MelGAN
A Pytorch Implementation of MelGAN (Mel Spectrogram --> Waveform)
Requirements
PyTorch 1.2.0 & python 3.6 & Librosa
Examples
Step 1. Download Dataset
- LJSpeech : https://keithito.com/LJ-Speech-Dataset/
Step 2. Preprocessing (Preparing Mel Spectrogram)
python preprocessing.py --in_dir ljspeech --out_dir DATASETS/ljspeech
Step 3. Train MelGAN
-c: configurations & hyper parameteres in json
-m: model directory name
python train.py -c configs/base.json -m test
Step 4. Synthesize
Run Sample Test.ipynb. If it doesn't work, please try nbviewer.
While improving, I share a temporary checkpoint of generator, which only runs about 140K steps: link
Put the checkpoint file into "./logs/test/".
You can get an synthesized audio sample of the temporary generator on the above notebook file.
References
- MelGAN: https://arxiv.org/abs/1910.06711
- Base codes are higly adopted from: https://github.com/ksw0306/ClariNet