When I started this project, the question I kept hearing was: "Okay, but what would I actually use it for?" Fair question. A drop-in board that replaces Google’s internals sounds cool in theory, but the real value is in what you can do once your Home Mini is running ESPHome and talking to Home Assistant locally, with no cloud in the middle. Here are some of the things that become possible.

Room-by-Room Voice Control Without Subscriptions

Say "turn off the lights" in the kitchen, and only the kitchen lights turn off. No wake word sent to Google’s servers, no account required, no internet dependency. The device processes the wake word locally on the ESP32-S3, and Home Assistant handles the rest, entirely on your own network.

Local Text-to-Speech

Home Assistant can push spoken notifications back through the speaker, appointment reminders, doorbell alerts, washing machine done, using a local TTS engine. No cloud TTS API calls, no latency spikes when your internet is slow.

Shopping Lists, Timers, and Custom Commands

You can add anything you’d wire up in Home Assistant via voice. Add items to a todo list, set a timer, trigger a script, or ask about the weather from your own local weather station. The command set is whatever you define, not whatever Google decided to support this month.

Automations Triggered by Voice, or Triggering Voice Notifications

Say "Hey Jarvis, goodnight" and your lights dim, the heating drops, the alarm arms. Or the other way: when the front door opens at 11pm, the bedroom device can say "front door opened." All bidirectional, local, and cloud-free.

Privacy by Architecture, not by Policy

The original Google Home Mini sends audio snippets to Google’s servers for processing; that’s how it works, by design. The MiciMike board changes the architecture entirely: wake word detection runs on-device, and nothing leaves your network unless you explicitly configure it to. You’re not trusting a privacy policy. You’re trusting physics.

One Device, Multiple Home Assistant Roles

The board appears in Home Assistant simultaneously as a voice assistant and a media player, the same way a native Voice PE device does. So you can use it for voice commands and push audio to it from HA automations, all from the same entity.

Keeping your data local and out of the cloud is beneficial for a variety of reasons, and we can’t wait to see what everyone creates with their MiciMike boards. If you haven’t already, please support our open source project and help us bring this to life.