on Mining Software Reposi- tories (MSR) (2024-04)

    Analyzing the Evolution and Maintenance of ML Models on Hugging Face. In 2024 IEEE/ACM 21st International Conference on Mining Software Reposi- tories (MSR) (2024-04). 607–618. https://ieeexplore.ieee.org/abstract/document/ 10555709 ISSN: 2574-3864

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.