The vast majority of what YouTube does on a technical level is ingesting a ton of uploaded user video, encoding it in dozens of combinations of resolution, framerate, quality, and codec, then seamlessly choosing which version to serve to requesting clients to balance bandwidth, perceived quality, power efficiency in the data center, power efficiency on client devices, and hardware support for the client. There’s a lot of stuff going on behind the scenes, and there’s a reason why the user experience is much more seamless on YouTube on a shitty data connection than, say, Plex on a good data connection.

No, it doesn’t need to be realtime, but people with metered or throttled bandwidth might benefit from downloading just in time video at optimized settings.