mumble.js

A simple Javascript framework for adding voice commands to a web site using the web speech recognition API. Supports RegEx command syntax and the CommonJS/AMD module syntax.

Forked from and inspired by https://github.com/TalAter/annyang.

Demo

For a demo of the original library (annyang.js), see https://www.talater.com/annyang. The end-user experience is basically the same.

Usage

• If using node.js with browserify for example you can run npm install mumble-js within your app to get the latest version. Then require('mumble-js').
• If you're using a plain web app include mumble.js or mumble.min.js directly on your site.

Then define some commands and start the device:

// if using node.js, else leave out
var Mumble = require('mumble-js');
 
// for all options, see the docs
var mumble = new Mumble({
    language: 'en-US',
    debug: false, // set to true to get some detailed information about what's going on
 
    // define some commands using regex or a simple string for exact matching
    commands: [{
        name: 'appointment',
        command: /^book (.+) for me (today|tomorrow) at (\d+)$/,
 
        action: function(type, date, hour) {
            console.log('Making an appointment for %s %s at %d', type, date, hour);
        }
    }, {
        name: 'google',
        command: /^google (.+) for me\s?(please)?$/,
 
        action: function(query, polite) {
            if (polite) {
                // google the query
            } else {
                console.log('I will google that for you but only if you say please');
            }
        }
    }],
 
    // define global callbacks (see docs for all)
    callbacks: {
        start: function(event) {
            console.log('Starting..');
        }
 
        // start, end, speech, recognizeMatch, etc
    }
});
 
// add a command afterwards, or anytime
mumble.addCommand('clock', 'what is the time', function() {
    console.log(new Date());
});
 
// start listening
mumble.start();

Microphone permissions and HTTP(S)

Chrome's implementation of SpeechRecognition behaves differently based on the protocol used:

• https:// Asks for permission once and remembers the choice.
• http:// Asks for permission repeatedly on every page load. Results are also returned significantly slower in HTTP.

For a great user experience, don't compromise on anything less than HTTPS (an SSL certificate can be as cheap as $5).

While developing, you can use a self-signed SSL certificate described here: https://www.sslshopper.com/article-how-to-create-and-install-an-apache-self-signed-certificate.html.

API Documentation

Check out mumble.md in the docs/ folder for the public API methods and all available options.

Authors

• Johan Johansson: https://github.com/swemaniac
• Tal Ater: https://github.com/TalAter (the original annyang.js author)

License

Licensed under MIT.