Peer-to-peer video calls: how they work and what they protect
Last updated: November 2026
In most video calls your camera does not go straight to the other person. It goes to the provider’s server, and the server sends it on. In a peer-to-peer call the computers talk to each other directly. Here is what that changes, and what it does not.
What happens in a peer-to-peer call
Two computers open a direct connection and send their audio and video over it. The technology is WebRTC, the real-time engine in every browser and in many desktop apps.
Before that, the two need to find each other. A signaling server passes on the connection details. A STUN server tells each computer the address under which the internet sees it. Once the call runs, audio and video no longer pass through either of them.
How the call is encrypted
WebRTC has no unencrypted mode. The two computers agree on keys with a DTLS handshake and encrypt audio and video with SRTP (together: DTLS-SRTP).
The keys exist only on the two computers. Anyone in between, your internet provider or a café Wi-Fi, sees encrypted packets.
What a server in the middle can see
It depends on the kind of server.
- A media server (an SFU, used by most meeting services) receives every stream and sends it on to the others. In the usual setup it decrypts the media to do that. Some services add end-to-end encryption on top: Zoom offers it as an option that turns off cloud recording, and Discord uses it by default for voice and video calls.
- A TURN relay only forwards packets when no direct path exists. It never gets the keys, so it moves the call without being able to read it. It does see who talks to whom, when and how much.
So "it goes through a server" alone says little. The question is whether that server holds the keys.
The limits of peer-to-peer
- It stays small. In a mesh every computer sends its stream to every other one. With five people, each computer sends four streams and receives four. That is why services switch to a media server as a call grows: Jitsi Meet uses peer-to-peer only for calls of two, Tuple up to three people, Whereby for small rooms.
- Some networks block it. Company firewalls and some routers do not allow a direct connection between two computers. Then the call needs a TURN relay, or it does not connect at all.
- The helpers see metadata. The signaling server and the STUN server see IP addresses and when you connect. They do not see the call itself.
How Møt does it
Møt is a peer-to-peer call for up to five people. Audio and video go encrypted from computer to computer. Our signaling server (signal.kveik.studio) introduces the computers; by default Google’s public STUN server tells each computer its address.
Møt has no relay of its own. If your network blocks direct connections, the call does not connect, and Møt does not hide that. You can enter your own TURN relay under Settings › General › Connection. It forwards the encrypted packets and cannot read them. Changes apply from the next call.
If you need a large meeting, a cloud recording or guests joining from a phone, a meeting service with its own servers is the better choice. Møt vs Zoom shows where each fits.
Questions, answered
What is a peer-to-peer video call?
A call in which the computers send audio and video directly to each other instead of through the provider’s server. A signaling server only helps them find each other.
Are peer-to-peer video calls encrypted?
Yes. WebRTC always encrypts media with DTLS-SRTP. The keys are agreed between the two computers, so only they can decrypt the call.
Are P2P calls more private than Zoom?
On the media path, yes: no provider server ever holds your audio and video. Zoom can come close with its optional end-to-end encryption, which turns off cloud recording. P2P stays small, though, and some networks need a relay.
Why does a peer-to-peer call sometimes not connect?
Some company firewalls and routers block direct connections between two computers. Then the call needs a TURN relay. Møt does not bring one; you can enter your own in the settings.