Transcoding is the process of decoding media from one codec format and re-encoding it into another. In WebRTC, transcoding is generally avoided because it is computationally expensive and adds latency.
When transcoding is needed in WebRTC
- PSTN interop: Converting between WebRTC's Opus audio and the PSTN's G.711
- Legacy system bridging: Connecting WebRTC endpoints to SIP devices that don't support modern codecs
- Codec mismatch: When two peers do not share a common video codec (rare in modern WebRTC)
- MCU mixing: MCUs inherently transcode as they decode, mix, and re-encode media
Alternatives to transcoding
WebRTC has developed several techniques to avoid transcoding:
- Simulcast: Sending multiple encodings so the SFU can forward without transcoding
- SVC: Scalable layers that can be selectively forwarded
- Codec negotiation: SDP ensures both sides agree on a common codec before media flows
Transcoding a video stream can increase server CPU costs by 10-100x compared to simple SFU forwarding, making it a significant factor in infrastructure cost calculations.
