Skip to content

Latest commit

 

History

History
93 lines (61 loc) · 3.62 KB

DOCS.md

File metadata and controls

93 lines (61 loc) · 3.62 KB
Raw

Home Assistant Add-on: snowboy

Installation

Follow these steps to get the add-on installed on your system:

  1. Navigate in your Home Assistant frontend to Settings -> Add-ons -> Add-on store.
  2. Add the store https://github.com/rhasspy/hassio-addons
  3. Find the "snowboy" add-on and click it.
  4. Click on the "INSTALL" button.

How to use

After this add-on is installed and running, it will be automatically discovered by the Wyoming integration in Home Assistant. To finish the setup, click the following my button:

Open your Home Assistant instance and start setting up a new integration.

Alternatively, you can install the Wyoming integration manually, see the Wyoming integration documentation for more information.

Configuration

Option: sensitivity

Activation threshold (0-1), where higher means fewer activations.

Option: debug_logging

Enable debug logging. Useful for seeing satellite connections and each wake word detection in the logs.

Custom Wake Words

This add-on will train custom wake words on start-up from WAV audio samples placed in /share/snowboy/train/<language>/<wake_word>

To get started, first record 3 samples of your wake word:

arecord -r 16000 -c 1 -f S16_LE -t wav -d 3 sample1.wav
arecord -r 16000 -c 1 -f S16_LE -t wav -d 3 sample2.wav
arecord -r 16000 -c 1 -f S16_LE -t wav -d 3 sample3.wav

Ideally, this should be recorded on the same device you plan to use for wake word recognition (same microphone, etc).

After your 3 samples are recorded, you will need to copy them to your Home Assistant server. You can use the Samba add-on to do this.

Copy the WAV files to /share/snowboy/train/<language>/<wake_word> where <language> is either en for English or zh for Chinese (other languages are not supported). <wake_word> should be the name of your wake word, such as hey_computer (spaces in the same are not recommended).

Your directory structure should look like this after copying the samples:

  • /share/snowboy/train/
    • en/
      • hey_computer/
        • sample1.wav
        • sample2.wav
        • sample3.wav

Restart the add-on and check the log for a message that your wake word was trained. Enable debug logging in the add-on configuration for more information.

After training, your wake word model (.pmdl) will be next to your samples:

  • /share/snowboy/train/
    • en/
      • hey_computer/
        • hey_computer.pmdl
        • sample1.wav
        • sample2.wav
        • sample3.wav

Copy your wake word model (e.g., hey_computer.pmdl) to /share/snowboy to start using it immediately.

If you'd like to retrain, delete the .pmdl file next to your samples and restart the add-on. You will need to copy the new model to /share/snowboy again after training.

Support

Got questions?

You have several options to get them answered:

In case you've found an bug, please open an issue on our GitHub.