In a nutshell

  • Deepfakes use AI to alter videos, often replacing faces, and are becoming more common and convincing.
  • While deepfakes can be used for entertainment, they can also cause serious harm by spreading false narratives and damaging reputations.
  • Signs of a deepfake include face discoloration, unrealistic eye blinking, poor lighting and unsynchronized audio and video.

Usually, when we think of technology progressing, we think of it as a positive thing. For instance, as the film industry continues to grow, more and more people gain access to higher-powered cameras that were only available to the top elite filmmakers just years ago. While technological advancements can be beneficial, even life-changing for our society, sometimes they can create issues that should be taken very seriously. One of the prime examples of this is the rapidly advancing technology behind deepfakes. So what are deepfake videos?

AI-generated fake videos are becoming more and more common in our current day of age. While some good can come from deepfakes, there are many things that we need to discuss and be mindful of as a society.

What are deepfakes?

Before we get into the moral and ethical implications of deepfakes, let’s first define a deepfake. Deepfakes are videos altered using AI technology, often switching out a subject’s face with another. Likely, you’ve seen memes or YouTube videos modified using a deepfake. Some of the most notorious deepfakes include putting Steve Buscemi’s face on Jennifer Lawrence or placing Nicolas Cage into every movie ever. A lot of these examples are made without malicious intent. They are meant to be entertaining, with the fact that they’re deepfakes being apparent.

More recently, deepfakes have made their way into Hollywood. For example, Disney is currently developing its own deepfake model. The company has also famously used deepfakes in recent “Star Wars” entries. Additionally, deepfakes have been used by studios to de-age actors, like how Disney de-aged Harrison Ford in “Indiana Jones and the Dial of Destiny” (2023).

However, as technology progresses and becomes more readily available to the public, deepfakes are being used in troubling ways. Some deepfakes have become nearly indistinguishable from reality, and that’s a frightening thought. Think about it. Practically anyone with the technology, or the money to hire someone with the skills, could fabricate a video of you saying or doing anything they choose. This could damage a person’s reputation, especially when the target can’t prove the video is fake. Deepfakes can even weaken democracy by promoting false narratives and framing politicians or public figures in ways that aren’t true.

Additionally, there’s been a lot of talk about how corporations, such as Disney, have (and will) misuse deepfakes. One controversial topic is the idea of using actors’ and actresses’ likenesses after they have passed away. Disney is currently facing a lawsuit over its recreation of Peter Cushing’s face for “Rogue One: A Star Wars Story” (2016). This raises questions about likeness rights in the conversation around AI technology.

How are deepfakes made?

While there are apps out there, like Face Swap Video by Deep Face, that can create deepfakes, they aren’t very effective. They aren’t that great at making convincing deepfakes. Convincing deepfakes require a lot more time and effort to make. However, technology is always improving. Improvements allow this kind of AI technology to work faster and more efficiently. Deepfakes have already become a problem in our society, and the technology is still in its infancy. And, as more easy-to-use deepfake apps release, the bigger problem we all will have to face.

The key to making deepfakes is machine learning. It’s what is making creating deepfakes faster and much more cost-efficient. To make a deepfake, the machine needs to have hours of real video footage of someone. This gives the machine a realistic understanding of what the subject looks like from different angles and lighting. After it understands what the person looks like, the creator combines the trained network with computer graphics techniques to superimpose a copy of the person onto a different actor. While the machine does speed up the processing time, creators still likely will need to make manual tweaks to avoid the obvious deepfake giveaways.

How do you determine if something is a deepfake?

While deep fakes are becoming more convincing as the technology develops, there are still ways you can tell if it’s a deepfake video. One of the first signs that the video’s been altered is if there is face discoloration and blurriness or weird blending where the face meets the subject’s neck and hair. Sometimes it’s hard to tell, but if you pay attention to the subject, you can see the face doesn’t match the surrounding skin or the lighting doesn’t match. Another sign is eye blinking. Typically people blink once every 2 to 10 seconds. Deepfakes often don’t follow this rule. So count the number of blinks throughout the video and see if it adds up.

Sounds can also be used to tell if it’s a deepfake video as well. For instance, deepfakes sometimes have poorly synched sound and video. So, if the sound feels off, it might be a deepfake when you notice a few of the other visual signs. Ultimately, when in doubt, try to find another version of the video online. Likely, the footage was taken from somewhere else on the internet, and you’ll find the original.

Tell signs of a deepfake video

  • Face discoloration
  • Eyes and teeth look unrealistic
  • Eye blinking
  • Lighting is off
  • Badly synced sound and video
  • Blurriness where the face meets the neck and hair
  • Weird blending along the face
  • Try to find another version of the video online

Knowing what’s real and what’s fake

There is a lot of controversy concerning deepfakes in our society. While deepfakes can be used to create some impressive video edits, they can pose a threat to society. As technology progresses, we are likely to see deepfakes more readily available to the public. This means that we as a society need to be prepared to spot deepfakes. It also needs to know how to push back against them when they are promoting a false narrative.

Image courtesy: Truth Syrup/YouTube; ScreenSlam/YouTube