23
A comment on my fact-check about the 'lost city' video made me rethink image verification
Last month I posted a breakdown of that viral clip claiming to show a 'lost city' found in the Amazon. I used reverse image search and it came up clean, so I called it plausible. A user named 'GeoVerify' commented, 'You only checked the stills. The cloud movement in the background of the video loop is identical every 7 seconds. It's a CGI asset.' I went back and checked frame by frame, and they were right. I was only looking for copied photos, not analyzing the video file itself. Now I always run videos through a metadata checker and look for repeating patterns before I even start on the story. It added about 20 minutes to my process, but it catches way more fakes. Has anyone else had a basic check fail them and had to add a new step?
4 comments
Log in to join the discussion
Log In4 Comments
holly7092mo agoMost Upvoted
Totally been there. I used to just check if a photo was online already. Missed a fake where someone mirrored the left side of a real disaster photo to make it look bigger. Now I always flip the image horizontally myself to check for perfect symmetry. It's a dead giveaway for lazy editing.
5
Close but not quite on that mirrored photo trick. Flipping it horizontally works for some stuff, but the real giveaway is checking brightness values across the image. That's the thing that catches those blended copies where someone just darkened one side and mirrored it. The human eye misses those little mismatches, but a histogram tool picks them up in seconds. Still, your idea is a good fast check for the obvious ones, I usually do that first too before digging into the numbers.
2
lucasw842mo ago
Adding 20 minutes to check every video seems like a lot of work for a clip of some trees and clouds. Most people just scroll past that stuff anyway.
1
zara_sanchez2mo ago
Yeah, the "trees and clouds" stuff is exactly what gets through... I run a quick reverse image search on a cropped section, like just the sky. It takes two minutes and catches a lot of reused backgrounds.
2