Summary

The technology of deepfakes is not itself anything new or unique. These techniques existed in various forms long before they were applied to face-swapping, but deepfakes have caught public attention in a way that other AI techniques have never really been able to. There is something very visceral about seeing a face where it doesn’t belong, seeing an actor in a role you know that they didn’t play, or seeing your own face doing something you’ve never done.

While the techniques that make up deepfakes have all existed previously on their own, together, they provide completely new possibilities. There are numerous use cases that deepfakes can be applied to, from stunt-double replacement to advertising. The technology is here, and its use will only grow as more and more industries find ways to use it.

There are still limits to the capabilities of generative AI. Knowing what a deepfake cannot do is as important as knowing what it can do. Especially regarding data, knowing how to work around those limitations is key to a quality result.

We’ve given an overview of deepfakes, covering what they are, what they can be used for, how they work, their limitations, and the existing software you can use to make them. In the next chapter, we’ll cover the potential dangers of deepfakes and talk about the ethical questions that the technology brings with it.