The AssemblyAI wrapper provides an easy to use interface to stream audio from the browser to the AssemblyAI API for near real-time speech-to-text. For a complete sample project, check out the AssemblyAI Sample WebRTC Project.

1. Clone this repo

2. Copy or link the cloned repo to your web project's public script folder Example: /js folder

3. Include the AssemblyAI javascript file in your HTML document

   <script src="public-script-folder/assembly-ai.js"></script>
   

var assemblyai = new AssemblyAI("your-secret-api-token");

Create an AssemblyAI instance with your API token.

Start recording

assemblyai.startRecording([autoStop, transcriptionCallback, uiCallback])

Begin recording audio stream from browser.

Optional parameters:

• autoStop: true | false Auto-detect when user is done speaking and stops recording. Requires transcriptionCallback parameter if true.

• transcriptionCallback: Function to handle result of transcription operation. For example:

  assemblyai.startRecording(true, function(response){
      console.log(response.text);
      console.log(response.confidence);
      console.log(response.id)
  });
  

  Required if autoStop is true.

• uiCallback: Function to handle any UI operations immediately after recording stops:

  assemblyai.startRecording(true, transcriptionCallback, function(){
      /* Code to update UI immediately after recording stops */
  
      // Change text of record button back from "Recording" to "Record"
      var button = document.getElementById("record-button");
      button.innerText("Record");
  });
  

Cancel recording

assemblyai.cancelRecording()

Cancel audio stream being recorded.

Stop recording

assemblyai.stopRecording(callback)

Stop audio stream being recorded. Recorded audio is sent to the AssemblyAI API for transcription.

• callback: Function to handle result of transcription operation. For example:

  assemblyai.stopRecording(function(response){
      console.log(response.text);
      console.log(response.confidence);
      console.log(response.id)
  });
  

• The response argument passed to your callback function will be the following response object returned from the AssemblyAI API:

  Example response object:
  {
    status: 'completed',
    model_id: 'None',
    confidence: '0.9',
    created: '2018-06-25 05:14:28.361123',
    text: 'the weather is nice today',
    id: 5
  }
  

More details can be found in the full AssemblyAI Documentation.

Save recording

assemblyai.saveRecording()

Save the recorded audio stream to disk in audio/wav format.

For a complete sample project, check out the AssemblyAI Sample WebRTC Project.

The wrapper utilizes WebRTC to capture audio from the browser. WebRTC requires web projects to be hosted on localhost or an SSL secured domain.