When Harry Potter first made it to Hogwarts, one of the things that stumped him the most were the animated paintings and pictures.
Well, it turns out wizards aren't so great, because mere Muggles are figuring out how to do with technology what they did with their magic.
Researchers at Samsung's AI Center have been working on an artificial intelligence?along the lines of DeepFakes. You know, the one that let you skillfully?edit faces of people onto other people's bodies?in footage.
Except here, the scientists aren't changing faces on footage. Instead, they're animating the faces in iconic paintings and photos, and essentially bringing those characters to life.
For instance, the team took the Mona Lisa, Marilyn Monroe, and Einstein's iconic photo. They used something called facial landmarks (basically defining points on the face for animation purposes) to sync the target faces with source faces (footage of someone talking for instance).
That way, when the source face does something, it's mimicked on the target face too.
This has of course been done before with AI, but it usually requires a massive amount of data to work with. In this case however, the Samsung AI Center was able to even work with a single image of a target face, making it yawn, roll its eyes, and much more. It's basically much more comprehensive control given even less data than before.
It's not flawless of course, but that's only a matter of time to change. Eventually, the magical talking portraits could actually be a thing, especially if the researchers figure out a way to also sync it with a conversational algorithm in real time.