Deepfaking Mona Lisa

From a Cornell University paper by Egor Zakharov, Aliaksandra Shysheya, Egor Burkov, and Victor Lempitsky titled "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models"

Several recent works have shown how highly realistic human head images can be obtained by training convolutional neural networks to generate them. In order to create a personalized talking head model, these works require training on a large dataset of images of a single person. However, in many practical scenarios, such personalized talking head models need to be learned from a few image views of a person, potentially even a single image. Here, we present a system with such few-shot capability. It performs lengthy meta-learning on a large dataset of videos, and after that is able to frame few- and one-shot learning of neural talking head models of previously unseen people as adversarial training problems with high capacity generators and discriminators. Crucially, the system is able to initialize the parameters of both the generator and the discriminator in a person-specific way, so that training can be based on just a few images and done quickly, despite the need to tune tens of millions of parameters. We show that such an approach is able to learn highly realistic and personalized talking head models of new people and even portrait paintings.

Image: Egor Zakharov/YouTube Read the rest

Watch a child magically materialize in the background of this BBC News clip

In this BBC News clip, a child seems to materialize just behind the woman speaking. WTF. Unfortunately this isn't likely a fun glitch in our simulated reality but rather something with much more insidious potential. From WAXY:

If you watch the woman’s face at the same time the boy appears, you can see her expression morph into a smile.

This technique is known as a Morph Cut, a feature added to Adobe Premiere Pro in 2015, intended to smooth transitions in interview footage, removing unwanted pauses, stutters, and filler words (“like,” “um,” and “uh”) without hard splices and cuts.

The results, when used appropriately in interview footage without a changing background, can be nearly seamless.

It’s likely that BBC News used a morph cut in the clip above to tighten up the interview without changing its meaning. But it’s also ripe for abuse and fully capable of altering the meaning of an interview, and in many cases, undetectable.

Another demonstration of the technology:

Read the rest

Faked video gets much better

Researchers have developed a generative neural network to make fake videos of talking heads. It's getting a lot harder to tell the difference between real and deep fakes.

We present a novel approach that enables photo-realistic re-animation of portrait videos using only an input video. In contrast to existing approaches that are restricted to manipulations of facial expressions only, we are the first to transfer the full 3D head position, head rotation, face expression, eye gaze, and eye blinking from a source actor to a portrait video of a target actor.

Read the rest