Before anything else: thank you.

The campaign has now crossed the 1000% funding mark, and none of this would have been possible without the trust and support from all of the backers so far. What started as a personal hobby project somehow turned into something much bigger than I ever expected, and I’m genuinely grateful for that.

One of the most common misconceptions about voice assistants is that they’re "always sending everything to the cloud." With most commercial smart speakers, that’s unfortunately true.

The MiciMike Home Mini Drop-In PCB was designed specifically to change that.

Local by Default

The original Google Home Mini depends heavily on Google’s servers. Audio leaves your home, gets processed remotely, and the response comes back a few seconds later. That’s how the product was designed. With the MiciMike board, the architecture is completely different. The wake word detection runs locally on the ESP32-S3. Once activated, Home Assistant handles the voice pipeline on your own hardware. Nothing is sent anywhere unless you explicitly configure external services.

That means:

• No mandatory cloud account
• No subscription fees
• No dependency on Google's ecosystem
• No random feature removals because a company changed direction
• Your automations continue working even if the internet goes down

More Than Just Privacy

Local processing isn’t only about privacy — it also changes how the device feels to use. When everything stays on your own network:

• Responses are faster
• Automations are more reliable
• Latency is predictable
• Your smart home keeps functioning during outages

And because the board integrates directly with ESPHome and Home Assistant, you’re not locked into predefined commands.

You define what the assistant can do.

Voice Assistant and Media Player

One thing many people don’t realize is that the board appears in Home Assistant as both:

• A voice assistant device and
• A media player

So the same speaker can…

• Answer voice commands
• Play TTS notifications
• Announce doorbell events
• Act as a room audio endpoint
• Trigger automations
• Receive audio streams from Home Assistant

…all from a single device.

Why XMOS Matters

A lot of DIY ESP32 voice assistants struggle in noisy environments. That’s exactly why this project uses an XMOS XU316 audio processor alongside the ESP32-S3. The XMOS handles the microphone audio front-end, including:

• noise suppression
• echo cancellation
• improved voice pickup reliability in real-world environments

This is the same general architecture used by the Home Assistant Voice Preview Edition. The goal wasn’t just to make a Home Mini replacement. The goal was to make one that actually feels usable in a real home.

The validation batch with the updated touch controller is currently in production, and I’ll share the results as soon as the boards arrive.