drc#

diffsptk.DRC#: alias of DynamicRangeCompression

class diffsptk.DynamicRangeCompression(threshold: float, ratio: float, attack_time: float, release_time: float, sample_rate: int, makeup_gain: float = 0, abs_max: float = 1, learnable: bool = False, device: device | None = None, dtype: dtype | None = None)[source]#

See this page for details.

Parameters:

thresholdfloat <= 0: The threshold in dB.
ratiofloat > 1: The input/output ratio.
attack_timefloat > 0: The attack time in msec.
release_timefloat > 0: The release time in msec.
sample_rateint >= 1: The sample rate in Hz.
makeup_gainfloat >= 0: The make-up gain in dB.
abs_maxfloat > 0: The absolute maximum value of input.
learnablebool: Whether to make the DRC parameters learnable.
devicetorch.device or None: The device of this module.
dtypetorch.dtype or None: The data type of this module.

References

[1]

C.-Y. Yu et al., “Differentiable all-pole filters for time-varying audio systems,” Proceedings of DAFx, pp. 345-352, 2024.

forward(x: Tensor) → Tensor[source]#

Perform dynamic range compression.

Parameters:

xTensor [shape=(…, T)]: The input waveform.

Returns:

outTensor [shape=(…, T)]: The compressed waveform.

Examples

>>> x = torch.randn(16000)
>>> x.abs().max()
tensor(4.2224)
>>> drc = diffsptk.DynamicRangeCompression(-20, 4, 10, 100, 16000)
>>> y = drc(x)
>>> y.abs().max()
tensor(2.5779)

diffsptk.functional.drc(x: Tensor, threshold: float, ratio: float, attack_time: float, release_time: float, sample_rate: int, makeup_gain: float = 0, abs_max: float = 1) → Tensor[source]#

Perform dynamic range compression.

Parameters:

xTensor [shape=(…, T)]: The input waveform.
thresholdfloat <= 0: The threshold in dB.
ratiofloat > 1: The input/output ratio.
attack_timefloat > 0: The attack time in msec.
release_timefloat > 0: The release time in msec.
sample_rateint >= 1: The sample rate in Hz.
makeup_gainfloat >= 0: The make-up gain in dB.
abs_maxfloat > 0: The absolute maximum value of input.

Returns:

outTensor [shape=(…, T)]: The compressed waveform.