NVIDIA has developed a technique of video compression using AI, replacing video codecs with a neural network. Instead of sending a compressed video stream, their algorithm sends the location of key facial points and then essentially animates a deepfake of yourself on the receiver’s computer. Additionally, their technique can also align faces so that you’re avatar is actually looking at the person you’re video conferencing with instead of appearing to stare off screen because your camera and monitor aren’t aligned right.

So obviously I’m replacing my face with Chris Hemsworth on all future Zoom calls because why would I use my own face when I can use his face? My girlfriend already makes me tape a printed out photo of him to my face so this feels like the next logical step.

Source: Geekologie – Nvidia creates AI video compression for video conferencing