Web Calls

This page covers the technical implementation and architecture for low-latency WebRTC browser audio streaming in Sayvy AI.

Latency Budget Breakdown (Sub-500ms)

Content coming soon.

Audio packetization (20ms frames)
Streaming Speech-to-Text (STT) pipeline
LLM time-to-first-token (TTFT) streaming
Incremental Text-to-Speech (TTS) chunking at sentence boundaries

WebRTC Transport & Audio Codecs

Content coming soon.

Opus audio codec (24kHz / 48kHz)
WebRTC Data & Media Channels configuration
Jitter buffer management and packet loss concealment (PLC)

Browser Client Audio Lifecycle

Content coming soon.

Microphone capture constraints (echoCancellation, noiseSuppression)
Audio track initialization and handshake
Network drop recovery and automatic session reconnection