cycle_gan_vc icon indicating copy to clipboard operation
cycle_gan_vc copied to clipboard

Reproducing PARALLEL-DATA-FREE VOICE CONVERSION USING CYCLE-CONSISTENT ADVERSARIAL NETWORKS (https://arxiv.org/pdf/1711.11293.pdf)

Python and Toolkit Version

Python:      '3.5.2'
numpy:       '1.13.1'
tensorflow:  '1.4'

Preprocess example script

Dataset: VCTK Example script: Gist link

Training

python3 gan_train.py

Inference

python3 gan_test.py

Hyperparameter

All hyperparameter tuning could be done in hyperparams.py.

Experiments and Samples

Currently done on VCTK dataset and can be found in results/

TODO

  • [ ] Remove per-speaker information from Graph eg. f0, mcep_mean...
  • [ ] Neater data preprocess