npm version

February 2019

The version 1.3 of the Cognitive Services Speech SDK is available.
The open-source Javascript version can be found here. For other languages and platform check out the Speech SDK home page.

NOTE: This repository is deprecated. Please use the new Cognitive Services Speech SDK!

September 2018: New Microsoft Cognitive Services Speech SDK available

We released a new Speech SDK supporting the new Unified Speech Service. The new Speech SDK comes with support for Windows, Android, Linux, Javascript and iOS.

Please check out Microsoft Cognitive Services Speech SDK for documentation, links to the download pages, and the samples.

NOTE: The content of this repository is supporting the Bing Speech Service, not the new Speech Service. Bing Speech Service has been deprecated, please use the new Speech Service.


Subscribe to the Speech Recognition API, and get a free trial subscription key

The Speech API is part of Cognitive Services. You can get free trial subscription keys from the Cognitive Services subscription page. After you select the Speech API, select Get API Key to get the key. It returns a primary and secondary key. Both keys are tied to the same quota, so you can use either key.

Note: Before you can use Speech client libraries, you must have a subscription key.

Get started

In this section we will walk you through the necessary steps to load a sample HTML page. The sample is located in our github repository. You can open the sample directly from the repository, or open the sample from a local copy of the repository.

Note: Some browsers block microphone access on un-secure origin. So, it is recommended to host the 'sample'/'your app' on https to get it working on all supported browsers.

Open the sample directly

Acquire a subscription key as described above. Then open the link to the sample. This will load the page into your default browser (Rendered using htmlPreview).

Open the sample from a local copy

To try the sample locally, clone this repository:

git clone

compile the TypeScript sources and bundle/browserfy them into a single JavaScript file (npm needs to be installed on your machine). Change into the root of the cloned repository and run the commands:

cd SpeechToText-WebSockets-Javascript && npm run bundle

Open samples\browser\Sample.html in your favorite browser.

Next steps

Installation of npm package

An npm package of the Microsoft Speech Javascript Websocket SDK is available. To install the npm package run

npm install microsoft-speech-browser-sdk

As a Node module

If you're building a node app and want to use the Speech SDK, all you need to do is add the following import statement:

import * as SDK from 'microsoft-speech-browser-sdk';

and setup the recognizer:

function RecognizerSetup(SDK, recognitionMode, language, format, subscriptionKey) {
    let recognizerConfig = new SDK.RecognizerConfig(
        new SDK.SpeechConfig(
            new SDK.Context(
                new SDK.OS(navigator.userAgent, "Browser", null),
                new SDK.Device("SpeechSample", "SpeechSample", "1.0.00000"))),
        recognitionMode, // SDK.RecognitionMode.Interactive  (Options - Interactive/Conversation/Dictation)
        language, // Supported languages are specific to each recognition mode Refer to docs.
        format); // SDK.SpeechResultFormat.Simple (Options - Simple/Detailed)

    // Alternatively use SDK.CognitiveTokenAuthentication(fetchCallback, fetchOnExpiryCallback) for token auth
    let authentication = new SDK.CognitiveSubscriptionKeyAuthentication(subscriptionKey);

    return SDK.Recognizer.Create(recognizerConfig, authentication);

function RecognizerStart(SDK, recognizer) {
    recognizer.Recognize((event) => {
            Alternative syntax for typescript devs.
            if (event instanceof SDK.RecognitionTriggeredEvent)
        switch (event.Name) {
            case "RecognitionTriggeredEvent" :
            case "ListeningStartedEvent" :
            case "RecognitionStartedEvent" :
            case "SpeechStartDetectedEvent" :
                console.log(JSON.stringify(event.Result)); // check console for other information in result
            case "SpeechHypothesisEvent" :
                console.log(JSON.stringify(event.Result)); // check console for other information in result
            case "SpeechFragmentEvent" :
                console.log(JSON.stringify(event.Result)); // check console for other information in result
            case "SpeechEndDetectedEvent" :
                console.log(JSON.stringify(event.Result)); // check console for other information in result
            case "SpeechSimplePhraseEvent" :
                UpdateRecognizedPhrase(JSON.stringify(event.Result, null, 3));
            case "SpeechDetailedPhraseEvent" :
                UpdateRecognizedPhrase(JSON.stringify(event.Result, null, 3));
            case "RecognitionEndedEvent" :
                console.log(JSON.stringify(event)); // Debug information
    .On(() => {
        // The request succeeded. Nothing to do here.
    (error) => {

function RecognizerStop(SDK, recognizer) {
    // recognizer.AudioSource.Detach(audioNodeId) can be also used here. (audioNodeId is part of ListeningStartedEvent)

In a Browser, using Webpack

Currently, the TypeScript code in this SDK is compiled using the default module system (CommonJS), which means that the compilation produces a number of distinct JS source files. To make the SDK usable in a browser, it first needs to be "browserified" (all the javascript sources need to be glued together). Towards this end, this is what you need to do:

  1. Add require statement to you web app source file, for instance (take a look at sample_app.js):

        var SDK = require('<path_to_speech_SDK>/Speech.Browser.Sdk.js');
  2. Setup the recognizer, same as above.

  3. Run your web-app through the webpack (see "bundle" task in gulpfile.js, to execute it, run npm run bundle).

  4. Add the generated bundle to your html page:

    <script src="../../distrib/speech.sdk.bundle.js"></script>

In a Browser, as a native ES6 module progress, will be available soon

Token-based authentication

To use token-based authentication, please launch a local node server, as described here


The SDK is a reference implementation for the speech websocket protocol. Check the API reference and Websocket protocol reference for more details.

Browser support

The SDK depends on WebRTC APIs to get access to the microphone and read the audio stream. Most of todays browsers(Edge/Chrome/Firefox) support this. For more details about supported browsers refer to navigator.getUserMedia#BrowserCompatibility

Note: The SDK currently depends on navigator.getUserMedia API. However this API is in process of being dropped as browsers are moving towards newer MediaDevices.getUserMedia instead. The SDK will add support to the newer API soon.


This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact with any additional questions or comments.