Pipecat: Real-time Voice AI

This talk covers building real-time voice AI applications, addressing latency, pipeline synchronization, and resource management using open source tools like Pipecat with practical demos.

Overview

I’m interested in presenting on real-time AI applications and developer tools with a focus on real-time voice AI applications. The presentation would cover:

Overview of real-time AI
Focusing on voice AI, what it takes to build out voice AI applications, including specific technical challenges around latency, pipeline synchronization, and resource management
Sharing the state of open source AI tools and frameworks
Using Pipecat, which is an open source framework that I help maintain, show how to build a real-time voice AI application, demonstrating:
- Frame-based pipeline architecture for handling concurrent processing
- Real-time speech-to-text, LLM processing, and text-to-speech integration
- Code walkthrough of a basic voice assistant implementation
Giving a quick voice AI demo showcasing natural, low-latency conversation flow

Links

https://github.com/pipecat-ai/pipecat
Pipecat is a Python framework for real-time voice/multimodal AI agents using pluggable pipelines.
https://demo.dailybots.ai/
This project demos leading LLMs: Claude, Llama, GPT, Grok, Gemini.

Tech stack