loristrck is a wrapper for the C++ partial-tracking library Loris.

It is written in cython and targets python 3. The source of the library is included as part of the project and does not need to be installed previously.

C++ Library Dependencies:

• fftw3

Additional Python Module Dependencies:

• Cython
• NumPy
• pysndfile + libsndfile
• sounddevice to be able to play samples (see loristrck_play)

Optional dependencies:

• csound (in order to play .mtx.wav files, as generated by loristrck_pack)

Install dependencies (see each platform):

• fftw
• libsndfile

git clone https://github.com/gesellkammer/loristrck
cd loristrck
pip install -r requirements.txt
pip install .

loristrck is a C/C++ extension for python 3 (>=3.6) and needs a compiler present. It is developed in linux but should work in macOS and Windows without problems

brew install fftw
brew install libsndfile
pip install numpy cython
pip install loristrck

apt install libfftw3-dev libsndfile1-dev
pip install numpy cython
pip install loristrck

• Install libsndfile from http://www.mega-nerd.com/libsndfile/#Download

• Install fftw3 from http://www.fftw.org/install/windows.html

• Download the 32-bit binary package. Unzip to a directory of your choice. Suggestion: C:\\src. You should have then a folder C:\\src\\fftw

• Put that directory in your PATH (Control Panel/System/Advanced/Environmental Variables/)

  python setup.py install -LC:\\my\\path\\to\\fftw

At the command-line, do:

git clone https://github.com/gesellkammer/loristrck
cd loristrck
pip install -r requirements.txt
pip install .

Otherwise, install it via pip:

pip install numpy cython
pip install loristrck

(this assumes that you put the fftw3 source under C:\\src)

import loristrck as lt

samples, sr = lt.sndreadmono("/path/to/sndfile.wav")
partials = lt.analyze(samples, sr, resolution=60)
# partials is a python list of numpy arrays
# select a subset of most significant partials
selected, noise = lt.select(partials, mindur=0.02, maxfreq=12000, minamp=-60, minbp=2)
# print each partial as data
for partial in selected:
    print(partial)
# plot selected partials
lt.plot_partials(selected)
# now resynthesize both parts separately
lt.partials_render(selected, outfile="selected.wav")
lt.partials_render(noise, outfile="noise.wav")

Each partial will be a numpy array of shape = (numbreakpoints, 5) with the columns:

time, frequency, amplitude, phase, bandwidth

See the example scripts in bin for more complete examples

sndtrck: https://github.com/gesellkammer/sndtrck

eduardo dot moguillansky @ gmail dot com

GPL