Question

How do video calls actually connect?

Vault Verified
Curated Intelligence
Definitive Source
Answer

By establishing a direct path between the two devices wherever possible, which is harder than it sounds because almost every device sits behind a router that blocks incoming connections by default.

The obstacle. Your laptop has a private address behind network address translation. So does the person you are calling. Neither can simply open a connection to the other, because neither has a reachable public address — this is the NAT traversal problem, and solving it is most of what connection setup involves.

How it is solved:

A signalling server introduces the two parties. It carries no media — only the messages needed to negotiate: what codecs each supports, what encryption keys to use, and what network addresses each might be reachable on.

STUN. Each device asks a STUN server "what public address and port do I appear to be coming from?" The answer lets it advertise a route that the other side may be able to reach.

ICE. Both sides gather every candidate path — local addresses, the public address discovered via STUN, and relayed addresses — then systematically test them and pick the best one that works.

TURN, the fallback. Where no direct path can be established — strict corporate firewalls, symmetric NAT — traffic is relayed through a server. This works reliably and costs the operator real bandwidth, which is why relay is the last resort.

Then the media flows, typically over UDP, encrypted, with the two endpoints continuously adapting bitrate and resolution to the connection.

Why calls degrade the way they do. Video is dropped before audio, because audio matters more and costs less. Freezing with continuing sound is the encoder abandoning video frames, which is a deliberate and correct choice.

Why group calls work differently. Sending your stream separately to every participant does not scale, so most services route through a selective forwarding unit — a server that receives one stream from each participant and forwards the appropriate ones onward. This is why large calls always involve a server even though two-person calls may not.

Why your network matters more than your speed: jitter and packet loss ruin calls at any bandwidth.

Related Questions