Skip to content
This repository has been archived by the owner on Jul 20, 2019. It is now read-only.


Repository files navigation

Watson Speech Android SDK

The Watson Speech SDK for the Android platform enables an easy and lightweight interaction with the IBM's Watson Speech-To-Text (STT) and Text-To-Speech (TTS) services in Bluemix. The SDK includes support for recording and streaming audio in real time to the STT service while receiving a transcript of the audio as you speak. This project includes an example application that showcases the interaction with both the STT and TTS Watson services in the cloud.

The current version of the SDK uses a minSdkVersion of 9, while the example application uses a minSdkVersion of 16.

Table of Contents


Using the library

  1. Download the speech-android-wrapper.aar
  2. Once unzipped drag the speech-android-wrapper.aar file into your Android Studio project view under the libs folder.
  3. Go to build.gradle file of your app, then set the dependencies as below:
    dependencies {
        compile fileTree(dir: 'libs', include: ['*.jar'])
        compile (name:'speech-android-wrapper',ext:'aar')
        compile ''
            dirs 'libs'
  1. Clean and run the Android Studio project

Getting credentials

  1. Create an account on Bluemix if you have not already.
  2. Follow instructions at Service credentials for Watson services to get service credentials.

A Quick Start Guide

To get started, you can also take a look at a quick start guide created by @KeyOnTech.

Speech To Text

Implement the ISpeechDelegate and SpeechRecorderDelegate in the MainActivity

These delegates implement the callbacks when a response from the server is received or when the recorder is sending back the audio data. SpeechRecorderDelegate is optional.

   public class MainActivity extends Activity implements ISpeechDelegate{}

Or with SpeechRecorderDelegate

   public class MainActivity extends Activity implements ISpeechDelegate, SpeechRecorderDelegate{}

Instantiate the SpeechToText instance

   SpeechToText.sharedInstance().initWithContext(new URI("wss://"), this.getApplicationContext(), new SpeechConfiguration());

Enabling audio compression

By default audio sent to the server is uncompressed PCM encoded data, compressed audio using the Opus codec can be enabled.

   SpeechToText.sharedInstance().initWithContext(this.getHost(STT_URL), this.getApplicationContext(), new SpeechConfiguration(SpeechConfiguration.AUDIO_FORMAT_OGGOPUS));

Or this way:

    // Configuration
    SpeechConfiguration sConfig = new SpeechConfiguration(SpeechConfiguration.AUDIO_FORMAT_OGGOPUS);
    // STT
    SpeechToText.sharedInstance().initWithContext(this.getHost(STT_URL), this.getApplicationContext(), sConfig);

Set the Credentials and the delegate


Alternatively pass a token factory object to be used by the SDK to retrieve authentication tokens to authenticate against the STT service

   SpeechToText.sharedInstance().setTokenProvider(new MyTokenProvider(this.strSTTTokenFactoryURL));

Get a list of models supported by the service

   JSONObject models = getModels();

Get details of a particular model

   JSONObject model = getModelInfo("en-US_BroadbandModel");

Pick the model to be used


Start Audio Transcription


If you implemented SpeechRecorderDelegate, and needs to process the audio data which is recorded, you can use set the delegate.


Delegate methods to receive messages from the sdk

    public void onOpen() {
        // the  connection to the STT service is successfully opened

    public void onError(String error) {
    	// error interacting with the STT service

    public void onClose(int code, String reason, boolean remote) {
        // the connection with the STT service was just closed

    public void onMessage(String message) {
        // a message comes from the STT service with recognition results

End Audio Transcription


Receive speech power levels during the recognize

The amplitude is calculated from the audio data buffer, and the volume (in dB) is calculated based on it.

    public void onAmplitude(double amplitude, double volume) {
        // your code here

Text To Speech

Instantiate the TextToSpeech instance


Set the Credentials


Alternatively pass a token factory object to be used by the SDK to retrieve authentication tokens to authenticate against the TTS service

   TextToSpeech.sharedInstance().setTokenProvider(new MyTokenProvider(this.strTTSTokenFactoryURL));

Get a list of voices supported by the service


Pick the voice to be used


Generate and play audio


Common issues