Silero Voice Activity Detector Extension

About Silero VAD

The Silero Voice Activity Detector Extension makes easy to detect speech in an audio signal. You can find more information at https://github.com/snakers4/silero-vad.\*

Example

ONNX.SileroVAD

In this example, we are extending an ONNX.MLSink to create an ONNX.SileroVAD node which we can use to detect specch on an incoming signal. The speech prediction is presented as a percentage.

Switchboard Editor Preview

Switchboard JSON

{
  "title": "ONNX.SileroVAD example",
  "nodes": [
    {
      "id": "superpoweredPlayer1",
      "type": "Superpowered.AdvancedAudioPlayer",
      "uiConfig": {
        "tracks": [
          {
            "label": "Clean speech",
            "url": "https://switchboard-sdk-public.s3.amazonaws.com/assets/audio/speech/clean-voice.wav"
          },
          {
            "label": "Noisy speech",
            "url": "https://switchboard-sdk-public.s3.amazonaws.com/assets/audio/speech/Noisy-Speech.wav"
          },
          {
            "label": "Solo guitar",
            "url": "https://switchboard-sdk-public.s3.amazonaws.com/assets/audio/music/guitar-48k.wav"
          },
          {
            "label": "Music",
            "url": "https://switchboard-sdk-public.s3.amazonaws.com/assets/audio/music/elevator-music-stereo.mp3"
          }
        ]
      }
    },
    {
      "id": "onnxNode",
      "type": "ONNX.SileroVAD",
      "assets": {
        "model": "https://switchboard-sdk-public.s3.amazonaws.com/assets/models/onnx/silero_vad.with_runtime_opt.ort"
      }
    }
  ],
  "connections": [
    {
      "sourceNode": "superpoweredPlayer1",
      "sourceBusIndex": 0,
      "destinationNode": "onnxNode",
      "destinationBusIndex": 0
    },
    {
      "sourceNode": "superpoweredPlayer1",
      "sourceBusIndex": 0,
      "destinationNode": "outputNode",
      "destinationBusIndex": 0
    }
  ]
}

Node types

The Silero Voice Activity Detector Extension provides the following audio nodes for a Switchboard SDK audio graph:

Node	Description
Onnx.SileroVad	A sink node that detects speech.