speech-to-text

readme

install SoX
create /data folder
create /config/default.json
npm install

index.js

var config = require('config');

// thanks to: https://gist.github.com/lukehoban/0ee5c1bef438dc5bd7cb
var fs = require('fs');
var util = require('util');
var request = require('request');

var clientId = 'test-app'; // Can be anything
var clientSecret = config.get('key'); // API key from Azure marketplace

var dirData = 'data/'; // set directory where thing will be saved and read
var locAudio = 'audio.wav'; // set where the audio file will be
var locTxt = 'text.txt'; // set where the test will be saved


getAccessToken(clientId, clientSecret, function(err, accessToken) {
  if(err) return console.log(err);
  console.log('Got access token: ' + accessToken);
  
    speechToText(dirData + locAudio, accessToken, function(err, res) {
      if(err) return console.log(err);

      var resConfidence = res.results[0].confidence; // result of confidence
      var resLexical = res.results[0].lexical; // result of written text from audio
      console.log('Confidence ' + resConfidence + ' for: "' + resLexical + '"');
      fs.writeFile(dirData + locTxt, resLexical, function(err) {
      	if(err) return console.log(err);

      	console.log('Written text into file');
      });
    });
});

// getAccessToken(clientId, clientSecret, function(err, accessToken) {
//   if(err) return console.log(err);
//   console.log('Got access token: ' + accessToken);
  
//   textToSpeech(str, dirData + locAudio, accessToken, function(err) {
//     if(err) return console.log(err);
//     console.log('Wrote out: ' + dirData + locAudio);
    
//     speechToText(dirData + locAudio, accessToken, function(err, res) {
//       if(err) return console.log(err);
//       console.log('Confidence ' + res.results[0].confidence + ' for: "' + res.results[0].lexical + '"');
//     });
//   });
// });

// ==== Helpers ====

function getAccessToken(clientId, clientSecret, callback) {
  request.post({
    url: 'https://oxford-speech.cloudapp.net/token/issueToken',
    form: {
      'grant_type': 'client_credentials',
      'client_id': encodeURIComponent(clientId),
      'client_secret': encodeURIComponent(clientSecret),
      'scope': 'https://speech.platform.bing.com'
    }
  }, function(err, resp, body) {
    if(err) return callback(err);
    try {
      var accessToken = JSON.parse(body).access_token;
      if(accessToken) {
        callback(null, accessToken);
      } else {
        callback(body);
      }
    } catch(e) {
      callback(e);
    }
  });
}

function textToSpeech(text, filename, accessToken, callback) {
  var ssmlTemplate = "<speak version='1.0' xml:lang='en-us'><voice xml:lang='%s' xml:gender='%s' name='%s'>%s</voice></speak>";
  request.post({
    url: 'http://speech.platform.bing.com/synthesize',
    body: util.format(ssmlTemplate, 'en-US', 'Female', 'Microsoft Server Speech Text to Speech Voice (en-US, ZiraRUS)', text),
    encoding: null,
    headers: {
      'Authorization': 'Bearer ' + accessToken,
      'Content-Type' : 'application/ssml+xml',
      'X-Microsoft-OutputFormat' : 'riff-16khz-16bit-mono-pcm',
      'X-Search-AppId': '07D3234E49CE426DAA29772419F436CA',
      'X-Search-ClientID': '1ECFAE91408841A480F00935DC390960',
    }
  }, function(err, resp, body) {
    if(err) return callback(err);
    fs.writeFile(filename, body, 'binary', function (err) {
      if (err) return callback(err);
      callback(null);
    });
  });
}

function speechToText(filename, accessToken, callback) {
  fs.readFile(filename, function(err, waveData) {
    if(err) return callback(err);
    request.post({
      url: 'https://speech.platform.bing.com/recognize/query',
      qs: {
        'scenarios': 'ulm',
        'appid': 'D4D52672-91D7-4C74-8AD8-42B1D98141A5', // This magic value is required
        'locale': 'en-US',
        'device.os': 'wp7',
        'version': '3.0',
        'format': 'json',
        'requestid': '1d4b6030-9099-11e0-91e4-0800200c9a66', // can be anything
        'instanceid': '1d4b6030-9099-11e0-91e4-0800200c9a66' // can be anything
      },
      body: waveData,
      headers: {
        'Authorization': 'Bearer ' + accessToken,
        'Content-Type': 'audio/wav; samplerate=16000',
        'Content-Length' : waveData.length
      }
    }, function(err, resp, body) {
      if(err) return callback(err);
      try {
        callback(null, JSON.parse(body));
      } catch(e) {
        callback(e);
      }
    });
  });
}

mic.js

var rec = require('node-record-lpcm16');
var fs = require('fs');

var file = fs.createWriteStream('./data/audio.wav', { encoding: 'binary' });

rec.start({
  sampleRate: 44100,
  verbose: true
})
.pipe(file);

setTimeout(function () {
  rec.stop();
}, 10000); // stop after 10 seconds

package.json

{
  "name": "speech-to-text",
  "version": "1.0.0",
  "description": "",
  "main": "index.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  },
  "author": "",
  "license": "ISC",
  "dependencies": {
    "config": "^1.19.0",
    "node-record-lpcm16": "^0.1.4",
    "request": "^2.67.0",
    "util": "^0.10.3"
  }
}