Picture → sound → picture
A picture is sent as a sound. Each column of the picture gets a tone of its own, and how loud that tone plays is how bright that pixel is — 480 of them sounding at once, a new row of the picture every 60 milliseconds. A second device listens, measures every tone, and draws the rows back out as they arrive. Ticking Lo-fi trades detail for reach: 320 lower tones that small speakers and cheap microphones can actually carry, taking twice as long — the listener hears which one is coming and follows.
The two devices are not connected to anything.
No pairing, no network, no second app. One of them makes a noise and the other one listens to it, so whatever the room does to the sound is what you see in the picture.
Nothing is asked for until Start listening is pressed. Stop ends the stream — the browser's recording indicator should go out with it.
Where the tones are meant to land, and what the row reader is indexing with. Compare it against the live spectrum above: energy on the teal comb and nothing between its teeth is what a clean picture looks like.