Project Oxford Speech APIs Node.js Sample

speech.js

var fs = require('fs');
var util = require('util');
var request = require('request');

var clientId = 'test-app';                             // Can be anything
var clientSecret = 'f6f0bfec08274b8790520a9079b808af'; // API key from Azure marketplace

var str = 'This is a cool demo to call Microsoft text to speach service in Node.js.';

console.log('Converting from text -> speech -> text.');
console.log('Input text: "' + str + '"');

getAccessToken(clientId, clientSecret, function(err, accessToken) {
  if(err) return console.log(err);
  console.log('Got access token: ' + accessToken)
  
  textToSpeech(str, 'test.wav', accessToken, function(err) {
    if(err) return console.log(err);
    console.log('Wrote out: ' + 'test.wav');
    
    speechToText('test.wav', 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);
      }
    });
  });
}

Bing speech API - same error

the body of created test.wav is "

Our services aren't available right now

We're working to restore all services as soon as possible. Please check back soon.

Ref A: 63E4B5D206F145EB87505A4D9945CC6A Ref B: 8699883F3C81321F54B2BE5A6EF4FC9D Ref C: Sat Apr 16 07:30:09 2016 PST".

It seems that everyone has the same error...

Problem with the text to speech code above is, that it is missing the User-Agent header.

It works for me as shown below with the User-Agent header set to some string (only required for the textToSpeech function).

`
headers: {
'X-Microsoft-OutputFormat' : 'riff-16khz-16bit-mono-pcm',
'Authorization': 'Bearer ' + accessToken,
'Content-Type' : 'application/ssml+xml',
'User-Agent': 'TTSWithNode' // or whatever else

  //'X-Search-AppId': '07D3234E49CE426DAA29772419F436CA',
  //'X-Search-ClientID': '1ECFAE91408841A480F00935DC390960',
}

`

This STT API still working??

Also getting "Services not available right now". Any idea why?

Hi Luke. Three things:

(1) It's not a good idea to post code with your clientSecret. You might want to remove it.
(2) Have you tried Microsoft's sample JS code on github? https://github.com/Microsoft/ProjectOxford-ClientSDK/tree/master/Speech/TextToSpeech/Samples-Http/NodeJS — all that is missing is a little code to write the data to disk (which you already have)
(3) The Microsoft sample code works for me, except that it seems to produces a wav file with a lot of static in the playback. Would love to hear from any of you that have the same issue or have found a workable solution. After messing with ffmpeg and sox, trying to figure it out, I am baffled.

I receive the "Services Not Working" error if I use the http: endpoint instead of https:, I'm still not getting successful responses, but I'm not getting That particular error.

I am also getting service not available .Anyone found resolution?. same error with micorsoft's sample code as well.

I dont get the problem when I test it on Postman. But I get the problem everywhere else lol

@myadhdoutlet did you fix the issue with the static? I had the same issue and I fixed it adding encoding: null to the request. This way you get the response as a Buffer. See https://github.com/palmerabollo/bingspeech-api-client/blob/master/src/client.ts#L102

@lukehoban From where did you get 'X-Search-AppId': '07D3234E49CE426DAA29772419F436CA' and 'X-Search-ClientID': '1ECFAE91408841A480F00935DC390960'?

Hi,
The voice recognition process works fine but only if my sentence contains at least 5 or 6 syllables. Is there any limitation that prevent shortest sentences to be recognized?
Thanks

From where did you get 'X-Search-AppId': '' and 'X-Search-ClientID': ''?

@miparnisari hello,From where did you get 'X-Search-AppId': '' and 'X-Search-ClientID': ''?