Mona Lisa Speaks!

Well, she looks like she speaks, anyway, after the iconic image is processed through a model developed by the Samsung AI Center.

The machine learning program maps facial landmarks on a source face and applies them to the landmarks on a target face, and then makes the source face do what the target face does. So the brief clip a little ways down the page shows Mona Lisa's face mimicking the speaking motions of a particular target face. The program has a couple of different ways of mapping, and some look better for some faces than others.

I'm not so good at lipreading, and of course I don't speak Italian, but if we were to get animation of the actual Mona Lisa speaking I bet it would be something like, "Can you hurry up, Leo? This robe is sweltering!"