JavaScript Speech SDK – API Reference

Comprehensive reference for the Azure Cognitive Services Speech SDK for JavaScript.

Installation

Install via npm or include directly from a CDN.

npm install microsoft-cognitiveservices-speech-sdk

Or use the CDN:

<script src="https://aka.ms/csspeech/jsbrowserpackageraw"></script>

Getting Started

Recognize speech from the microphone and synthesize text to speech.

const SpeechSDK = window.SpeechSDK || require('microsoft-cognitiveservices-speech-sdk');

const speechConfig = SpeechSDK.SpeechConfig.fromSubscription("YOUR_SUBSCRIPTION_KEY", "YOUR_REGION");
speechConfig.speechRecognitionLanguage = "en-US";

const audioConfig = SpeechSDK.AudioConfig.fromDefaultMicrophoneInput();
const recognizer = new SpeechSDK.SpeechRecognizer(speechConfig, audioConfig);

recognizer.recognizeOnceAsync(result => {
    console.log(`RECOGNIZED: Text=${result.text}`);
    recognizer.close();
});
Demo unavailable in static preview.

Core Classes

Configuration for speech recognition and synthesis.

Method / PropertyDescription
static fromSubscription(key, region)Creates a config from an Azure subscription.
speechRecognitionLanguageLocale for recognition, e.g., "en-US".
speechSynthesisVoiceNameName of the voice used for synthesis.
outputFormatAudio output format (e.g., SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm).

Defines audio input/output sources.

MethodDescription
static fromDefaultMicrophoneInput()Uses the system microphone.
static fromDefaultSpeakerOutput()Uses the default speaker for synthesis.
static fromWavFileInput(file)Audio from a WAV file (browser only).

Performs speech‑to‑text.

MethodDescription
recognizeOnceAsync(callback, errorCallback)Recognizes a single utterance.
startContinuousRecognitionAsync()Begins continuous recognition.
stopContinuousRecognitionAsync()Stops continuous recognition.
close()Releases resources.

Events:

  • recognizing – interim results.
  • recognized – final result.
  • canceled – errors or cancellations.

Synthesizes text to speech.

MethodDescription
speakTextAsync(text, callback, errorCallback)Synthesizes plain text.
speakSsmlAsync(ssml, callback, errorCallback)Synthesizes SSML.
close()Releases resources.

Events:

  • speakingStarted
  • speakingCompleted
  • canceled

Error Handling

All async methods accept an errorCallback. The error object contains errorDetails and errorCode.

recognizer.recognizeOnceAsync(
    result => console.log(result.text),
    err => console.error(`Error ${err.errorCode}: ${err.errorDetails}`)
);

Additional Resources