Brussels / 4 & 5 February 2023


Real-time audio/video conferences in Linphone thanks to a modern SFU server

Existing audio/video conferencing solutions of type “Multipoint control unit” or “Mediamixing mixer”, have proved to be costly in terms of resources in the case of video processing, due to the algorithm complexity of video codecs. This constraint leaded to the emergence of new solutions transmitting a useful subset of the streams received from all the participants, to each participant. For example: only the video of the actively talking participant is dispatched to each participants, together with some very low resolution video streams of the other participants. This technology is referred to as “Selective Forwarding Unit” (SFU) or “Selective Forwarding Middlebox” (SFM). These solutions open the path to conference sessions managed by a server that does not decode the media content. The main benefit of a Selective Forwarding Unit is that it is capable of receiving multiple media streams and then decide which of these media streams should be sent to which participant. That way it is technically possible to enhance the number of participants to a group call.

Linphone is likely to be the first open source SIP-based conferencing solution powered by a modern SFU server. As the server solution already used for Instant Messaging features was already compatible with open standards for the function of managing group chatrooms and their participants., the idea was to enhance this component with media capabilities. Therefore the Linphone team developed a SFU algorithm on top of mediastreamer2 and ortp (Linphone’s media processing libraries), inside of a new software component called “MS-SFU”.

This talk will focus on the following topics:

  • compatibility of the conference establishment mechanism with SIP and the RFC 4579 (SIP Call Control - Conferencing for User Agents)

  • stream selection performed by the SFU based on the current audio level of participants

  • implemented standards to allow the communication between the Linphone's library and the Flexisip conference server in SFU mode

  • how this technology paves the way for a future major innovation: end-to-end encryption for real-time audio and video conferences


Photo of Jehan Monnier Jehan Monnier