Comment “pulse” and I’ll send you a guide on how to build your it yourself for…

Your face changes color every time your heart beats. Your eyes can't see it, but a camera can, and you can use a tiny change in color to detect deepfake videos. Intel built a deepfake detector using this, so I looked into whether it works. Every time your heart beats, more blood flows through your face, causing tiny changes in how your skin reflects light. But how does a computer catch it, you may ask?

First, it finds your face and picks a few patches of skin, like your forehead and cheeks. A video is just a bunch of pictures, about 30 every second, and every pixel is three numbers, how red, green, and blue it is. So for each patch, it tracks how the green value changes from frame to frame, because that channel usually shows the pulse most clearly.

Line those numbers up, and you get a messy wave. Most of it is noise from things like lighting changes and movement. But the interesting thing is, your heart can only beat about 40-200 times a minute, so the computer filters out everything outside that range. And what's left is your pulse. On a real face, different parts of your skin show the same heartbeat. But a deepfake is generated by AI, not a heart, so those patches won't match.

And when Intel tested its system, it reported 96% accuracy. If you want to build your own deepfake detector to put on your resume, comment the word PULSE and I'll send you a full breakdown on how you can do this.

and I'll send you a full breakdown on how you can do this.