SFU
What is an SFU?
A selective forwarding unit (SFU) is a media server that receives each participant's audio and video once and forwards it to everyone else in the call, choosing which streams and which quality each person gets. Each person uploads one stream instead of one per participant, which keeps group calls stable on ordinary connections.
How does an SFU work?
In an SFU call, every participant sends their camera, microphone and any shared screen once, to the server. The server does not decode the video or paint a combined picture. It forwards the packets to the other participants and chooses, for each viewer separately, which streams to send and at what quality. A participant whose tile is hidden does not need to receive that video; a viewer on a weak connection can receive a lighter version when the sender uses simulcast.
| Design | How the video travels |
|---|---|
| Peer-to-peer (mesh) | Each device sends a separate copy to every other participant. Fine for two people; the upload grows with every extra person. |
| SFU | Each device uploads once; the server forwards streams and picks the right ones for each viewer. Each viewer keeps their own layout. |
| MCU | The server decodes everyone and mixes one combined picture. Light for viewers, heavy for the server, and everyone sees the same layout. |
Why does an SFU matter in a video call?
Upload speed is usually the scarce part of a home or mobile connection. In a mesh call with six people, each device has to upload five copies of its video; with an SFU it uploads one. That difference decides whether a group call on an ordinary connection stays smooth.
The server also gives the call a place to adapt. It can send a phone on mobile data a smaller picture while a desktop on fibre keeps full quality, and it can stop sending video nobody is looking at. The trade-off is that media passes through a server: its distance from participants affects delay, and because each connection's encryption ends at the server, an SFU-based service is not end-to-end encrypted unless it adds a separate layer of encryption on top.
How does Meeto use an SFU?
Every Meeto call runs through Meeto's media servers. When a meeting is created, it is pinned to the nearest media region, Europe or South-East Asia, and everyone in that meeting connects there. Media uses WebRTC (DTLS-SRTP) encryption in transit; Meeto does not add an end-to-end layer, so it is not end-to-end encrypted.
Each viewer receives only the quality their tiles need. On a weak network the call steps down by itself, first lowering the picture, then showing only the active speaker's video, then switching to audio only, and it steps back up when the connection recovers. Layouts such as Grid, Speaker and Sidebar change only your own screen. The laggy call guide explains what you see when this happens.
Frequently asked questions
Is an SFU the same as a TURN server?
No. A TURN server only relays the packets of one connection when a network blocks the direct route, without deciding what to send. An SFU takes part in every connection of the call and decides which streams each person receives. One call can use both: device to TURN relay to SFU.
Does Meeto switch to peer-to-peer when only two people are in a call?
No. Every Meeto call, including one-to-one meetings and direct calls between contacts, goes through Meeto's media servers. The network handling is the same for two people as for a full room.
Does an SFU record the meeting?
Not by itself. Forwarding media and saving it are separate jobs. In Meeto, cloud recording is an opt-in feature for Pro and Team hosts, and everyone in the call sees an indicator while it runs.
How many people can an SFU call hold?
It depends on the server capacity and on each participant's bandwidth, so services set a practical limit per plan. A Meeto call holds up to 10 people on Free and up to 25 on Pro and Team.
Try your first browser meeting
Create a room, share the link and invite someone to join you.