I’ve worked a bit in the computer vision field. I know people who are experts in the field. Expert opinion from former colleague is that the Tom Hanks SNL video is a deep fake.
Perhaps some of you have seen the video of the “disappearing Pope”. Probably also a deep fake (IMO).
That opens up a boatload of questions.
Think of all the people we’ve rarely seen over the last few weeks/months/years, except on the occasional video. Or still shot. How many of them are likely deep fakes? Still shots are MUCH easier to fake.
Klintoons, 0vomits, Biden, Brennan, Comey, Podestas, Gates, NP, .....
If I were making deep fakes to fool the public, I’d put my max effort into those that get the most views and those that are considered the most important people (Pope, Biden) or those whose existence/location are being questioned the most (TH).
A “talking head” video of someone with a static background is much easier to fake than a full body shot of someone walking around. The tech is much more focused on getting the face right than on details like the hands. That’s why in the TH SNL video, the young actor’s hands aren’t changed and a giveaway that this is not TH who is over 60.
Videos of Biden, 0vomit are going to get LOTS of views, so if they are deep fakes, it’s a good reason to make them head shots only. Fewer giveaways that they are fake. No hands. No walking so you can’t do a gait comparison. Not a lot of body movement, so no comparison of mannerisms, stance, etc. Not a lot of head motion, just a face staring straight at the camera.