FB-MelGAN

A pytorch implementation of the FB-MelGAN(https://arxiv.org/pdf/2005.05106.pdf)

Prepare dataset

• Download dataset for training. This can be any wav files.
• Edit configuration in utils/audio.py
• Process data: python process.py --wav_dir="wavs" --output="data"

Pre_train & Train & Tensorboard

• python pre_train.py --input="data/train"
• python train.py --input="data/train"
• tensorboard --logdir logdir

Inference

• python generate.py --input="data/test"

Reference

• Multi-band MelGAN: Faster Waveform Generation for High-Quality Text-to-Speech(https://arxiv.org/pdf/2005.05106.pdf)
• MelGAN: Generative Adversarial Networks for Conditional Waveform Synthesis(https://arxiv.org/pdf/1910.06711.pdf)
• kan-bayashi/ParallelWaveGAN(https://github.com/kan-bayashi/ParallelWaveGAN)
• Parallel WaveGAN(https://arxiv.org/pdf/1910.11480.pdf)