According to a survey conducted in March 2025, ** percent of adult female respondents in the United States expressed concerns about the spread of artificial intelligence (AI) video and audio deepfakes. Similarly, nearly ** percent of men shared this concern. In contrast, only *** percent of adult women and *** percent of adult men in the U.S. reported that they were not concerned at all.

Artificial intelligence-generated deepfakes are videos or photos that can be used to depict someone speaking or doing something that they did not actually say or do. Deepfakes are being used more frequently in cybercrime. A global 2022 survey found that 71 percent of consumers claimed they did not know what a deepfake video was, while 29 percent claimed to be familiar with the term.

This study conducts research on deepfakes technology evolution and trends based on a bibliometric analysis of the articles published on this topic along with six research questions: What are the main research areas of the articles in deepfakes? What are the main current topics in deepfakes research and how are they related? Which are the trends in deepfakes research? How do topics in deepfakes research change over time? Who is researching deepfakes? Who is funding deepfakes research? We have found a total of 331 research articles about deepfakes in an analysis carried out on the Web of Science and Scopus databases. This data serves to provide a complete overview of deepfakes. Main insights include: different areas in which deepfakes research is being performed; which areas are the emerging ones, those that are considered basic, and those that currently have the most potential for development; most studied topics on deepfakes research, including the different artificial intelligence methods applied; emerging and niche topics; relationships among the most prominent researchers; the countries where deepfakes research is performed; main funding institutions. This paper identifies the current trends and opportunities in deepfakes research for practitioners and researchers who want to get into this topic.

Deepfake AI Market Analysis The global deepfake AI market is poised for significant growth, with a market size valued at USD XXX million in 2025 and projected to reach USD XXX million by 2033, exhibiting a CAGR of XX% during the forecast period 2025-2033. Key drivers of this expansion include rising concerns over privacy and misinformation, the proliferation of social media, and the increasing availability of user data used for deepfake creation. Market Segments, Trends, and Restrains The deepfake AI market is segmented based on type (software and service), application (finance and insurance, telecommunications, government and defense, health care, and others), and region. Software solutions dominate the market currently, driven by the growing demand for advanced deepfake detection and protection technologies. Key trends in the market include the emergence of deepfake-as-a-service (DaaS) models, the integration of AI and machine learning for enhanced deepfake detection, and increased regulatory scrutiny aimed at mitigating potential risks associated with deepfake technology. However, concerns about ethical implications, legal liability, and technical challenges in detecting highly sophisticated deepfakes pose potential restraints to market growth.

A survey held on AI and journalism in August 2023 in the United States found that adults were mostly united on their attitude to using AI to create deepfakes in journalism, with the majority saying that this would not be morally acceptable. Democrats were less likely to feel this way than other voters, with ** percent saying they did not think this would be acceptable behavior, compared to close to ** percent of Republicans.

The main purpose of this data set is to facilitate research into audio DeepFakes. We hope that this work helps in finding new detection methods to prevent such attempts. These generated media files have been increasingly used to commit impersonation attempts or online harassment.

The data set consists of 104,885 generated audio clips (16-bit PCM wav). We examine multiple networks trained on two reference data sets. First, the LJSpeech data set consisting of 13,100 short audio clips (on average 6 seconds each; roughly 24 hours total) read by a female speaker. It features passages from 7 non-fiction books and the audio was recorded on a MacBook Pro microphone. Second, we include samples based on the JSUT data set, specifically, basic5000 corpus. This corpus consists of 5,000 sentences covering all basic kanji of the Japanese language (4.8 seconds on average; roughly 6.7 hours total). The recordings were performed by a female native Japanese speaker in an anechoic room. Finally, we include samples from a full text-to-speech pipeline (16,283 phrases; 3.8s on average; roughly 17.5 hours total). Thus, our data set consists of approximately 175 hours of generated audio files in total. Note that we do not redistribute the reference data.

We included a range of architectures in our data set:

MelGAN

Parallel WaveGAN

Multi-Band MelGAN

Full-Band MelGAN

WaveGlow

Additionally, we examined a bigger version of MelGAN and include samples from a full TTS-pipeline consisting of a conformer and parallel WaveGAN model.

Collection Process

For WaveGlow, we utilize the official implementation (commit 8afb643) in conjunction with the official pre-trained network on PyTorch Hub. We use a popular implementation available on GitHub (commit 12c677e) for the remaining networks. The repository also offers pre-trained models. We used the pre-trained networks to generate samples that are similar to their respective training distributions, LJ Speech and JSUT. When sampling the data set, we first extract Mel spectrograms from the original audio files, using the pre-processing scripts of the corresponding repositories. We then feed these Mel spectrograms to the respective models to obtain the data set. For sampling the full TTS results, we use the ESPnet project. To make sure the generated phrases do not overlap with the training set, we downloaded the common voices data set and extracted 16.285 phrases from it.

This data set is licensed with a CC-BY-SA 4.0 license.

This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy -- EXC-2092 CaSa -- 390781972.

'In-the-Wild' Dataset We present a dataset of audio deepfakes (and corresponding benign audio) for a set of politicians and other public figures, collected from publicly available sources such as social networks and video streaming platforms. For n = 58 celebrities and politicians, we collect both bona-fide and spoofed audio. In total, we collect 20.8 hours of bona-fide and 17.2 hours of spoofed audio. On average, there are 23 minutes of bona-fide and 18 minutes of spoofed audio per speaker.

The dataset is intended to be used for evaluating deepfake detection and voice anti-spoofing machine-learning models. It is especially useful to judge a model's capability to generalize to realistic, in-the-wild audio samples. Find more information in our paper, and download the dataset here.

The most interesting deepfake detection models we used in our experiments are open-source on GitHub:

RawNet 2 RawGAT-ST PC-Darts This dataset and the associated documentation are licensed under the Apache License, Version 2.0.

The advent of deepfakes – the manipulation of audio records, images and videos based on deep learning techniques – has important implications for science and society. Current studies focus primarily on the detection and dangers of deepfakes. In contrast, less attention is paid to the potential of this technology for substantive research – particularly as an approach for controlled experimental manipulations in the social sciences. In this paper, we aim to fill this research gap and argue that deepfakes can be a valuable tool for conducting social science experiments. To demonstrate some of the potentials and pitfalls of deepfakes, we conducted a pilot study on the effects of physical attractiveness on student evaluations of teachers. To this end, we created a deepfake video varying the physical attractiveness of the instructor as compared to the original video and asked students to rate the presentation and instructor. First, our results show that social scientists without special knowledge in computational science can successfully create a credible deepfake within reasonable time. Student ratings of the quality of the two videos were comparable and students did not detect the deepfake. Second, we use deepfakes to examine a substantive research question: whether there are differences in the ratings of a physically more and a physically less attractive instructor. Our suggestive evidence points towards a beauty penalty. Thus, our study supports the idea that deepfakes can be used to introduce systematic variations into experiments while offering a high degree of experimental control. Finally, we discuss the feasibility of deepfakes as an experimental manipulation and the ethical challenges of using deepfakes in experiments. This is the provision of the data and code.

Keywords: deepfakes, face swap, deep learning, experiment, physical attractiveness, student evaluations of teachers

According to a survey conducted in March 2025 in the United States, ** percent of respondents aged 65 and older in the United States reported being very concerned about the spread of video and audio deep fakes generated via artificial intelligence (AI), compared to ** percent of those aged 18 to 29 years. Overall, the majority of U.S. citizens across all age groups expressed worries about AI-generated deep fakes.

This study investigates the content and changes in deepfakes-related discussions on 5,220 Turkish Reddit posts from October 2019 to August 2023. Although the academic community has shown an increasing interest in deepfakes since 2017, focusing on detection methods and the technology itself, scant attention has been paid to public perceptions and online debate. The analysis reveals that 69.4% of the examined posts feature deepfake content with sexual themes, with celebrity women being the primary targets in 60.2% of cases. In contrast, 22% of the content is about politics and political figures, while 8.6% provides technical guidance on creating deepfakes. The study also observes content changes over time, noticing a rise in sexually explicit deepfake posts, particularly involving celebrities. However, in May 2023, coinciding with the presidential and general elections in Türkiye, discussions about politics and political figures have significantly increased. This study sheds light on the changing landscape of discussions, emphasizing the predominant presence of sexual content and the increasing prevalence of political content, particularly during election seasons.

Deepfake Detection Market is estimated to reach USD 5,609.3 Million By 2034, Riding on a Strong 47.6% CAGR throughout the forecast period.

Description

SWAN-DF: the first high fidelity publicly available dataset of realistic audio-visual deepfakes, where both faces and voices appear and sound like the target person. The SWAN-DF dataset is based on the public SWAN database of real videos recorded in HD on iPhone and iPad Pro (in year 2019). For 30 pairs of manually selected people from SWAN, we swapped faces and voices using several autoencoder-based face swapping models and using several blending techniques from the well-known open source repo DeepFaceLab and voice conversion (or voice cloning) methods, including zero-shot YourTTS, DiffVC, HiFiVC, and several models from FreeVC.

For each model and each blending technique, there are 960 video deepfakes. We used three types of models of the following resolutions: 160x160, 256x256, and 320x320 pixels. We took one pre-trained model corresponding for each resolution, and tuned it for each of the 30 pairs (both ways) of subjects for 50K iterations. Then, when generating deepfake videos for each pair of subjects, we used one of the tuned models and a way to blend the generated image back into the original frame, which we call blending technique. SWAN-DF dataset contains 25 different combinations of models and blending, which means the total number of deepfake videos is 960*25=24000.

We generated speech deepfakes using four voice conversion methods: YourTTS, HiFiVC, DiffVC, and FreeVC. We did not use text to speech methods for our video deepfakes, since the speech they produce is not synchronized with the lip movements in the video. For YourTTS, HiFiVC, and DiffVC methods, we used the pretrained models provided by the authors. HiFiVC was pretrained on VCTK, DiffVC on LibriTTS, and YourTTS on both VCTK and LibriTTS datasets. For FreeVC, we generated audio deepfakes for several variants: using the provided pretrained models (for 16Hz with and without pretrained speaker encoder and for 24Hz with pretrained speaker encoder) as is and by tuning 16Hz model either from scratch or starting from the pretrained version for different number of iterations on the mixture of VCTK and SWAN data. In total, SWAN-DF contains 12 different variations of audio deepfakes: one for each of YourTTS, HiFiVC, and DiffVC and 9 variants of FreeVC.

Acknowledgements

If you use this database, please cite the following publication:

Pavel Korshunov, Haolin Chen, Philip N. Garner, and Sébastien Marcel, "Vulnerability of Automatic Identity Recognition to Audio-Visual Deepfakes", IEEE International Joint Conference on Biometrics (IJCB), September 2023. https://publications.idiap.ch/publications/show/5092

A May 2024 survey of IT and cybersecurity professionals worldwide found that around ** percent of organizations in the United States were targeted by AI-assisted deepfakes in the past ** months, compared to the global average of ** percent.

• Enhancements: Derived from the Fakeddit fake news dataset, the DeepFakeNews dataset comprehends a total of 509,916 images and has been enriched with 254,958 deepfake images generated using three different generative models: Stable Diffusion 2, Dreamlike, and GLIDE.
• Balance and Composition: The dataset is perfectly balanced, containing an equal number of pristine (authentic) and generated (deepfake) images.
• Removal of Hand-Modified Content: The original "manipulated content" category from Fakeddit, which consisted of images altered or modified by hand, has been removed. These have been replaced with deepfakes to provide a more relevant and challenging set of synthetic images.
• Cleaning and Quality Control: The Fakeddit dataset was thoroughly cleaned, removing any images that were not found, contained only logos, or were otherwise unsuitable for deepfake detection. This cleaning process ensures a higher quality and more reliable dataset for training and evaluation.
• Application: The DeepFakeNews dataset is suitable for both deepfake detection and fake news detection. Its diverse and balanced nature makes it an excellent benchmark for evaluating multimodal detection systems that analyze both visual and textual content.

The dataset comes with three CSV files for training, testing, and validation sets, along with corresponding zip files containing the split images for each set. The deepfake images are named in both the CSV files and the image filenames following a specific format based on the generative model used: "SD_fake_imageid" for Stable Diffusion, "GL_fake_imageid" for GLIDE, and "DL_fake_imageid" for Dreamlike.

The Deepfake Generation Pipeline involves a 2 steps approach:

1. first generating a caption for a pristine image using a captioning model.
2. then feeding this caption into a generative model to create a new synthetic image.

By incorporating images from multiple generative technologies, the dataset is designed to prevent any bias towards a single generation method in the training process of detection models. This choice aims to enhance the generalization capabilities of models trained on this dataset, enabling them to effectively recognize and flag deepfake content produced by a variety of different methods, not just the ones they have been exposed to during training. The other half consists of pristine, unaltered images to ensure a balanced dataset, crucial for unbiased training and evaluation of detection models.

The dataset has been structured to maintain retrocompatibility with the original Fakeddit dataset. All samples have retained their original Fakeddit class labels (6_way_label), allowing for fine-grained fake news detection across the five original categories: True, Satire/Parody, False Connection, Imposter Content, and Misleading Content. This feature ensures that the DeepFakeNews dataset can be used not only for multimodal and unimodal deepfake detection but also for traditional fake news detection tasks. It offers a versatile resource for a wide range of research scenarios, enhancing its utility in the field of digital misinformation detection.

For full info and details about dataset creation, cleaning pipeline, composition and generation process please refer to my Master Thesis.

Replication data for Political Deepfakes are as Credible as Other Fake Media and (Sometimes) Real Media. Primary data analysis and code work done by my co-authors Soubhik Barari and Christopher Lucas.

The Deepfake AI market is experiencing rapid growth, driven by advancements in artificial intelligence and machine learning. While precise market sizing data is unavailable, considering the burgeoning interest in AI-powered media manipulation and its applications across various sectors, a reasonable estimate for the 2025 market size could be around $500 million. This is based on the increasing adoption of deepfake technology in entertainment, marketing, and education, as well as its potential in security and defense applications. The Compound Annual Growth Rate (CAGR) is projected to remain strong, potentially exceeding 25% through 2033, indicating significant market expansion. Key drivers include the decreasing cost and increasing accessibility of deepfake creation tools, coupled with rising demand for realistic video and audio content across various industries. However, ethical concerns surrounding misuse, such as the creation of fraudulent content and identity theft, present significant restraints to market growth. Regulatory frameworks and technological solutions aimed at detecting and mitigating deepfakes are crucial for fostering responsible innovation within the sector. The market is segmented by application (entertainment, advertising, security, etc.) and technology (generative adversarial networks (GANs), autoencoders, etc.). The competitive landscape is marked by a diverse range of companies offering both deepfake creation and detection solutions. Major players include Synthesia, Pindrop, Reface, and others, each focusing on specific niches within the market. The future of the Deepfake AI market hinges on the balance between technological advancement, ethical considerations, and robust regulatory oversight. Further growth will be significantly impacted by developments in AI detection technologies, increased public awareness, and clear industry guidelines aimed at minimizing malicious applications while encouraging beneficial uses. The geographical distribution is expected to show a strong concentration in North America and Europe initially, but significant growth potential exists in other regions as technology adoption matures globally. Furthermore, advancements in areas like AI-driven content verification and authentication are crucial for mitigating the risks associated with deepfake technology and unlocking its wider potential.

Deepfake Detection Accelerator Market Outlook

According to our latest research, the global Deepfake Detection Accelerator market size in 2024 is valued at USD 1.23 billion, reflecting a robust response to the growing threat of synthetic media and manipulated content. The market is expected to expand at a remarkable CAGR of 28.7% from 2025 to 2033, reaching a forecasted value of USD 10.18 billion by 2033. This substantial growth is driven by increasing awareness of the risks associated with deepfakes, rapid advancements in artificial intelligence, and a surge in demand for real-time content authentication across diverse sectors. As per our latest research, the proliferation of deepfake technologies and the resulting security and reputational risks are compelling organizations and governments to invest significantly in detection accelerators, thereby propelling market expansion.

One of the primary growth factors for the Deepfake Detection Accelerator market is the exponential increase in the creation and dissemination of deepfake content across digital platforms. As deepfakes become more sophisticated and accessible, businesses, media outlets, and public institutions are recognizing the urgent need for robust detection solutions. The proliferation of social media, coupled with the ease of sharing multimedia content, has heightened the risk of misinformation, identity theft, and reputational damage. This has led to a surge in investments in advanced deepfake detection technologies, particularly accelerators that can process and analyze vast volumes of data in real time. The growing public awareness about the potential societal and economic impacts of deepfakes is further fueling the adoption of these solutions.

Another significant driver is the rapid evolution of artificial intelligence and machine learning algorithms, which are the backbone of deepfake detection accelerators. The ability to leverage AI-powered hardware and software for identifying manipulated content has substantially improved detection accuracy and speed. Enterprises and governments are increasingly relying on these accelerators to safeguard sensitive information, ensure content authenticity, and maintain compliance with emerging regulations. The integration of deepfake detection accelerators into existing cybersecurity frameworks is becoming a standard practice, especially in sectors such as finance, healthcare, and government, where data integrity is paramount. This technological synergy is expected to sustain the market’s momentum throughout the forecast period.

The regulatory landscape is also playing a critical role in shaping the growth trajectory of the Deepfake Detection Accelerator market. Governments across major economies are enacting stringent policies and guidelines to combat the spread of malicious synthetic content. These regulations mandate organizations to implement advanced detection mechanisms, thereby driving the demand for high-performance accelerators. Furthermore, industry collaborations and public-private partnerships are fostering innovation in the development of scalable and interoperable deepfake detection solutions. The increasing frequency of high-profile deepfake incidents is prompting regulatory bodies to accelerate the adoption of these technologies, ensuring market growth remains on an upward trajectory.

From a regional perspective, North America currently leads the global deepfake detection accelerator market, accounting for the largest share in 2024. This dominance can be attributed to the presence of key technology providers, a mature cybersecurity ecosystem, and proactive regulatory initiatives. Europe follows closely, driven by strict data protection laws and increased investments in AI research. The Asia Pacific region is emerging as a high-growth market, fueled by the rapid digital transformation of its economies and rising concerns about deepfake-related cyber threats. Latin America and the Middle East & Africa are also witnessing increased adoption, albeit at a slower pace, as awareness and infrastructure development continue to progress. Overall, the global market is poised for sustained growth, with regional dynamics playing a pivotal role in shaping future trends.

Recent technological advancements in Artificial Intelligence make it easy to create deepfakes, hyper-realistic videos in which images and video clips are processed to create fake videos that appear authentic. Many of them are based on swapping faces without the consent of the person whose appearance and voice are used. As emotions are inherent in human communication, studying how deepfakes transfer emotional expressions from original to fakes is relevant. In this work, we conduct an in-depth study on facial emotional expression in deepfakes using a well-known face swap-based deepfake database. First, we extracted the photograms from their videos. Then, we analyzed the emotional expression in both the original and the faked versions of the video recordings for all performers in the database. Results show that emotional expressions are not adequately transferred between original recordings and the deepfakes created from them. The high variability in emotions and performers detected between original and fake recordings indicates that performer emotion expressiveness should be considered for better deepfake generation or for detecting them.

In 2023, the police in South Korea recorded *** individual cases of illegally created deepfake sexual content on the internet. The number of such reports has increased slightly over the last three years. Recently, the issue of illegally created deepfakes has gotten more attention in South Korea as a set of Telegram rooms distributing AI deepfake pornography has been discovered.

A 2023 survey found that approximately ** percent of respondents in Indonesia reported having heard of deepfakes. Meanwhile, around ** percent of respondents were unaware of the technology. Deepfakes are realistic yet deceptive images, audio, or videos generated using artificial intelligence technology.