Apple Patents Deepfake Tech to Manipulate Photos
The U.S. Patent Office has granted Apple a patent for a deepfake technology that can alter someone’s face in a photograph. Apple may fold the “Face Image Generation with Pose and Expression Control” patent into a new tool for creating synthetic media just as the AI image generation industry continues its recent rapid expansion.
Synthetic Face Control
The patent describes how Apple can apply AI to a reference image to adjust a human face’s expression and bearing in subtle and overt ways. The picture of a smiling person could become a photo of someone frowning or sticking their tongue out. It could theoretically be used to produce enough images strung together to mimic someone talking in a film. Apple makes a point in the patent that the technology can change and alter an existing image, but doesn’t produce the images from scratch in the way OpenAI’s DALL-E, Midjourney, and other platforms operate. Like those tools, however, the manipulated image is not a real photograph.
“At inference time, a single reference image can generate an image that looks like the person (i.e., the subject) of the reference image, but shows the face of the subject according to an expression and/or pose that the system or method has not previously seen,” the patent explains. “Thus, the generated image is a simulated image that appears to depict the subject of the reference image, but it is not actually a real image.
The best comparison might be Apple’s Animoji and Memoji apps, which use a video camera to overlay a user’s face with a cartoonish 3D character capable of following mouth movements well enough to imitate someone speaking. The tool is also sensitive enough to follow along when someone sticks their tongue out or raises their eyebrows, though not much beyond that. Deepfake technology to play with human faces is just starting to crop up in DALL-E’s features. Apple’s description also sounds like what the British deepfake tech developer Flawless did for survival action movie “Fall” by changing the mouth movements of actors to not swear as a way to lower the film’s R-rating.