---
license: mit
tags:
  - text-to-speech
  - grad-tts
  - hifi-gan
  - vocoder
language:
  - en
---

# Grad-TTS (LJSpeech) + HiFi-GAN checkpoints — redistribution

A convenience mirror of the pretrained **Grad-TTS** (LJSpeech, `pe_scale=1000`)
and **HiFi-GAN** checkpoints, hosted so the
[fast-indic-tts](https://github.com/Abhijit89Kumar/fast-indic-tts) demo Space can
fetch them via `huggingface_hub`. All credit to the original authors.

| File | Source | License |
|---|---|---|
| `grad-tts.pt` | [Grad-TTS, Huawei Noah's Ark Lab](https://github.com/huawei-noah/Speech-Backbones/tree/main/Grad-TTS) | MIT |
| `hifigan.pt`, `hifigan-config.json` | [HiFi-GAN, jik876](https://github.com/jik876/hifi-gan) | MIT |

These weights are **not** trained or modified by me; this repository only mirrors
them. Please cite the original Grad-TTS and HiFi-GAN works.