WebRTC Aggressively Drops Audio Packets to Keep Latency Low
Check your WebRTC implementation for packet drops; add a 200 ms buffer to improve audio quality.
Patch your WebRTC implementation to add a 200 ms buffer for audio packets.
Summary
WebRTC is engineered to drop audio packets when network conditions worsen, prioritizing low latency over audio fidelity.
The article explains that during a conference call, distorted audio is a common symptom because WebRTC refuses to wait for packet retransmission. This design choice is intentional: real‑time communication cannot afford the delay that would come from buffering or retransmitting. Users who value accurate prompts would prefer a 200 ms wait, but WebRTC’s hard‑coded logic prevents any such delay. Even attempts to retransmit packets in browsers, such as Discord’s experiment, failed because the implementation is locked to real‑time latency. The result is that voice AI services like OpenAI’s rely on WebRTC for speed, but at the cost of occasional audio distortion. Developers must be aware that WebRTC’s packet‑dropping behavior is a trade‑off baked into the protocol, not a bug that can be patched in the browser.
Key changes
- WebRTC drops audio packets when network degrades to keep latency low
- Audio distortion occurs during conference calls
- Cannot retransmit audio packets within browser
- Implementation hard‑coded for real‑time latency or else
- Users would prefer 200 ms wait for accurate prompt
- OpenAI’s voice AI uses WebRTC for low latency
- Discord attempted retransmission but failed