grafx.processors.stereo

class StereoGain

Bases: Module

A simple stereo-to-stereo or mono-to-stereo gain.

We use simple channel-wise constant multiplication with a gain vector. The gain is assumed to be in log scale, so we apply exponentiation before multiplying it to the stereo signal.

$$y[n]=\exp (g_{\log })\cdot u[n].$$

Hence, the learnable parameter is $p=\{g_{\log }\}.$

forward(input_signals, log_gain)

Processes input audio with the processor and given parameters.

Parameters:

• input_signals (FloatTensor, $B\times C\times L$) – A batch of input audio signals, either mono or stereo.

• log_magnitude (FloatTensor, $B\times 2$) – A log-magnitude vector of the FIR filter.

Returns:

A batch of output signals of shape $B\times 2\times L$.

Return type:

FloatTensor

parameter_size()

Returns:

A dictionary that contains each parameter tensor’s shape.

Return type:

Dict[str, Tuple[int, ...]]

class SideGainImager

Bases: Module

Stereo imager with side-channel loudness control.

We multiply the input’s side signal $u_{\mathrm{s}}[n]$, i.e., left $u_{\mathrm{l}}[n]$ minus right $u_{\mathrm{r}}[n]$, with a gain parameter $g\in \mathbb{R}$ to control the stereo width. The mid and side outputs are given as

$$\begin{array}{rl}{y}_{\mathrm{m}}[n]& =u_{\mathrm{l}}[n]+u_{\mathrm{r}}[n],\\ y_{\mathrm{s}}[n]& =\exp (g)\cdot (u_{\mathrm{l}}[n]-u_{\mathrm{r}}[n]).\end{array}$$

Hence, the learnable parameter is $p=\{g_{\log }\}.$

forward(input_signals, log_gain)

Processes input audio with the processor and given parameters.

Parameters:

• input_signals (FloatTensor, $B\times C\times L$) – A batch of input audio signals; must be stereo.

• log_magnitude (FloatTensor, $B\times P$) – A log-magnitude vector of the FIR filter.

Returns:

A batch of output signals of shape $B\times 2\times L$.

Return type:

FloatTensor

parameter_size()

Returns:

A dictionary that contains each parameter tensor’s shape.

Return type:

Dict[str, Tuple[int, ...]]

class MonoToStereo

Bases: Module

A simple mono-to-stereo conversion.

forward(input_signals)

Processes input audio with the processor and given parameters.

Parameters:

input_signals (FloatTensor, $B\times 1\times L$) – A batch of input audio signals; must be mono.

Returns:

A batch of output signals of shape $B\times 2\times L$.

Return type:

FloatTensor

parameter_size()

Returns:

A dictionary that contains each parameter tensor’s shape.

Return type:

Dict[str, Tuple[int, ...]]

class StereoToMidSide(normalize=True)

Bases: Module

A simple stereo-to-mid-side conversion.

forward(input_signals)

Processes input audio with the processor and given parameters.

Parameters:

input_signals (FloatTensor, $B\times 2\times L$) – A batch of input audio signals; must be stereo.

Returns:

A batch of output signals of shape $B\times 2\times L$.

Return type:

FloatTensor

parameter_size()

Returns:

A dictionary that contains each parameter tensor’s shape.

Return type:

Dict[str, Tuple[int, ...]]

class MidSideToStereo(normalize=True)

Bases: Module

A simple mid-side-to-stereo conversion.

forward(mid, side)

Processes input audio with the processor and given parameters.

Parameters:

• mid (FloatTensor, $B\times 1\times L$) – A batch of mid audio signals.

• side (FloatTensor, $B\times 1\times L$) – A batch of side audio signals.

Returns:

A batch of output signals of shape $B\times 2\times L$.

Return type:

FloatTensor

parameter_size()

Returns:

A dictionary that contains each parameter tensor’s shape.

Return type:

Dict[str, Tuple[int, ...]]