Irishman deepfake de-age

Did a deepfake artist take Robert De Niro, Al Pacino and Joe Pesci back in time more successfully than The Irishman?

It’s in the eye of the beholder. But a viral video by the YouTube account “iFake” raises highlights the infinite possibilities for using deepfakes to de-age actors quickly and at minimal expense. The account boasts that it took seven days to make the above video, with free software.

With more time, iFake promised, “the results could turn out even better.”

Martin Scorsese’s gangster epic The Irishman cost in the neighborhood of $150 million.

Also Read: Here’s How Old Robert De Niro, Al Pacino and Joe Pesci Are Supposed to Be During Key Scenes

A few quick observations about the deepfake video: It very successfully de-ages DeNiro, Pacino and Pesci. But one of the challenges The Irishman faced was to de-age DeNiro in character as gangster Frank Sheeran. That means he couldn’t just look as lean and mean as De Niro did when he shot Taxi Driver in his early 30s. He had to look like the burly, brawling Frank Sheeran.

And, of course, the YouTube video entitled “The Irishman De-Aging: Netflix Millions VS. Free Software!” isn’t a fair fight, in that it de-ages De Niro, Pacino and Pesci in just a few scenes, not over a three-and-a-half-hour movie.

Still: The actors do look strikingly youthful in the YouTube video. And the creator says it was all free.

Whether or not big-budget studios are kicking themselves for using expensive CGI when they could be using deepfake technology, the iFake video is sure to fire the imaginations of moviemakers working on lower-budget projects.

Deepfake videos take a person in an existing video and replace the person with someone else’s likeness, using machine learning techniques. They often combine and superimpose older images onto the media they seek to change.

For The Irishman video, “iFake” appears to have replaced the stars in their Irishman scenes with images drawn from their decades-old films.

The creator of the video, which has been viewed for than 700,000 times, didn’t respond to our request for comment.