WS::AudioModel Class Reference

Main entry point class to handle AI audio processing. More...

#include <AudioModel.h>

Public Member Functions
	AudioModel (std::string const &activation="tanh", unsigned int sampleRate=44100)
	Constructor for AudioModel.
	~AudioModel ()
	Destructor for AudioModel.
	AudioModel (const AudioModel &)=delete
AudioModel &	operator= (const AudioModel &)=delete
bool	prepare (std::string const &modelName)
	Loads and prepares the model for processing.
void	setNewParam (std::string const &param)
	Adds a new parameter to the model.
bool	setLicense (std::string const &license)
	Adds license.
void	setParamValueAt (size_t param, double value)
	Sets the value of a parameter by index.
void	setParamValueAt (size_t param, float value)
	Sets the value of a parameter by index.
void	setParamValueAt (std::string const &param, double value)
	Sets the value of a parameter by name.
void	setParamValueAt (std::string const &param, float value)
	Sets the value of a parameter by name.
bool	process (const float *input, float *output)
	Processes audio through the model.
bool	loadJsonEQParameters (std::string const &inJsonConfigPathName, int samplingRate)
	Loads EQ parameters from a JSON configuration file.
size_t	getFrameLength () const
	Gets the frame length required for processing.
size_t	getNumberOfFilters () const
	Gets the number of filters in the model.
size_t	getDenseLocalSize () const
	Gets the dense local size of the model.
size_t	getMaxPoolingPoolSize () const
	Gets the max pooling pool size.
size_t	getNumberOfParams () const
	Gets the number of parameters supported by the model.
void	setLogger (WS::Logger *logger)
	Sets a logger for the model.
float *	getValidationValues (const std::string &bufferName, size_t &filterCnt, size_t &sampleCnt)
	Gets validation values for a specific buffer.

Private Member Functions
float	oneminus (float x)
float	softplus (float x)
float *	getBreakPoints (size_t numBreakPoints)

Private Attributes
Impl *	pImpl
WS::Logger *	mLogger {nullptr}

Friends
class	::AudioModelUT

Detailed Description

Main entry point class to handle AI audio processing.

The AudioModel class provides the primary interface for loading and using AI models for audio processing. It supports various activation functions, parameter setting, and model processing capabilities.

// Basic usage example
static const std::string ACTIVATION{"tanh"};
std::unique_ptr<AudioModel> audioModel{new AudioModel(ACTIVATION)};
 
// Load and prepare the model binaries (.dat files)
std::string modelName = "/path/to/model/";
if(!audioModel->prepare(modelName)) {
    std::cout << "ERROR: Could not prepare the model properly." << std::endl;
    return 1;
}
 
// Set parameters if needed
if(audioModel->getNumberOfParams() > 0) {
    audioModel->setParamValueAt(0, 0.5f);
}
 
// Process audio
audioModel->process(inputBuffer, outputBuffer);

Constructor & Destructor Documentation

AudioModel() [1/2]

WS::AudioModel::AudioModel	(	std::string const &	activation = "tanh",
		unsigned int	sampleRate = 44100
	)

Constructor for AudioModel.

Parameters

activation

The activation function to use (default: "tanh")

Creates a new AudioModel instance with the specified activation function. Unless otherwise stated, most of the WS AI models use "tanh"

~AudioModel()

WS::AudioModel::~AudioModel

(

)

Destructor for AudioModel.

AudioModel() [2/2]

WS::AudioModel::AudioModel ( const AudioModel &  )

delete

Member Function Documentation

getBreakPoints()

float * WS::AudioModel::getBreakPoints ( size_t  numBreakPoints )

private

getDenseLocalSize()

size_t WS::AudioModel::getDenseLocalSize

(

)

const

Gets the dense local size of the model.

Returns

Dense local size

getFrameLength()

size_t WS::AudioModel::getFrameLength

(

)

const

Gets the frame length required for processing.

Returns

Number of samples per frame

getMaxPoolingPoolSize()

size_t WS::AudioModel::getMaxPoolingPoolSize

(

)

const

Gets the max pooling pool size.

Returns

Max pooling pool size

getNumberOfFilters()

size_t WS::AudioModel::getNumberOfFilters

(

)

const

Gets the number of filters in the model.

Returns

Number of filters

getNumberOfParams()

size_t WS::AudioModel::getNumberOfParams

(

)

const

Gets the number of parameters supported by the model.

Returns

Number of parameters

getValidationValues()

float * WS::AudioModel::getValidationValues	(	const std::string &	bufferName,
		size_t &	filterCnt,
		size_t &	sampleCnt
	)

Gets validation values for a specific buffer.

Parameters

bufferName	Name of the buffer
filterCnt	Output parameter to receive filter count
sampleCnt	Output parameter to receive sample count

Returns

Pointer to the validation values array

loadJsonEQParameters()

bool WS::AudioModel::loadJsonEQParameters	(	std::string const &	inJsonConfigPathName,
		int	samplingRate
	)

Loads EQ parameters from a JSON configuration file.

Parameters

inJsonConfigPathName	Path to the JSON configuration file
samplingRate	Sample rate of the audio to be processed

Returns

true for success, false otherwise (no EQ effects will be applied on failure)

If EQ effects are to be applied, a JSON formatted config file for the parameters needs to be loaded prior to calling process().

std::string eqConfigFile = "eq_settings.json";
if(!audioModel->loadJsonEQParameters(eqConfigFile, 44100)) {
    std::cout << "Warning: Could not load EQ parameters." << std::endl;
}

oneminus()

float WS::AudioModel::oneminus ( float  x )

private

operator=()

AudioModel & WS::AudioModel::operator= ( const AudioModel &  )

delete

prepare()

bool WS::AudioModel::prepare

(

std::string const &

modelName

)

Loads and prepares the model for processing.

Parameters

modelName

Path to the model directory

Returns

true if successful, false if the model cannot be found or loaded

Prepare needs to be called once to allocate the proper layers before calling process() multiple times for processing audio frames.

std::string modelPath = "/path/to/model/";
if(!audioModel->prepare(modelPath)) {
    std::cout << "Error: Could not prepare the model properly." << std::endl;
    return 1;
}

process()

bool WS::AudioModel::process	(	const float *	input,
		float *	output
	)

Processes audio through the model.

Parameters

input	Pointer to input buffer of audio samples (size must match getFrameLength())
output	Pointer to output buffer where processed audio will be stored

Returns

true if processing was successful, false otherwise

// Process example
float* inputBuffer = new float[audioModel->getFrameLength()];
float* outputBuffer = new float[audioModel->getFrameLength()];
// Fill inputBuffer with audio samples...
 
if(audioModel->process(inputBuffer, outputBuffer)) {
    // Use processed audio in outputBuffer...
}

setLicense()

bool WS::AudioModel::setLicense

(

std::string const &

license

)

Adds license.

Parameters

license

License to use

This is just a placeholder for a license checker We need a way to manage the license

setLogger()

void WS::AudioModel::setLogger

(

WS::Logger *

logger

)

Sets a logger for the model.

Parameters

logger

Pointer to a Logger instance

setNewParam()

void WS::AudioModel::setNewParam

(

std::string const &

param

)

Adds a new parameter to the model.

Parameters

param

Name of the parameter

Parameters will be added in order and indexed according to their order:

• First parameter added is index 0
• Second parameter added is index 1
• And so on...

setParamValueAt() [1/4]

void WS::AudioModel::setParamValueAt	(	size_t	param,
		double	value
	)

Sets the value of a parameter by index.

Parameters

param	Index of the parameter
value	Double value to set (0.0 to 1.0 recommended)

setParamValueAt() [2/4]

void WS::AudioModel::setParamValueAt	(	size_t	param,
		float	value
	)

Sets the value of a parameter by index.

Parameters

param	Index of the parameter
value	Float value to set (0.0 to 1.0 recommended)

setParamValueAt() [3/4]

void WS::AudioModel::setParamValueAt	(	std::string const &	param,
		double	value
	)

Sets the value of a parameter by name.

Parameters

param	Name of the parameter
value	Double value to set (0.0 to 1.0 recommended)

setParamValueAt() [4/4]

void WS::AudioModel::setParamValueAt	(	std::string const &	param,
		float	value
	)

Sets the value of a parameter by name.

Parameters

param	Name of the parameter
value	Float value to set (0.0 to 1.0 recommended)

softplus()

float WS::AudioModel::softplus ( float  x )

private

Friends And Related Function Documentation

::AudioModelUT

friend class ::AudioModelUT

friend

Member Data Documentation

mLogger

WS::Logger* WS::AudioModel::mLogger {nullptr}

private

pImpl

Impl* WS::AudioModel::pImpl

private

---

The documentation for this class was generated from the following file:

• AudioModel.h