Deepfakes have emerged as one of the most significant developments in contemporary computational media, representing a sophisticated convergence of machine learning, computer vision, and audiovisual synthesis. Enabled primarily by deep neural networks such as generative adversarial networks (GANs) and transformer-based architectures, Deepfakes are realistic video fabrications through sound and image alteration and substitution that synthesises human likeness, speech, and behaviours. Deepfakes function simultaneously as creative tools, political instruments, security risks, and epistemic disruptors. They have generated widespread scholarly, regulatory, and public concern by contributing to the reshaping of visual communication and posing significant challenges to established norms of authenticity. This entry defines Deepfakes, outlines their technological foundations, synthesises insights from current research and assesses implications for media industries, journalism, documentary, disinformation, governance, and digital culture.
Deepfakes refer to algorithmically generated or manipulated audiovisual artefacts produced through deep learning systems initially involving encoder–decoder models and more recently generative adversarial networks (GANs). Deepfakes are synthetic representations of people, voices, or events with high degrees of realism making them increasingly difficult to distinguish from authentic representations. Prominent U.S. politicians Nancy Pelosi, Barack Obama and Joe Biden have been the subject of Deepfakes alongside countless numbers of female celebrities and anonymous women who have been subjected to Deepfakes for the purposes of revenge-porn. While traditional audiovisual manipulation has long existed, deepfakes represent a paradigmatic shift due to their automation, scalability, believability, and accessibility, allowing non-experts to create and distribute photorealistic fabrications with consumer-level tools.
Within media and communication studies, Deepfakes are understood as a form of synthetic media capable of eroding visual evidence standards, destabilising epistemic trust, and reshaping public communication practices
[1][2][3][4][5][1,2,3,4,5]. Deepfakes is a compound noun derived from Deep and Fake: deep refers to deep learning methods, and fake, referring to inauthentic. Deepfakes refers to the products generated by machine learning computational models that enable face replacement, voice cloning, mouth reanimation, full-body puppetry, and multimodal synthetic content generation.
The following entry is arranged accordingly: Technological foundations presented in
Section 2. Expanding into media, communication, and the epistemic challenges of deepfakes in
Section 3. The social, political and cultural implications of Deepfakes are examined in
Section 4. Issues stemming from Detection, Forensics, and Synthetic Media Literacy and explored in
Section 5. Key measures of control are the focus of
Section 6 with governance, regulation, and ethical frameworks.
Section 7 addresses contemporary developments and future trajectories and conclusions are in
Section 8.
