Voice enabled SmartMirror² - snips inside


I recently stumbled upon a nice Mini MagicMirror²

It is built around a Pi Zero but the recently released Pi3 A+ is a better fit.

Printing the enclosure is straightforward on a Prusa i3 MK3. There is something missing. Snips !

I had a Google AIY Voice hat (gen 1) laying around which has been used for a Snips satellite.

The voice hat fits nicely onto the Pi3 A+. It also has connectors for 5V and GND which is used for powering the screen

The HDMI cable is built from a flat ribbon cable and two angled connectors.

The mic board of the voice hat is attached to the top. The biggest challenge was to find a speaker that would fit into the enclosure. Finding a 4Ohm speaker small enough to fit in it not easy. The smaller ones are usually below 1 Watt.

I didn’t know about a special type of speaker called exciter.

The Dayton Audio DAEX19CT-4 Coin Type 19mm Vented Exciter 5W 4 Ohm is a nice fit.

Last piece is a suitable two-way mirror. Forget about foil. The guy in the video bought pre-cut plastic/acrylic mirror but I did not find a good source. You can supposedly buy custom made ones here.

From a recent MagicMirror project I learned to go for glass mirrors.

There is company in Germany GlasStar that makes these according to the dimensions you need them in. I ordered 3 mirrors for a price of 89€ (inluces VAT and shipping). They sent me 4 mirrors. Same quality like the bigger one I ordered recently.

Worked out perfectly.

Why would you need snips in a magic mirror ? Have a look at this video

It simply makes sense to control your mirror with your voice rather then a touchscreen.

What else is it used for ? It is used to check on the hot water temperature (eBUS-Vaillant) and the heating system.

Another idea would be to leave out the speaker and print the text from snips onto the screen. That way the MagicMirror could be used as snips satellite without the need for a speaker.

Join the Snips universe and benefit from a privacy by design voice assistant.

Snips GUI or visual display of assistant

Really nice :slight_smile:

You provide lots of information about the hardware - but what about software? Are you using snips apps for sending information to the screen? Or is there a specific display software written by you that listens to the detected intents directly?

I’m asking because I think a display is a must for my assistant - but I’m still not sure about the concept. (My TV-App will send a MQTT message containing a HTML page, that can be displayed in any browser running on the raspi for example)


The software is the MagicMirror² project. I followed these instruction for installing on a Pi Zero W. I later changed this to a RPi 3 A+ (without the USB and Ethernet connectors)

The MagicMirror² project has good documentation. I have yet to integrate the display software-wise but don’t see huge problems since MM has modules for that





Awesome project! Congrats

What TTS provider are you using?


wohoo someone linked my magic mirror + snips video :smiley:
and @valf the video actually uses the standard TTS
Since I was asked for the source code (still needs some work and I did not have to much time lately) I just published it aswell

The mqtt bridge delivers messages from mqtt topics to mm2 internal notification system (and the other way around). That way it can be used for multiple mm2 modules and not just for the hide/show/move module. (I will probably have time to finish a few things next month, since the TLS stuff is not working)


I can report that it now (13 Mar 2019) works after extensive work by the author. When you set it up, triple check the MQTT configuration (topics etc.)

The browser console will give you feedback.


Seems like the Google AIY voice kits are no longer available. However, there seems to be a clone (original?) which should work.

To me it looks identical.