Metadata-Version: 2.1
Name: hifigan-vocoder
Version: 0.1.1
Summary: A wrapped hifi-gan vocoder for easy use.
Home-page: https://github.com/mushanshanshan/hifigan
Author: Mushan
License: MIT
Keywords: None
Platform: UNKNOWN
Classifier: Environment :: Console
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3.7
Description-Content-Type: text/markdown
License-File: LICENSE

# hifigan
A wrapped hifi-gan vocoder for easy use. 

Source: https://github.com/jik876/hifi-gan

# Install

```bash
pip install hifigan-vocoder
```

# Usage
```python

from hifigan_vocoder import hifigan, get_sample_mel

mel = get_sample_mel()
print(mel.shape) # torch.Size([1, 80, 2062])

model = hifigan(dataset='uni', device='cpu') # dataset in ['uni', 'vctk']; device in ['cpu', 'cuda']; checkpoint will be downloaded from google driver.

audio = model.infer(mel)
print(mel.shape) # (527872,)

```

