Get Started

XiaoZhi (xiaozhi-esp32) is open-source voice interaction firmware for ESP32 boards: the device wakes, captures and plays audio, while recognition and LLM conversation happen on the server. This page walks you from preparation to your first conversation.

First, understand who does what: wake word detection runs offline on the device; recognition, LLM conversation and speech synthesis run on the server. You do not need your own server to get started — see Self-Hosted Server when you want control over your own voice data.

Have a device? Three steps

1. Download & flash firmware
Pick the prebuilt variant for your board; flash without a toolchain
2. Wi-Fi setup & activation
Hotspot or BluFi provisioning, activation code and console binding
3. Start using it
Voice dialogue, device control and personalization

For network issues start with Network Setup; for anything else see the FAQ.

DIY route: from parts to conversation

What XiaoZhi can and cannot do
Five minutes on the capability boundary to avoid misconceptions
Build one on a breadboard
Parts, wiring, flashing, provisioning and activation in one pass
Hardware Setup Guide
ESP32-S3 assembly, wiring diagrams and parts list
ESP32 Budget Version
Low-cost hardware option

Go further