Home Assistant voice assistant for ESP32-S3-BOX-3

What you’re building

Home Assistant’s first-party tutorial explicitly supports ESP32-S3-BOX-3 and BOX-3B, using a dedicated ESPHome configuration for the board’s display, microphones, speaker and controls. The precompiled browser installer joins 2.4 GHz Wi-Fi, discovers the device through ESPHome and turns it into an Assist satellite that shows listening, processing, reply and timer states. Users can run wake-word detection on the BOX-3 or in Home Assistant, then send recognized requests through the configured Assist pipeline and hear its response. The expected outcome is voice control of authorized Home Assistant entities from a self-contained desk terminal. This is not automatically a fully local assistant: speech recognition, conversation and text-to-speech follow the selected cloud or local pipeline. Initial flashing requires desktop Chrome or Edge, direct USB rather than the dock, and the exact BOX-3 target.

Key steps

1. Confirm the label is ESP32-S3-BOX-3 or BOX-3B, prepare Home Assistant Assist, and open the first-party installer in desktop Chrome or Edge.
2. Connect USB-C directly to the BOX-3 rather than its blue dock; if no serial port appears, hold Boot while tapping Reset to enter flash mode.
3. Choose Install Voice Assistant, flash the precompiled image, then provision the device on the documented 2.4 GHz Wi-Fi network.
4. Select Add to Home Assistant, approve the discovered ESPHome device, and assign the BOX-3 to the intended room.
5. Choose on-device or Home Assistant wake-word processing, then verify a non-critical command, spoken response, display states, mute control and a timer.

Original evidence: Home Assistant and ESPHome ↗ · checked 2026-08-08