fftune

This is an audio pitch detection library written in modern C++20, providing multiple pitch detection methods and a ready to use binary to convert an input WAV file into a MIDI file.

Building

Note that Arch Linux users can simply install the native fftune-git package from the AUR.

The following libraries are required to be installed:

fftw3f - For fast Fourier transformations
libsmf - For Midi file support
libsndfile - For audio file support
sfizz - For sfz file support

fftune is built using cmake:

cmake -B build
cmake --build build

# To install
cmake --install build

Dockerfile

A Dockerfile is provided for distributions that can not satisfy the dependency requirements. To build the container, use podman or docker:

podman build -t fftune .
# Convert audio.wav in current directory to audio.midi
podman run -v "$PWD:/tmp/data" fftune

Usage

The most basic usecase is:

audio-to-midi /path/to/audio.wav

It is also possible to stream data in realtime via stdin in an OS-independent way. For example on Linux with Pulseaudio, you can process a live recording in realtime with:

parec --rate 48000 --format s32le --channels 1 | \
	sox -t raw -r 48000 -b 32 -e signed-integer -c 1 - -t wav - | \
	audio-to-midi --verbose -

There are many more options available, view them by showing the help with audio-to-midi -h or by looking at the provided man-page with man audio-to-midi.

Documentation

This library comes with a developer API documentation. To build it use:

cd doc
doxygen

You can then open doc/html/index.html in a web browser.

If you are a developer and want to use this library in your own program, you can include the header with #include <fftune.hpp> and link against libfftune.so. Examples for using this library can be found in the examples directory. This library comes with pkg-config support, so it is not necessary to manually link against it.

Tests

fftune ships with a testing suite, for more information checkout the tests directory.

fftune / fftune