Here's the thing. It does look fake. I could see how anybody would think to yourself: "why do his hands appear over the microphone? Oh, I know: green screen." Which could be possible, but there's something else weird about that hypothesis. Why doesn't his jacket/waist/etc. also cover the microphone when he walks close enough? When you composite a green screen shot, imagine a pizza: there are layers. The grass and background (Biden included) are the bottom, crust layer. The microphones would be the sauce layer, on top of the crust. If/when Biden comes close enough to the microphone layer, he would be overlaid by them, thereby looking real. Obviously, it looks like that didn't happen. What did? The crust covered the sauce?
If it was an actual green screen, the only way to get the effect shown in the video is to add another layer: cheese. So what's the cheese in this shot? Biden's hands. Not his whole body, only his hands. If you mask out his body and leave his hands, you can use these hands as the cheese. That way, they would appear to cover over the microphones (sauce). The rest of the time we only see Biden's hands from the crust layer.
Okay, so that's how it could be done, but why do it? It would be extra work for the editor in order to oust themselves as using green screen??? That's what I don't understand.
Yes, why add hands? Or what were they obstructing from view?
There was no extra work; just clever technology wielded by incompetent hands.
Greenscreens and cameras have come a long way. Sophisticated filming rigs can capture distance information for every pixel, along with the visual image. Even modern phones have this feature these days, which is why they have multiple cameras. Greenscreen software uses the distance information to decide what parts of the actor(s) go over or under the scene. This way you could easily add someone to a scene with columns, for example, becoming visible in the gaps between them and going out of view "behind" them as they walk. No need for layers if you know how "far" each pixel is. AI can achieve similar effects without the measured distance information, by doing its best to estimate how far things are from just the visual information, just as we do.
Notice that Biden's hands aren't always over the mics. If you look closely, his left hand is actually always behind them (the tip of his middle finger is covered at one point). And his right hand is in front of them only when he leans forward and reaches out to point. When he pulls back and retracts his hand at the end, it goes behind the mics. Looks like the distances just weren't calibrated or calculated correctly.
(post is archived)