Recognize audio speech and turn it into text

The SpeechIn widget performs speech recognition and outputs text. It supports multiple languages. Currently only works in the NTK app, and in Google Chrome in the Command Line version of NTK.


1 Connect

Connect the trigger input to the first inlet. When the value from this input goes above the threshold, voice recognition will start. When the input value goes below the threshold, recognition will stop.

2 Configure the widget

In the more section, set the following:

  • language – sets the language to be recognized
  • threshold – sets the level the input must reach for speech recognition to start
  • continuous
    • unchecked - recognition will end after a period of silence, even if the input level is above the threshold.
    • checked – after a period of silence the recognized text will be output, and then a new speech phrase will be recognized. This can go on for a period, but if there is a long silence, recognition will end.
  • text box – this is the recognized text. You can manually edit this and that text will be output for testing purposes


3 Speak

Set the level of the input above the threshold and start speaking

  • The header of the widget will have red text while recognition is available
  • Yellow text in the widget body is the interim recognition and will be replaced by the final recognition
  • Green text indicates the final recognition – at the point the text turns green, it is output from the widget and is usable in the next widget

4 Example

This example patch shows how to use the SpeechIn widget to generate two different outcomes, depending on the words said by the user. It uses the IfThen and Text widgets.

Download: speechIn.ntk


NETLabTK - Tangible IoT