This sample application was created using the express application generator, is heavily based on the Speech to Text Browser Application code and so can be set up and used by following the instructions on that page.
The aim here is to provide a very cut down set of code that can be used to add voice input for search style applications in the web browser. As such, it simply:
-
controls the microphone on/off state;
-
sends audio to the Speech to Text service over WebSockets;
-
receives transcribed text;
-
automatically detects when the speaker has stopped speaking;
-
times out after 5 seconds of inactivity if no speech is heard;
-
provides a sample web interface that redirects your web browser to an search of ibm.com;
-
provides a NodeJS interface for getting a BlueMix speech to text token;
-
demonstrates how to get a BlueMix speech to text token from the local server.
There are two reusable components that you can use to easily integrate search style Speech to Text into your application. These are:
- Microphone.js
This is more or less a carbon copy of the Microphone.js object from the [Speech to Text Browser Application][speech-to-text-nodejs]. It's an HTML 5 web audio interface for accessing your microphone using a web browser.
All you need to do here is create an instance of a Microphone object that you'll pass in when you create a SpeechToText object.
e.g. `var mic = new Microphone();`
- SpeechToText.js
A very simple and massively cut down implementation of the web sockets code from the [Speech to Text Browser Application][speech-to-text-nodejs].
Create an instance of a SpeechToText object and pass it:
* a token for your BlueMix speech to text service
* the Microphone instance you created
* a callback for when recording has started (useful for clearing any previous search and changing your UI)
* a callback for transcription events (this will periodically receive a String containing the latest transcription)
* a callback for when recording has stopped (useful for changing your UI)
e.g. `var s2t = new SpeechToText(token,mic,onStarted,onTranscript,onStopped);`
*Note:* The default voice model is set to US English (broadband). No interface has been provided here to allow a user to change this. Adding that feature should be trivial or you could automatically detect the region of your user and set the appropriate model automatically.
The code in speechsearch.js demonstrates how the above objects can be used. The HTML found in index.html shows how they can all be tied together into a simple browser interface.
Licensed under Apache 2.0. Full license text is available in LICENSE.