Audio Streaming

Real-time bidirectional voice streaming over WebSocket for AI voice-bot integrations. Fonada streams live call audio to your own ASR → LLM → TTS stack over a single WebSocket. Fonada runs the telephony, carrier connectivity, and transfer to agents; you own the voice intelligence.

On Fonada's side the call path is Asterisk plus a media connector: the caller talks to Fonada, and Fonada forwards PCM16 over wss:// to your WebSocket server.

This guide is for the engineering and support teams integrating a voice bot with the Fonada calling platform. It describes the full integration surface: the supported call flows, the message protocol, the audio format, and the operational requirements your endpoint must satisfy.

Fonada provides the end-to-end telephony — phone numbers, carrier connectivity, and call handling for both inbound and outbound traffic — together with live media streaming and transfer to human agents. You provide a single WebSocket endpoint containing the voice intelligence: speech recognition (ASR), the conversation and business logic, and speech synthesis (TTS).

Division of responsibilities

In short, Fonada owns the telephony and audio transport; you own the voice intelligence and the decisions that drive the conversation.

Fonada is responsible forYou are responsible for
Phone numbers and carrier connectivitySpeech recognition (ASR)
Inbound call answering and routingConversation and business logic
Outbound call origination and diallingSpeech synthesis (TTS)
Audio transport in both directionsDeciding when to transfer or end a call
Transfer to human agentsDetecting caller interruption (barge-in), if required
Call teardown and cleanupHosting and operating the WebSocket endpoint

The integration surface

The integration surface consists of two components:

  • WebSocket endpoint (you host). Fonada connects to this endpoint for every call and streams audio in both directions. This is the bulk of the integration work and the primary subject of this guide.
  • REST API (Fonada hosts). Your bot calls this API to bring a second party onto a live call. See Consult & conference.

Start with the Quickstart echo bot, then the protocol and audio pages. In-call control is on Consult & conference. REST details are on the API reference.

Sandbox

The sandbox lets you exercise the integration end to end before going live. Request a sandbox credential from your onboarding contact. Sandbox calls are billed at zero and are isolated from your production numbers.

Use the API reference for consult, conference, and session endpoints, and the echo bot to point Fonada at a sample server and hear yourself on a test call.