arguments_classes/vad_arguments.py

from dataclasses import dataclass, field


@dataclass
class VADHandlerArguments:
    thresh: float = field(
        default=0.3,
        metadata={
            "help": "The threshold value for voice activity detection (VAD). Values typically range from 0 to 1, with higher values requiring higher confidence in speech detection."
        },
    )
    sample_rate: int = field(
        default=16000,
        metadata={
            "help": "The sample rate of the audio in Hertz. Default is 16000 Hz, which is a common setting for voice audio."
        },
    )
    min_silence_ms: int = field(
        default=250,
        metadata={
            "help": "Minimum length of silence intervals to be used for segmenting speech. Measured in milliseconds. Default is 250 ms."
        },
    )
    min_speech_ms: int = field(
        default=500,
        metadata={
            "help": "Minimum length of speech segments to be considered valid speech. Measured in milliseconds. Default is 500 ms."
        },
    )
    max_speech_ms: float = field(
        default=float("inf"),
        metadata={
            "help": "Maximum length of continuous speech before forcing a split. Default is infinite, allowing for uninterrupted speech segments."
        },
    )
    speech_pad_ms: int = field(
        default=500,
        metadata={
            "help": "Amount of padding added to the beginning and end of detected speech segments. Measured in milliseconds. Default is 250 ms."
        },
    )
    audio_enhancement: bool = field(
        default=False,
        metadata={
            "help": "improves sound quality by applying techniques like noise reduction, equalization, and echo cancellation. Default is False."
        },
    )