https://github.com/revsic/torch-nansypp
NANSY++: Unified Voice Synthesis with Neural Analysis and Synthesis
https://github.com/revsic/torch-nansypp
Last synced: 14 days ago
JSON representation
NANSY++: Unified Voice Synthesis with Neural Analysis and Synthesis
- Host: GitHub
- URL: https://github.com/revsic/torch-nansypp
- Owner: revsic
- License: mit
- Created: 2022-12-08T00:52:51.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2023-02-11T14:28:44.000Z (over 2 years ago)
- Last Synced: 2025-05-05T17:32:05.535Z (14 days ago)
- Language: Python
- Size: 136 KB
- Stars: 147
- Watchers: 27
- Forks: 11
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# torch-nansypp
Torch implementation of NANSY++: Unified Voice Synthesis with Neural Analysis and Synthesis, [[openreview](https://openreview.net/forum?id=elDEe8LYW7-)]
## TODO
1. breathiness perturbation
2. DEMAND based noise addition
## Requirements
Tested in python 3.7.9 conda environment.
## Usage
Initialize the submodule.
```bash
git submodule init --update
```
Download LibriTTS[[openslr:60](https://www.openslr.org/60/)], LibriSpeech[[openslr:12](https://www.openslr.org/12)] and VCTK[[official](https://datashare.ed.ac.uk/handle/10283/2651)] datasets.
Dump the dataset for training.
```
python -m speechset.utils.dump \
--out-dir ./datasets/dumped
```
To train model, run [train.py](./train.py)
```bash
python train.py
```
To start to train from previous checkpoint, `--load-epoch` is available.
```bash
python train.py \
--load-epoch 20 \
--config ./ckpt/t1.json
```
Checkpoint will be written on TrainConfig.ckpt, tensorboard summary on TrainConfig.log.
```bash
tensorboard --logdir ./log
```
[TODO] To inference model, run [inference.py](./inference.py)
[TODO] Pretrained checkpoints will be relased on [releases](https://github.com/revsic/torch-nansypp/releases).
To use pretrained model, download files and unzip it. Followings are sample script.
```py
from nansypp import Nansypp
ckpt = torch.load('t1_200.ckpt', map_location='cpu')
nansypp = Nansypp.load(ckpt)
nansy.eval()
```
## [TODO] Learning curve and Figures
## [TODO] Samples