How To Build

Instructions

The build system is based on cmake.

mkdir build && cd build
cmake ..
cmake --build .

The best options for real-time integration would be

OSX

mkdir build && cd build
cmake -GXcode -DCMAKE_INSTALL_PREFIX=$(pwd)/install -Dspleeter_enable_tests=OFF -Drtff_use_mkl=ON ..
cmake --build .

Windows (MSVC2019)

mkdir build && cd build
cmake -DCMAKE_INSTALL_PREFIX=$(pwd)/install -Drtff_use_mkl=ON ..
cmake --build . --config Release --target INSTALL

Options

spleeter_enable_tests

Enable or disable the unit tests. Default is ON

spleeter_enable_high_resolution

Enable or disable the process extension up the 16KHz. Default is OFF

spleeter_regenerate_models

Uses the cmake/export_spleeter_models python scripts to run the conversion of Spleeter models for c++. Default is OFF

Requires Conda

spleeter_enable_filter

Enable the filter interface. Default is ON

spleeter_input_frame_count

Set spleeter T parameter

Available options are 64 / 128 / 256 / 512. Default is 64

rtff_use_mkl

Use the Inter MKL to speed up the Fourier transform and the matrix operations.

Requires a properly install Intel MKL

rtff_use_fftw

Use the FFTW for Fourier transform computation.

Only for OSX and Linux.

Tensorflow

The project relies heavily on the tensorflow library. We use the Tensorflow C API for portability. Check the cmake/add_tensorflow.cmake script for more.