Real time voice AI agents are coming to town (and well before Christmas)

This talk demonstrates how to build a low-latency conversational voice AI agent using WebRTC and smart turn-taking, leveraging open source Voice AI tools.

Overview

Pipecat and LiveKit are the leading open source projects that fuel the major players inside Voice AI arena (including OpenAI’s ChatGPT voice models) and their adoption is increasing over time due to their stability, scalability and continous development.
At the same time operational costs for conversational real-time Voice AI models (Speech-To-Speech) are constantly decreasing and quality of voice recognition and speech generation is everyday better and better.
This situation enables an accelerated adoption rate of these technologies in Enterprises and SMEs and can grow substantially in the next six months.
During the presentation we’re going to implement a conversational voice AI agent based on WebRTC protocol and smart turn-taking capabilities.

Links

https://github.com/pipecat-ai/pipecat
Pipecat is a Python framework for real-time voice/multimodal AI agents using pluggable pipelines.
https://github.com/livekit/livekit
LiveKit: Go-powered WebRTC SFU media server and client/server SDKs.

Tech stack