WebRTC egress flips WebRTC's traditional peer-to-peer model: instead of two browsers talking to each other, an SFU (Selective Forwarding Unit) acts as the server source, and each viewer is a WebRTC client receiving from it. The viewer's browser uses the same RTCPeerConnection API as for a normal call, but the SDP offer/answer is now one-directional (server-to-client only). WHEP (WebRTC-HTTP Egress Protocol) standardised the signalling so any compliant player can pull from any compliant SFU.
The killer feature is latency. WebRTC delivery hits 200 ms to 1 s glass-to-glass — categorically lower than LL-HLS's 2–4 s. That makes it the only viable protocol for sports betting, real-money auctions, interactive game shows, classroom video, and any "react together" use case where a 2-second gap breaks the user experience. Twitch, Kick, DAZN's betting feeds, and major auction platforms run WebRTC egress in 2026.
The cost is operational complexity. WebRTC egress requires SFU infrastructure (mediasoup, LiveKit, Janus, Pion, Cloudflare Calls), TURN bandwidth budget, ICE handling, and a player stack that knows about NACK, PLI, simulcast and SVC. Multi-CDN distribution is harder because every SFU is stateful per-viewer. Most operators that need WebRTC egress run it from a small handful of regional POPs (5–20 globally) rather than the hundreds-of-POPs CDN model that HLS enjoys.
