IF YOU HAVE NOT CONFIGURED SOUND DO THAT PRIOR TO CONFIGURING SPEECH
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:
- In the Cloud Platform Console, go to the Projects page and select or create a new project.
GO TO THE PROJECTS PAGE
- Enable billing for your project.
- 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)
- Create a new JSON service account key and download it to the
Credentials Guide: On Your Own Server.
When prompted to create a new service account select "Project Owner"
Keep this information for later. You'll need your projectID and keyfile to configure the smart mirror.
Train your own Keyword
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:
npm run train-model
This command launches the snowboy site in embedded chromium.
Note: there is currently a snowboy bug that requires you to click the logo in the top left, login, close the window, and re-run the command
Or, you can train your model here: https://snowboy.kitt.ai/hotword/47
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