How do deepfakes work?
The creator gathers a trove of photos or videos of the target—so it helps if it’s a famous person—along with the video to be doctored. The video maker then feeds the data into the app, which uses a form of artificial intelligence (AI) known as “deep learning”—hence deepfake—to combine the face in the source images with the chosen video. This process requires a sizable graphics processing unit and a vast amount of memory. It’s time-consuming—the Obama/Peele video took 56 hours to make—and the quality is variable. But the technology is improving fast. Tech expert Antonio García Martínez, writing for Wired, says