Comment by bityard
4 years ago
> but you insert ultrasound with encoded data
Others in these comments have also suggested steganography in both the video and audio streams. The problem with that is that when you retrieve a video from YouTube, you never get the original version back. You only get a lossy re-encoded version, and the very definition of lossy encoding is to toss out details that humans can't (or wouldn't easily) perceive, including ultra-sonic audio.
It might be ridiculous, but how about uploading a computer-generated video of a human saying 0 and 1 very quickly, to encode binary file.
Or better yet, the file could be one third the size if the human says the numbers 0 to 7.