In recent years, artificial intelligence has made significant strides in generating realistic media through machine learning algorithms. This technology has opened up new avenues for creativity and innovation. At the forefront of jaw-dropping AI advancements is the emergence of deepfakes.
Deepfakes are AI-generated media that are created by manipulating existing data to generate new content that appears to be authentic but is entirely fabricated. The term “deepfake” is coined from the combination of “deep learning” and “fake”.
In today’s digital age, seeing is no longer believing. But while deepfakes have received much scrutiny and concern, it is important to note that there are flip sides to the use of this technology.
Whether you’re a tech enthusiast or a creative wizard, deepfakes offer an exciting new frontier for expression. In our comprehensive guide, we will explore the inner workings of deepfakes, their types, and the legalities involved in creating one. So buckle up and get ready to dive headfirst into the world of AI-generated media!
How Does Deepfake Work?
Ever scrolled past a video of Kobe Bryant in a Kendrick Lamar music video? No, you’re not experiencing a fever dream, you’ve just stumbled upon a deepfake! So how does it work exactly?
Deepfake technology uses dual machine learning models and a framework called generative adversarial networks or simply GANs. GANs further consist of two neural networks called a generator and a discriminator.
The generator creates deepfake media, while the discriminator attempts to identify whether it is real or fake. Through a process of trial and error, the generator continually improves its output until it can create deepfakes that are virtually indistinguishable from authentic media.
To understand how deepfake works, think of GANs as a pair of detectives, one creates fake media, and the other tries to tell if it is fake. The first detective, known as the generator, creates a fake image or video by studying a real image or video and identifying key features, such as the shape of a person’s face, lips, eyes, etc.
It then uses this information to create a new image or video that is similar but not identical to the original. The second detective called the discriminator, is then tasked with making the generated media as authentic-looking as possible.
It looks for inconsistencies, such as facial features that don’t match their bodies or movements that look unnatural. This process continues until the final results look convincing.
If you want to learn how deepfake media is made, click here.
So How Many Images Are Required To Create DeepFakes?
One of the critical components of a deepfake is the images used to train the algorithm. The question then arises, how many pictures do I actually need to create a deepfake?
The answer depends on various factors, including the complexity of the task at hand, the quality of the source images, and the sophistication of the deepfake algorithm. Generally between 500 – 100k images can create a convincing deepfake.
The more images you have, the better the deepfake will be. Think of it as adding more ingredients to a recipe to create a richer flavor. However, there is no particular set number of images required to create a deepfake. For instance, to create a convincing deepfake of a celebrity, a vast amount of high-quality images of the target individual would be required.
This is where the power of machine learning comes in, as the deepfake algorithm can use these images to create a realistic digital version of the celebrity.
Different Types of Deepfake Media
As deepfake technology keeps advancing at an exponential rate, it is important to be aware of the different types of deepfakes and their potential implications. There are currently 5 major types of deepfakes available, they are as follows:
Deep Fake Videos
Deepfake video is what usually comes to mind when the term “deepfake is used”. With photos or videos, deepfake programs can use artificial intelligence to manipulate a person’s appearance or actions in a video.
This can range from adding a smile to a person’s face to changing their facial features. Want to see what it would look like if your favorite celebrity starred in a different movie? A deepfake video can make that happen.
Textual deepfakes are created by using natural language processing (NLP) algorithms to generate convincing text that appears to be written by a specific person.
Thanks to the advancement in language models, creative works like stories, news articles, and poems can now be generated within seconds with a given prompt. One recent example of a textual deepfake is content produced by the popular ChatGPT, an AI language model that is trained to generate human-like responses to user inputs called prompts.
ChatGPT is capable of producing coherent responses that can trick users into believing that they are conversing with a real person.
With data from a few minutes of audio or video recordings of a person’s voice, deepfake algorithms can generate speech patterns, tones, and pitch that sound just like the person speaking. This can be used to create deep fake voice messages or even impersonate someone. Check out these advanced AI Voice generators.
Want to see what you would look like as Batman or wonder woman? Deepfake images can make that happen. Deepfake images can be created by using machine learning algorithms to manipulate photos or images.
This can range from changing a person’s facial expressions to creating entirely new people from scratch. For example, deepfake can be used in the fashion industry to create virtual models that don’t exist in real life.
To learn more about AI-generated images, read our 10-minute guide on creating art with Midjourney or Dall-E 2.
Real-Time or Live DeepFake
Real-time or live deepfake technology uses computer vision to manipulate live video or audio in real-time. This type of deepfake is often used in video conferences or live interviews, where the speaker’s appearance or voice can be altered in real-time.
5 Best Examples of Deepfake Media
While we’ve discussed the different types of deepfakes and their implications, seeing them in action is even eerier. From harmless comedic parodies to dangerous impersonations, the use of deepfakes has raised significant ethical concerns. Here are some of the most popular examples of deepfake media pieces today.
1. Donald Trump in the Blockbuster Show “Breaking Bad”
This deepfake video features the former US President Donald Trump inserted into scenes from the hit TV show “Breaking Bad.” Created by YouTube creators Ctrl Shift Face.
The video features Trump’s face seamlessly superimposed onto the character of shady lawyer Saul Goodman conversing with dope pusher Jesse Pinkman about money laundering.
The video was created using DeepFace Lab and AI model called Stable Voices, which allows for more realistic and consistent audio and lip-syncing. While the video was created for comedic purposes, it highlights the technology’s potential to manipulate reality in a chilling way. Of course, the video went viral on social media platforms, attracting views and criticism.
2. Pope Francis in an Ankle-Length Puffy Jacket
If you’ve been on social media for the past few weeks, you must have seen the image of Pope Francis rocking a white puffer jacket, news flash, AI strikes again!
The deepfake image was created by Reddit user Pedro Xavier using the generative AI program, Midjourney. The image quickly went viral on social media platforms as most netizens thought the outfit looked dabber and was real.
3. Obama Giving a Public Announcement
A video of former US President Barack Obama, in which he appears to be delivering a public speech was produced by BuzzFeed and comedian Jordan Peele. The deepfake was generated using Adobe After Effects CC and FakeApp, a deep learning tool for creating realistic face-swapped videos.
In the video, Peele’s voice is used to provide a convincing Obama impersonation, while the AI technology is used to overlap his facial features with Obama’s. On a positive note, the video was seemingly made without malicious intent but was created by the media giant to raise awareness about the dangers of fake news.
4. Unreal Keanu Reeves Videos on Tiktok
A-list celebrities are mostly the target of deepfake media as hundreds to thousands of pictures are usually needed for hyper-realistic rendering of their faces and movements. Deepfakes of Kim Kardashian, Tom Cruise, Nicki Minaj, Adele, and many others can be found on social media platforms.
However, one of the most popular deepfakes is that of Keanu Reeves, created by the TikTok account @Unreal_Keanu. The account features deepfake videos of Reeves participating in challenges and engaging in regular day-to-day activities.
The videos are incredibly convincing, and many viewers have been left amazed by the level of detail and accuracy of the deepfake technology. The popularity of @Unreal_Keanu again highlights the growing fascination with deepfakes and the potential for technology to create disturbingly realistic content.
5. DeepFake Audio Track of Rihanna Featuring Drake
On April 25th, a supposed track titled “Por Que” featuring popular artists Bad Bunny and Rihanna emerged on the internet. The song was uploaded directly to Tiktok, YouTube, and Sound Cloud by an anonymous creator.
The song not officially endorsed by either Bad Bunny or Rihanna was taken down eventually and further sparked important questions about copyright, ownership, and artistic integrity.
Days before this saga, a similar AI-produced song allegedly by the same creator made waves on Tik Tok. The track featured Drake and The Weekend and was titled “Heart on my sleeve”. It garnered millions of views before most people realized it was fake. As expected, Apple and other streaming platforms pulled the music citing copyright claims.
Legalities of DeepFake Technology
As the popularity of deepfakes continues to rise, so too does the concern around their legality. While the technology in and of itself is not illegal, some of its applications can lead to various legal consequences. Here are some instances where deepfakes can be illegal.
Deepfakes used to create explicit videos or images of individuals without their consent for the purpose of humiliation or extortion are illegal in many countries.
46 states in the United States have placed a ban on revenge porn. But the conversation around the laws surrounding deep-faked illicit content is still ongoing. So far, only the state of Virginia and California have laws directly addressing deepfake media.
As of November 2022, the UK amended its online safety bill, which recognizes AI-powered revenge porn as a criminal offense. The bill is expected to come into full effect in the coming months.
Fraud and Identity Theft
Using deepfakes to impersonate someone for financial gain, whether through identity theft, fake credentials, or forged documentation, is illegal. In many countries, including the United States, the use of deepfakes to commit fraud or identity theft can result in criminal charges under federal and state laws.
Additionally, the technology is used to create fake credentials, such as diplomas or certifications is also illegal and can result in criminal charges for forgery. As the days go by, proving the use of deepfakes in committing fraud is becoming increasingly challenging, as technology is constantly evolving and becoming more sophisticated.
However, law enforcement agencies and other organizations are still working to develop new methods of detecting and preventing deepfake-related crimes.
Defamation and Copyright Infringement
Deepfakes used to spread false information that harms someone’s reputation or livelihood can lead to legal action. In this case, the person or entity being defamed can take legal action against the creator of the deepfake.
As seen recently In the entertainment industry, deepfakes have been used to create audio recordings of artists that they never produced.
The use of deepfakes in political manipulation of any kind is a no-no and can have serious legal consequences. For example, in the United States, the Federal Communications Commission has regulations on political advertisements and their content, and deepfakes that violate these regulations can also be deemed illegal.
In addition to these laws, there may be civil and criminal consequences for those who create and distribute political deep fakes. For example, individuals who create deepfakes that intentionally spread false information about political candidates or parties may be liable for defamation or libel.
Additionally, those who distribute political fake media with the intent to influence an election outcome may be in violation of election tampering laws.
Is Watching Deepfake Illegal?
Watching deepfakes is generally not illegal, as long as the content being viewed does not violate any other laws or regulations. However, there are some exceptions to this rule.
For example, in some jurisdictions, it may be illegal to possess or distribute certain types of deepfakes, such as those that involve child pornography or revenge porn.
In these cases, watching deepfakes could be considered aiding and abetting in the commission of a crime. It is also worth noting that while synthetic media can be used maliciously, they also have positive applications. For example, deepfakes can be used in the entertainment industry to enhance movies, television shows, and video games.
They can also be used in research and education, such as creating realistic simulations of historical figures or extinct animals. As with any technology, it is important to use deepfakes responsibly and ethically. To learn more, read our guide on the commercial use of AI-generated images.
As we wrap up our descent into the world of deepfakes, it is hard not to feel a mix of excitement and concern. On the one hand, we can’t help but marvel at the incredible potential of deepfake technology to create new forms of media and entertainment. On the other hand, it’s not all fun and giggles as we’re acutely aware of the risks and dangers that come with this powerful tool.
So, how do we move forward? Well, regardless of how we feel about AI advancement, it is up to us to stay vigilant and use deepfakes responsibly and ethically to combat the spread of misinformation and avoid legal implications.
If you want to learn how to detect deepfake media, check out this article.
Join our newsletter as we build a community of AI and web3 pioneers.
The next 3-5 years is when new industry titans will emerge, and we want you to be one of them.
- Receive updates on the most significant trends
- Receive crucial insights that will help you stay ahead in the tech world
- The chance to be part of our OG community, which will have exclusive membership perks