A smart robot that raps like your favorite artist.
Install the header files for Python 2.7 and the HDF5 library:
sudo apt-get -y install python2.7-dev
sudo apt-get install libhdf5-dev
Install the following libraries using pip:
Cython
numpy
argparse
h5py
six
Install torch.
Install Mathematica through any means you would like.
Place all text training data in lyrics.txt and run in the /torch-rnn directory.
python scripts/preprocess.py --input_txt lyrics.txt --output_h5 lyrics.h5 --output_json lyrics.json --val_frac 0.01 --test_frac 0.01
th train.lua -gpu -1 -input_h5 lyrics.h5 -input_json lyrics.json -rnn_size 512 -num_layers 2 -checkpoint_every 20 -checkpoint_name cv/checkpoint
th sample.lua -gpu -1 -checkpoint cv/`ls -v cv | tail -n 1` -sample 1
Place the desired lyrics in generated_lyrics.txt.
Run wolframscript make_rap.wls in the root directory of the project.
The script will produce rap.wav, an audio file containing the rap demo.
A pre-made sample is already stored in the current rap.wav.