How We Created a Deepfake of a Deepfake in 20 Minutes

We jumped into the fascinating yet alarming world of deepfakes, a technology that's quickly reshaping our digital environment. We started this experience when Deutsche Telekom released a video about this topic, and we asked ourselves, how convincing can a video created by us be? (low budget, no support, small team). So, we embarked on the journey.

Deepfakes, for those unfamiliar, are synthetic media where the likeness of one person is replaced by another, using artificial intelligence (AI). From Hollywood celebrities to politicians – to children, anyone can be a target, and the results are so convincing it's often challenging to distinguish between reality and deepfake. In the past, such high-level manipulation required substantial resources and technical expertise. Today, thanks to advancements in AI, it's become frighteningly easy to create copies of us online, saying things we would – or would not say – at lightning speed.

The Process of Crafting a Deepfake

Creating a deepfake, despite its complex final output, is relatively straightforward. The process involves feeding a machine learning algorithm, typically a variant of a Generative Adversarial Network (GAN), with numerous images or videos of the target. The more data you provide, the more accurate the deepfake tends to be. Once the machine has learned the target's features, it can generate new images or videos that are practically indistinguishable from the real thing. This process is called 'training,' and it can take anywhere from a few hours to a few days, depending on the complexity of the deepfake and the processing power available. However, the real challenge lies in making the deepfake convincing. This involves fine-tuning the AI model, adjusting the lighting, and syncing the audio. While there are pre-built solutions available for creating deepfakes, creating a convincing and untraceable one requires a certain level of technical expertise.

The Experiment

We decided to experiment and create our own deepfake of Telekom’s video, and we wanted to do it in just 20 minutes. First, we started with their existing video. In less than 20 minutes, our system learned the facial features of our target. and was able to generate a new video. The result was a fascinating, albeit slightly unnerving, demonstration of just how powerful this technology can be.

Deepfakes and Sharenting

Now, let's talk about a relatively new term - sharenting. Sharenting is the practice of parents sharing their children's photos or videos online. While adorable and seemingly harmless, sharenting can inadvertently provide a treasure trove of data for deep fake creators. Consider this: with every picture or video shared, you're providing more data that can be used to train a deepfake algorithm. As the technology becomes more accessible, the risk of these images being misused grows. What’s worse is that bad players can also use what parents say online too. Let’s say; for instance, you have a parent that is controversial on social media – all that information follows the parent and can also follow their children – forever. The threat is real, and the implications, especially for children who don't have the agency to consent to their images being shared, are concerning. Parents need to be aware of the potential risks of sharenting and take steps to protect their children's privacy. This could include limiting the number of photos shared, only sharing on channels with known followers, and using privacy settings on social media – all while being very careful about which moments they choose to share online.

Societal Impact of Deepfakes

The societal impact of deepfakes is profound and multi-faceted. On one hand, they offer exciting possibilities for entertainment and creative expression. Imagine being able to insert yourself into your favorite movie, or having a personalized greeting from your favorite celebrity. These are the fun, harmless applications of deepfakes. On the other hand, the potential for misuse is staggering. Deepfakes can be used to create fake news, manipulate public opinion, and even incite violence. They can be used to blackmail individuals, damage reputations, and commit fraud. The threat to personal privacy and societal stability is real and significant. Additionally, deepfakes can also undermine trust in digital media. As deepfakes become more common and more convincing, people may begin to doubt the authenticity of all digital content, leading to a 'reality apocalypse' where nothing can be trusted.

How to Protect Yourself from Deepfakes

Protection against deepfakes involves a combination of technological solutions, media literacy, and legal measures. On the technological front, researchers are developing AI tools that can detect deepfakes. These tools analyze the media for signs of manipulation, such as inconsistencies in lighting or unnatural movements. However, technology alone is not enough. We need to develop our media literacy skills, learn how to critically evaluate digital media, and be aware of the potential for manipulation. This includes checking the source of the media, looking for signs of manipulation, and cross-referencing information from multiple sources. Legal measures are also crucial. Laws need to be updated to address the unique challenges posed by deepfakes, including issues of consent, defamation, and fraud. Individuals who are victims of deepfakes should have legal recourse available to them.

Legal and Ethical Considerations of Deepfakes

The legal and ethical landscape surrounding deepfakes is complex and evolving. From a legal perspective, deepfakes raise issues around consent, defamation, and fraud. Producing a deepfake without the individual's consent can be seen as a violation of their privacy rights. If a deepfake causes harm to a person's reputation, it could be considered defamation. And if a deepfake is used to deceive or commit fraud, it could be deemed a criminal act. Ethically, deepfakes raise questions about truth, trust, and the nature of reality. In a world where deepfakes are prevalent, how do we determine what's real and what's not? How do we trust digital media? And how do we protect the rights of individuals whose likeness is used in a deepfake? These are difficult questions, and there are no easy answers. However, it's crucial that we continue to engage in these discussions as a society, and work towards legal and ethical frameworks that protect individuals and uphold the integrity of our digital world.


The Future of Deepfakes and Privacy Deepfakes are here to stay, and their impact on everyone will be significant. As this technology continues to evolve, so too must our approaches to protecting privacy. This includes developing more sophisticated detection tools, promoting media literacy, and updating our legal frameworks. However, it's not all doom and gloom. With awareness and action, we can mitigate the risks associated with deepfakes. We can enjoy the creative possibilities of this technology without falling prey to its potential for misuse. As a parting thought, we invite you all to watch this Wake-up call video. It's a visual journey into the world of deepfakes and a reminder of the need for vigilance and awareness. Stay safe, stay informed, and let's navigate the digital future together.