Conversational Phone Agent with Natural Interruptions

A voice agent that handles inbound and outbound phone calls with natural conversation flow, supporting user interruptions and multi-turn dialogue. Uses streaming speech recognition and text-to-speech for low-latency responses over standard telephone lines.

No demo yetView source Fork template

The numbers

latencyReal-time streaming with built-in VAD

cost / minLow cost per minute (GPT-4o-mini + Deepgram + OpenAI TTS)

frameworkPipecat

The stack

telephonyPlivo

speech-to-textDeepgram Nova-3

llmGPT-4o-mini

text-to-speechOpenAI gpt-4o-mini-tts

System prompt

No prompt published.

Config

config.json

{
  "tts_voice": "alloy",
  "sample_rate": 8000,
  "vad_enabled": true,
  "audio_format": "mulaw",
  "add_wav_header": false,
  "available_voices": [
    "alloy",
    "ash",
    "ballad",
    "coral",
    "echo",
    "fable",
    "nova",
    "onyx",
    "sage",
    "shimmer",
    "verse"
  ],
  "allow_interruptions": true,
  "vad_audio_passthrough": true
}

Voice AI recipes, picks, and analysis.

Get the useful new templates plus the occasional teardown of what’s working in production voice AI.