Wilde

Wilde is a command line tool to find and extract music from podcasts

Usage: wilde [OPTIONS] PATH

Options:
  -t, --file_type [mp3|wav]       The file format. Defaults to mp3.
  -s, --size INTEGER              Segmentation size in seconds. Defaults to
                                  600.

  -o, --output PATH               Path for the output files.
  -v, --verbose                   Output verbose logging information.
  -md, --minimal_duration INTEGER
                                  Minimal duration music has to be when
                                  exporting

  -jd, --join_distance INTEGER    The maximum time between joined segments
  --help                          Show this message and exit.

Installation

Download the latest release. Make sure you have python 3, ffmpeg and libsndfile installed.

Ubuntu

Unpack the release into a directory and run the following commands:

python -m pip install --upgrade pip setuptools wheel
python setup.py install

Windows

python -m pip install --upgrade pip setuptools wheel
pip install torch==1.4 torchvision==0.5 -f https://download.pytorch.org/whl/torch_stable.html
python setup.py install

Installation might take a while, since it has to install multiple large machine learning libraries.

Troubleshooting

ffmpeg

If you get errors such as The system cannot find the file specified you are missing ffmpeg. Do one of the following:

Install ffmpeg from http://www.ffmpeg.org/
If you are using conda then you can get the package from conda-forge by running:

conda config --add channels conda-forge
conda config --set channel_priority strict
conda install --yes --quiet ffmpeg

Libsndfile

If you get an error about missing libsndfile do one of the following:

Install from the official website here on both Linux and Windows
Run sudo apt-get install libsndfile on Ubuntu
If you are using conda then you can get the package from conda-forge by running:

conda config --add channels conda-forge
conda config --set channel_priority strict
conda install --yes --quiet libsndfile

Tensorflow

If you are missing tensorflow upon install, go to tensorflow.org and get the link for your wheel package. Then run the following command, replacing the link with your new link:

python -m pip install --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-2.3.0-cp38-cp38-manylinux2010_x86_64.whl

PyTorch on Conda

If you get an error regarding torch during the setup installation, try running the following command before running the setup again:

conda install pytorch

Using your GPU

By default Wilde will not use your GPU. If you want to speed up Wilde, you have to enable GPU support for tensorflow. This can be done by following these instructions:

Windows Tutorial

Ubuntu Tutorial