Deepfake is one of the most sophisticated as well as terrifying inventions that has become possible due to great advancements in artificial intelligence (AI), machine learning (ML), and other IT sectors. The origin of a deepfake can be traced back to 2017 when a user posted pornographic videos on Reddit and replaced the faces of the actual porn stars with celebrities.
Since then, the concept of deepfake, especially celebrity deepfakes, has continued to become more and more popular as deepfaked videos of various celebrities like Kylie Jenner, Tom Cruise, Taylor Swift, and many others have been released.
In this article, you will get to know all about celebrity deepfakes and the basic working of deepfake technology that has both pros and cons.
Let's start with some basics.
What Is Deepfake?
The term 'deepfake' itself originates from the combination of 'Fake' and 'Deep Learning'. The purpose of deepfake technology is to superimpose face images of a target person onto a source person to create a fake video. Due to the great advancements in AI, most people cannot distinguish fake videos from real ones.
Now that you are familiar with the basics of deepfake, you will be able to understand its fundamental working procedures as well.
Working of Deepfake
An autoencoder is the foundation of deepfake technology. An autoencoder is a deep neural network that understands how to take an input, compress it into a manageable representation or encoding, and then regenerate the original input from this encoding.
The network is forced to construct these pictures rather than just returning what it sees when there is congestion in the middle. The encodings aid in capturing more extensive patterns.
A deepfake celebrity video typically makes use of two ML models. The forgeries are made by one model using a database of example movies, while the other model attempts to determine if the video is indeed a fake. The deep fake is likely convincing enough to a human viewer when the second model can no longer determine if the video is fake. The term "generative adversarial network" refers to this method (GAN).
When given a huge data set to work with, GAN performs better. That is why politicians and Hollywood stars are often seen in the early deepfake films. GAN can leverage its extensive video library to produce very lifelike deepfakes.
What Are the Applications of Deepfake in the Real World?
If used correctly, deepfake can be a highly useful tool. Deep learning has myriad applications in complex real-world problems such as big data analytics, computer-vision perception, and state-of-the-art control systems.
However, the cons of deep learning and especially deepfake images and videos cannot be ignored as they pose a serious threat to individuals' privacy as well as the stability of the ML-based systems.
Celebrity Deepfakes in the Entertainment Industry
While on an individual level, celebrity deepfakes have the potential to damage a celebrity's reputation to a great extent as the majority of people are likely to believe such videos and images and not verify them.
Yet, swapping faces and celebrity deepfakes are beneficial in the entertainment industry. Numerous movies and TV series have started using deepfake technology for shooting purposes and to de-age the celebrities as per the narrative.
For instance, Disney's The Mandalorian, a highly popular Star Wars series, used deepfake technology to de-age the actor Mark Hamill to allow him to play the younger version of the character Luke Skywalker.
Other than the big-budget movies and series, many fans and avid film watchers also continue to use deepfake technology to create different types of deepfake celebrity videos.
Let's explore some of the other specific examples of celebrity deepfakes.
Celebrity Deepfakes on Social Media Platforms
1. TikTok Tom Cruise
In 2021, videos of "Tom Cruise'' went viral on the popular social network TikTok. He was seen doing different types of coin tricks and goofing around, which is unlikely for the Hollywood star. After a few videos, users realized that these were deepfake videos, and the doppelganger was using advanced AI and deepfake technology to mimic Tom Cruise's face as well as voice.
2. Obama's Public Service Message
Comedian Jordan Peele and Buzzfeed invested over 70 hours in creating a deepfake video of President Obama. Computer scientists at the University of Washington used neural networks AI to model the shape of President Obama's mouth and used his previous public addresses to create the voice.
3. Fake Roundtable
As Netflix and other streaming services compete for your attention, the entertainment sector is now experiencing a great transformation. In light of this, Collider created a funny celebrity deepfake featuring the voices of Tom Cruise, Robert Downey Jr., George Lucas, Ewan McGregor, and Jeff Goldblum debating streaming and the future of film.
4. Donald Trump in Breaking Bad
While some deepfakes aim to deceive the audience, Better Call Trump: Money Laundering 101 is a pure satire. In this video, Donald Trump is introduced as corrupt attorney Saul Goodman using a clip from the wildly successful Breaking Bad television series.
In the sequence, Aaron Paul's character Jesse Pinkman is taught the fundamentals of money laundering by Goodman. Jared Kushner, Donald Trump's deepfaked son-in-law, replaces Paul in the deepfaked scenario to give it a more realistic feel, making the spoof seem almost intimate.
Celebrity Deepfakes Using Deepswap
Deepswap is a popular free deepfake online tool that allows you to swap faces within seconds without being familiar with the technicalities of deepfake technology. A lot of people use it to create celebrity deepfake.
Let's explore some celebrity deepfakes made by Deepswap.ai.
This video of Taylor Swift playing tennis is one of the many examples of such celebrity deepfakes. Here, we can see Taylor Swift as a tennis coach rather than a singer.
Here, we can also see a deepfaked video in which Scarlett Johansson is in an alluring swimsuit.
Generally, it is easier to create celebrity deepfakes and target high-profile individuals because a lot of material, in the form of video and audio, is available about the public figures. As a result, it becomes easier to train the AI models.
However, it is important to note that with the rising trend of taking and uploading selfies and further advancements in AI and ML, it is becoming possible to create a deepfake of an individual even with a limited amount of data.
Therefore, legislation should be in place to prevent the malicious use of deepfake technology rather than banning it altogether, considering its great value in the entertainment industry.