tangkk / audiomod

audiomod is a project for audio modifications, including audio manipulators such as time-stretching, pitch-shifing, formant-changing, and audio filters such as vibrato, tremolo, ring-modulation, compression, reverb, equalizer, etc. It is a good starting point for audio signal processing developpers to build more advanced applications, or for students to learn and practice their audio coding techniques.

audiomod quickstart

Introduction

audiomod is a project for audio modifications, including audio manipulators such as time-stretching, pitch-shifing, formant-changing, and audio filters such as vibrato, tremolo, ring-modulation, compression, reverb, equalizer, etc. It is a good starting point for audio signal processing developpers to build more advanced applications, or for students to learn and practice their audio coding techniques.

References

This project refers deeply to the algorithms and code of the following opensourced projects:

• Rubberband
• Audio Effect Book Code
• Dafx Code

This project tries to unified these three sources into one framework, and provide "build-abilities" for desktop and mobile platforms as well.

License

Since the core part of the code are modified based on Rubberband, the license of this project follows GNU General Public License (GPL).

Build standalone executables for Linux and Mac

under the source folder do:

mkdir build
cd build
cmake ..
make

This will process an executable called audiomod-exe, as well as a static library libaudiomod.a.

usage:

usage: ./audiomod-exe dafx_name infile outfile <args> (dafx: constant, time_stretch, normal_pitchshift, formant_pitchshift, gender_change, vocoder, vocoder_chord, robotic, whisper, vibrato, delay, flanger, chorus, ringmod, tremolo, compressor, limiter, autogain, reverb, autowah, phaser, loudnessmeter, equalizer, gain, envelope, )

The <args> here can be looked up in main.cc. For example: ./audiomod-exe normal_pitchshift stereo.wav out.wav 4 1 2048 runs a 4-semitone up pitch-shift to stereo.wav, using a phasevocoder with a 2048-point FFT. Note that currently only .wav format is supported.

Build for Android

For an android build, you need to use cmake's cross-platform build process, which depends on the Android NDK. Here is an example for arm64:

cmake .. -DCMAKE_TOOLCHAIN_FILE=/path/to/ndk/18.1.5063045/build/cmake/android.toolchain.cmake \
    -DANDROID_NDK=/path/to/ndk/18.1.5063045/ \
    -DANDROID_ABI="arm64-v8a" \
    -DANDROID_STL="c++_shared" \
    -DCMAKE_BUILD_TYPE=Release \
    -DBUILD_LIBS=ON
	
cmake --build . --config Release

Please refer to build-android.sh for more details on this process.

Build for iOS

For an ios build, you need to have the Xcode installed, and then make use of cmake's cross-platform build process, which depends on the ios-cmake. Here is an example:

cmake .. -G Xcode -DCMAKE_TOOLCHAIN_FILE=../ios-cmake/ios.toolchain.cmake -DPLATFORM=OS -DBUILD_LIBS=ON
cmake --build . --config Release

Please refer to build-ios.sh for more details on this process.

SDK usage

To incorporate audiomod into your own project, first you need to build it for your target platform. After this step you get a libaudiomod.a or libaudiomod.so (for Android)

Include the audiomod.h in your source code. Then you create an object of the sound filter, and initialize it like this:

modbase * m_modbase = nullptr;

std::unique_ptr<phasevocoder> m_phasevocoder;

m_phasevocoder = std::make_unique<phasevocoder>(sample_rate, num_channels, 1, shift_pitch, NORMAL_SHIFT, coremode, fftsize);
        m_modbase = m_phasevocoder.get();

you should provide the parameters needed accordingly. Then you could process the sound with the following code:

for (int i = 0; i < file_length; i+=block_size) {
	int num_samples = input->read(buff, block_size);
	m_modbase->processBlock(buff, num_samples);
	if (m_modbase->outputReady()) {
		output->write(buff, num_samples);
	}
}

Refer to main.cc for more details on the exact procedure.

Finally, make sure to link the libaudiomod library in your building process.