Skip to content
A PyTorch implementation of "WaveFlow: A Compact Flow-based Model for Raw Audio"
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md init Oct 7, 2019
data.py
model.py
modules.py cleanup Oct 7, 2019
preprocessing.py init Oct 7, 2019
synthesize.py cleanup Oct 7, 2019
train.py --load_param_only Oct 8, 2019
utils.py init Oct 7, 2019

README.md

WaveFlow : A Compact Flow-based Model for Raw Audio

This is an unofficial PyTorch implementation of a paper "WaveFlow : A Compact Flow-based Model for Raw Audio".

Currently WIP. The implementation details may not be faithful.

Requirements

PyTorch 1.1.0 or later (tested on 1.2.0) & python 3.6 & Librosa

Examples

Step 1. Download Dataset

Step 2. Preprocessing (Preparing Mel Spectrogram)

python preprocessing.py --in_dir /path/to/ljspeech/data/root --out_dir ./ljspeech_data

Step 3. Train the Model

python train.py --model_name waveflow_h8_r64 --n_height 8 --res_channels 64 --n_layer_per_cycle 1

python train.py --model_name waveflow_h64_r64 --n_height 64 --res_channels 64 --n_layer_per_cycle 5

python train.py --model_name waveflow_h32_r128 --n_height 32 --res_channels 128 --n_layer_per_cycle 3

Step 4. Synthesize

Specify --load_step and --num_samples that looks like:

python synthesize.py --model_name waveflow_h8_r64 --n_height 8 --res_channels 64 --n_layer_per_cycle 1 --load_step 100000 --num_samples 5

References

You can’t perform that action at this time.