The bizarre 1990s action film Face/Off offered viewers the rare spectacle of John Travolta playing Nicolas Cage, and Nicolas Cage playing John Travolta. The two main characters in the film underwent a procedure in which their faces were transposed. It was weird. Ultimately, though, the two actors was only acting like each other. Now, a user experience designer is digitally merging actors’ features and voices to form one talking, gesticulating monstrosity, and the results are mesmerizing.
"I was in the process of teaching myself how to use Adobe After Effects, and realized that this sort of face swapping would be moderately easy to accomplish with the program," says Chase, the semi-anonymous creator of FaceMashups. Chase, a former touring musician with a rock band, is making videos that feature the body and head of, say, Natalie Portman, with the lower half of Will Ferrell’s face blended in, fielding interview questions. These videos are kind of grotesque, but viewers can’t seem to look away: The YouTube channel has been live for less than a month and already has nearly 2 million views.
While still-photo mashups of celebrity faces have been popular for some time, the idea of creating the same effect with live-action video is a recent phenomenon. FaceMashups is similar to the recurring Conan O’Brien segment, "If They Melded," except with more of a concentration on movement. In a mashup of two A Few Good Men actors, Tom Cruise and Jack Nicholson square off in a courtroom, with exchanged face parts and uncanny eyebrow acting. A lot of the face swapping is between opposite-gender stars, though, like Jonah Hill and Jennifer Lawrence.
"When picking the celebrities, I am mainly considering two things. Their relevance and popularity, as well as the availability of unique, high-quality footage in which the actor is looking mostly towards the camera," Chase says. "Mashing up footage in which the characters are constantly looking side to side is much more difficult and usually results in a less convincing final product." He adds. "There have a been a few After Effects sessions that ended up in the recycle bin because of this."
After Chase decides on a target and finds the best footage to use, he takes the video of the subject providing the mouth and voice, and edits this person’s footage down to just the desired clips—excising any parts that may not work. Next, he takes the footage of the other celebrity—the body and upper face—and looks for any movements or gestures that might match how the other celebrity is talking. These clips are then cut and slipped behind the first footage. Indeed, the funniest moments in the videos are when a change in the voice’s inflection goes along with sudden movements from the body. It’s this touch that separates FaceMashups from anything similar.
"Through some clever—and tedious—masking, motion stabilization, and motion tracking, I am able to lock the superimposed mouth onto the other actor’s head so that it follows their movements realistically. A typical one- to two-minute video takes about four hours to make, but I’m getting faster at it every time."
Have a look at some of the best Face Mashups in the slides above.