Audio Synthesis with continuous Matrix Product States https://arxiv.org/abs/1911.11879
- TensorFlow
- Python 3.6.6
- model.py
It defines our RNN model and contains other methods, to sample from the trained model for example.
- train.py
It performs the training.
- data.py
It creates a synthetic dataset or reads in real data.
- reader.py
When using real data (like Nsynth), it picks out a subset of the data, that contains a desired feature: pitch, instrument ...
- make-small-dataset.py
When using real data (like Nsynth), it imports reader.py to make the dataset.