Class DenoiserConfig (2.33.0)

DenoiserConfig(mapping=None, *, ignore_unknown_fields=False, **kwargs)

Denoiser config. May not be supported for all models and may have no effect.

Attributes

Name Description
denoise_audio bool
Denoise audio before sending to the transcription model.
snr_threshold float
Signal-to-Noise Ratio (SNR) threshold for the denoiser. Here SNR means the loudness of the speech signal. Audio with an SNR below this threshold, meaning the speech is too quiet, will be prevented from being sent to the transcription model. If snr_threshold=0, no filtering will be applied.