---
language: en
datasets:
- ljspeech
tags:
- audio
- text-to-speech
---

# HiFi-GAN

[HiFi-GAN](https://arxiv.org/abs/2010.05646) vocoder trained on the [LJ Speech dataset](https://keithito.com/LJ-Speech-Dataset/). The modeling code is based on the [official implementation](https://github.com/jik876/hifi-gan) and the [fairseq adaptation](https://github.com/pytorch/fairseq).

## Usage

```python
from transformers import AutoModel

model = AutoModel.from_pretrained("jaketae/hifigan-lj-v1", trust_remote_code=True)
```