Authors: Wei-ning WANG, Jia-hui LI, Yi-fan LI, Xiao-fen XING
Official PyTorch implementation of paper "SCTG: Style Conditioned Music Generation With Transformer-GANs".
Some generated demo can be downloaded from here
The generated demo in mp3 format shared here
See the paper for details of the model.
The code for training is coming soon.
For inference, you can see the details in main.py
The Emopia dataset is from Emopia The Pianst8 dataset is from Zenodo
The code for preparing the training data is in the dataset folder.
The checkpoints for SCTG can be downloaded from here
- PyTorch codes for Transformer Generator is modified from YatingMusic/compound-word-transformer.