1 of 2
Screenshot
An advanced AI algorithm can be used to convert the "Mona Lisa" into a convincing moving portrait.

SALT LAKE CITY — Deepfakes are getting smarter, and can create moving pictures using a single portrait, according to a new research study.

The Verge reports that a new artificial intelligence algorithm developed by Samsung’s AI Center can convincingly animate a model face using a single image. The AI has been trained to recognize basic landmark facial features using 7,000 images of celebrities.

A piece of reference video footage can then be used to dictate how the still image is animated, resulting in some passable doppelgangers.

A video on YouTube summarizing the research shows how the model can be applied to iconic portraits to create moving images.

It’s not quite up to the quality of “Harry Potter” magical photos, but it’s fascinating to see Albert Einstein, Marilyn Monroe and the Mona Lisa look around, smile and talk — complete with terrifyingly real teeth.

Comment on this story

This isn’t the only advancement in deepfakes recently. I previously reported on two other cases for Deseret News. Last week an AI was trained to imitate standup comedian Joe Rogan using over 1,000 hours of audio with convincing results.

The Dali Museum in Florida also has an exhibit in which Salvador Dali greets visitors and talks about current events and the weather before snapping a selfie. The interactive display was made using existing footage of Dali paired with a body double and voice actor.