Loading…
1
Question about how I flipped my take on blurry footage in deep fake spotting
Last Tuesday I was at a bar in Kansas City and this guy at the next stool kept insisting any blurry clip with a missing ear shadow is a deep fake, so I told him that was nonsense. Then he pulled up a 2019 clip of a mayor whose jawline shifted on frame 412 and matched it to the same glitch on a known fake from the same software, and I sat there quiet for a minute. I had always brushed off low res video as just bad compression, but that one frame pairing made me think about how fakers hide flaws in soft focus on purpose. Now I go through grainy stuff frame by frame instead of skipping it. Anybody got a go to way to tell a compression smear from a real jaw warp?
1 comments
Log in to join the discussion
Log In1 Comment
fionafoster25d ago
That frame 412 trick works the other way too, since compression smear drifts across the whole face while a real jaw warp stays locked to the bone. Pull the clip into any free tool that shows motion vectors, and if the smear moves one way while the jaw moves another, that tells you plenty.
3