OpenAI's WebRTC Problem

WebRTC prioritizes low latency above all, as in meeting calls, so it has the characteristic of aggressively discarding audio packets when network conditions are poor.

However, in Voice AI environments, rather than slower response speed, corrupted voice prompts due to packet loss can be a more fatal factor that degrades the model's response quality.

Since TTS has the ability to generate audio faster than real-time playback, buffering can be used on the client side to mitigate quality degradation caused by network instability.