Mostly it is latency added by webcams, microphones(But our ears are used to delayed audio anyway) and the software. Your webcam probably has >100ms of latency then the app adds >50ms then multiply that by 2x (for two way conversation). Even a display like a TV can have loads of latency.

If you just open up a camera view there's huge latency without it even sending data anywhere. The internet latency is probably only a small portion of over all latency and WiFi latency isn't much either. There's already lots of latency with audio (Perhaps having a microphone could make audio travel faster then through air?)

I wish people would take latency more seriously. Using a rolling shutter, encoding a single line at a time and sending it via UDP could probably save almost all the latency. Sub 30ms latency IMO would make it much better.