RED (Redundant audio data, RFC 2198) is a simple, effective way to survive packet loss in real-time audio: each RTP packet carries not only the current frame but also a copy of one or more recent frames. If a packet is lost, its payload arrives anyway inside the next packet, so the receiver reconstructs the gap with no retransmission and no delay. It is codec-agnostic — wrapping whatever payload you use — and trades bandwidth for resilience, the redundancy roughly multiplying the audio bitrate. WebRTC uses RED, often combined with Opus's own in-band FEC, on lossy paths. The more redundancy you add the more loss you survive, but the more bandwidth you spend.
