pi-voice-command-google

Recognize a single voice command on a Raspberry Pi using the Google Speech API.

~~ WARNING ~~

Google demands personal information, including credit card billing information in order to use their service. Further, at time of writing, only 60 minutes of speech is free per month before you are charged. This is soooo far from a good solution, but it's the best I've been able to find for free voice recognition. To help the development of Free/Libre Open Source voice recognition, please contribute your voice to the Common Voice project!

Setup

1. Install NodeJS 8.x.x or newer.
2. Follow Step 1 and Step 2 on this Google Speech API quickstart page. Save the JSON file it gives you somewhere handy.
3. Have ALSA set up for audio. If the arecord command is not available, you might need a package like alsa-utils.
4. Run npm install pi-voice-command-google to install the module.

Usage

var voice = require('pi-voice-command-google')
var fs = require('fs')
 
var keysPath = './speech_keys.json'
 
voice(keysPath, function (err, res) {
  if (err) throw err
  console.log('Transcript:', res)
})

and saying 'hello world' aloud, outputs

hello world

API

var voice = require('pi-voice-command-google')

var cancel = voice(keysPath[, ready], cb)

Returns a function cancel that can be used to terminate microphone recording and upload to Google servers.

keysPath is a path to a JSON file: whatever Google gives you when you set up a Speech API project here.

Otherwise, cb(err, res) is called. res will be a string with the contents of what the Google machinery heard you say.

ready is optional, and is a function that will be called once the mic is set up and recording is starting.

Right now, this function is hardcoded to abort after 3 seconds of silence (no command given; null as res), or after 1 second of silence after a command is given.

License

MIT