This repository uses CycleGAN to recreate Piano and/or Guitar Sounds from an unpaired pool of data.
Link to Presentation: https://docs.google.com/presentation/d/1gcknmPha5GIwbzGci1zKCR1IkCU0p2PI67rvmIrL4-U/edit?usp=sharing
Link to Soundcloud of Results: https://soundcloud.com/eee298deeplearning
Link to Results: https://drive.google.com/open?id=1e5-mZzr2t0CrSoa-7Ql5LjKFcIKiCXXz https://drive.google.com/open?id=1erY3FnCvJ3kRz2sKzQxtVuzsRKuNdH7X
Link to Processed Audio: https://drive.google.com/open?id=1b0TnImwQsykZLj6yTD6WuSHV9fk9q6Qw
Link to Dataset: https://drive.google.com/open?id=1tWAUjwLRPwd2yu78yVaC2f3znv9Y3N4S
To run:
- Install Dependencies
- Download Dataset
- Run process_audio.py
- Train with cyclegan.py
- To synthesize data, run synthesizer.py