文件列表:

main.py (11199, 2018-11-13)
runme.sh (426, 2018-11-13)
utils/ (0, 2018-11-13)
utils/config.py (458, 2018-11-13)
utils/datasets.py (3207, 2018-11-13)
utils/models.py (14885, 2018-11-13)
utils/mu_law.py (2239, 2018-11-13)
utils/utilities.py (1544, 2018-11-13)

### WaveNet pytorch implementation This code implements speech synthesis with WaveNet using pytorch. This code is partly translated from the tensorflow implementation: https://github.com/ibab/tensorflow-wavenet ## Requirements pytorch: 0.4.0 ## Data preparation VCTK Corpus includes 44,352 speech sentences uttered by 109 native English speakers. Each utterance last for a few seconds. Download VCTK dataset from https://homepages.inf.ed.ac.uk/jyamagis/page3/page58/page58.html. The dataset looks like:

dataset_dir
├── wav48
│    ├── p225
│    │     ├── p225_001.wav
│    │     └── ...
│    ├── p226
│    │     ├── p226_001.wav
│    │     └── ...
│    └── ...
├── txt
│    ├── p225
│    │     ├── p225_001.txt
│    │     └── ...
│    ├── p226
│    │     ├── p226_001.txt
│    │     └── ...
│    └── ...
├── speaker-info.txt
└── ...

## Run Modify the paths in runme.sh Run commands in runme.sh line by line. ## Results ## FAQ This code runs wtih a single GPU card with 12 GB memory. If you are running out of GPU memory, then modify Dataset to shorten the audio clip. ## References [1] Van Den Oord, Aron, et al. "WaveNet: A generative model for raw audio." SSW. 2016. [2] Paine, Tom Le, et al. "Fast wavenet generation algorithm." arXiv preprint arXiv:1611.09482 (2016).

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：