Fun fact, live A/V can not be cryptographically verified in real time.

You have you batch up 5-10 seconds of data and sign that.

If you try to sign each frame/packet, your CPU won't be able to sign fast enough. This is why e2ee videoconferences are symmetrically encrypted and all content is unsigned. You know it came from someone in the conference, but you can't prove from whom (cryptograohically speaking).

Maybe with custom encryption hardware this could be done fast enough that humans wouldn't notice the lag of the pipeline:

capture frame -> hash -> sign -> encrypt -> tx -> rx -> decrypt -> verify -> play

It'd still be tricky to get the data around to each component fast enough. Having an FPGA with an integrated NIC that would accept raw audio and video frames and deal with hashing, signing, encryoting, packing them up and shoving them over to the switch all in one device would be pretty slick.

In a world of quality deepfakes made easy, this could be just the ticket to restoring faith that something is genuine.