MatrixRTC is the Matrix ecosystem's architecture for real-time voice and video communication, including group calling. It uses Matrix rooms and membership signalling to coordinate participants while real-time media is carried using technologies such as WebRTC and selective forwarding infrastructure.
How Matrix coordinates a real-time session
A call is represented through Matrix state and events so participants can discover that a real-time session exists and determine who is taking part. For scalable group calls, clients connect media through an SFU rather than sending a separate full-quality stream directly to every other participant. MatrixRTC work has been integrated into the broader Matrix 2.0 direction, alongside newer clients and faster synchronisation. The approach keeps call membership and room identity inside Matrix while using specialised media infrastructure where low latency demands it.
Why media needs its own infrastructure
Voice and video are difficult to federate because media bandwidth behaves very differently from small chat events. MatrixRTC tries to preserve Matrix's room model and decentralised identity without pretending that high-quality multiparty video can be routed like text messages. The result is a layered system: homeservers coordinate the social context, while real-time components optimise media transport.
Federated signalling is not peer-to-peer media
Decentralised signalling does not mean every media packet passes through a user's own homeserver. Production group calling can depend on SFUs and TURN infrastructure, with associated hosting and trust considerations. MatrixRTC has also evolved rapidly, so operators should follow current Matrix documentation rather than older Jitsi-based or experimental calling guides when deploying it.
Related reading
- What Is Simplified Sliding Sync in Matrix and Why Did It Replace the Old Proxy?
- What Is Blossom? Nostr's Hash-Based Media Hosting Protocol Explained