We proceed with the recipe as follows:
- Install NSynth by creating a separate conda environment. Create and activate a Magenta conda environment using Python 2.7 with Jupyter Notebook support:
conda create -n magenta python=2.7 jupytersource activate magenta
- Install the Magenta pip package and the librosa, which is used for reading audio formats:
pip install magentapip install librosa
- Install a pre-built model from the internet http://download.magenta.tensorflow.org/models/nsynth/wavenet-ckpt.tar and download a sample sound https://www.freesound.org/people/MustardPlug/sounds/395058/. Then run the notebook contained in the demo directory (in my case http://localhost:8888/notebooks/nsynth/Exploring_Neural_Audio_Synthesis_with_NSynth.ipynb ...