Cloud Speech Recognition

If you have trouble setting up speech recognition try heading over to the troubleshooting section. If that doesn't work drop into the discord chat.

The smart mirror uses [Sonus](https://github.com/evancohen/sonus) for speech recognition. The following steps will get you everything you need to make this work:

Setting up Speech Recognition

The smart mirror uses Sonus with Google Cloud Speech for keyword spotting and recognition. To set that up, you'll need to create a new project in the Cloud Platform Console:

  1. In the Cloud Platform Console, go to the Projects page and select or create a new project. GO TO THE PROJECTS PAGE

  2. Enable billing for your project. ENABLE BILLING

  3. Enable the Cloud Speech API. ENABLE THE API - For more info see Cloud Speech API Pricing (for normal smart mirror usage it will be free)

  4. Create a new JSON service account key and download it to the smart-mirror folder... Credentials Guide: On Your Own Server.

When prompted to create a new service account select "Project Owner"

New service account

Keep this information for later. You'll need your projectID and keyfile to configure the smart mirror.

Setting up your API key

After setting up your project you will also need an API key for

  • geolocation

  • geocoding

  • Javascript Static Map

  • youtube (if you intend to use this function on your mirror)

google apis.

from the 3 bar menu on the left, select APIs and Services, then Library

enter part of the name of each api in turn, and select it, and click enable repeat for the other apis listed after completing the enabling for all the listed apis

  • from the 3 bar menu on the left, select APIs and Services, then Credentials

  • in the Create credentials dropdown, select "API key",

  • save the value that is shown on the message window that pops up

The API key should be a long string, something like this 'AIzaSyBGWRB2oC1P_xYteMMsdTRV3Kv3aEBPQg'

[Optional] Train your own Keyword

Note: This is no longer required, the mirror comes with a pre-trained universal model! If you're having trouble with the universal model you can follow these steps

Training your own keyword will drastically increase the accuracy of keyword detection. This will be most accurate if you do this on your Pi using the microphone that you'll be using to trigger the mirror.

From the smart mirror directory run:

as of Mid 2019 local model training no longer works see

you can train your model here: [https://github.com/seasalt-ai/snowboy]https://github.com/seasalt-ai/snowboy)

Once trained, download the model and save it to the root of the smart-mirror directory.

Next Step: Run The Mirror For the First Time

PreviousConfiguring SoundNextFirst Time Running Smart Mirror

Last updated

Was this helpful?