py-webrtcvad Documentation

To use the WebRTC Voice Activity Detector, follow these steps:

1. Initialize the VAD: Create a webrtcvad.Vad() object.
2. Set Aggressiveness (Optional): Use vad.set_mode(mode) to set the filtering aggressiveness. The mode must be an integer between 0 and 3.
   • 0: Least aggressive (least likely to filter out non-speech).
   • 3: Most aggressive (most likely to filter out non-speech).
   • You can also pass the mode directly during initialization: webrtcvad.Vad(mode).
3. Process Audio Frames: Use vad.is_speech(frame, sample_rate) to classify a segment of audio.

Audio Requirements:

• Format: 16-bit mono PCM audio.
• Sample Rates: Must be exactly 8000, 16000, 32000, or 48000 Hz.
• Frame Duration: Each frame must be exactly 10, 20, or 30 ms in duration.

Install the webrtcvad module using pip:

pip install webrtcvad

Sets the aggressiveness mode for the VAD.

Parameters:

• mode: An integer between 0 and 3.
  • 0: Least aggressive.
  • 3: Most aggressive.

Note: The mode can also be set during the instantiation of the Vad object.

vad.set_mode(1)

Classifies a short segment of audio as being voiced or unvoiced.

Parameters:

• frame: A bytes object containing 16-bit mono PCM audio.
• sample_rate: The sample rate of the audio (must be 8000, 16000, 32000, or 48000).

Returns:

• bool: True if the frame contains speech, False otherwise.

vad.is_speech(frame, sample_rate)