Skip to content

IndexTTS

A GPT-style TTS model with pinyin and pause control

Visit official site

What IndexTTS is

The model is described as a GPT-style text-to-speech system built mainly on XTTS and Tortoise, with character-pinyin mixed modelling for Chinese and BigVGAN2 for audio quality. It is trained on tens of thousands of hours of data and described as open source for academic and commercial work.

What you can do with it

  • Correct Chinese character pronunciation with pinyin
  • Control pauses through punctuation
  • Clone a voice from a zero-shot sample
  • Synthesize speech in Chinese and English

Application scenarios

  • Supply voice assistants with high-quality speech output
  • Build audiobooks with reading in several languages
  • Add narration quickly during video production

Getting started

  • Clone or download the code from the GitHub repository
  • Install dependencies such as PyTorch
  • Prepare an audio dataset and preprocess it
  • Train the model with the provided script or load a pretrained one

Pros & cons

✓ What we like

  • Character-pinyin mixed modelling is used for Chinese
  • Chinese and English synthesis are both listed
  • The training set covers tens of thousands of hours

! What to watch out for

  • The model is built mainly on XTTS and Tortoise
  • Training or loading a pretrained model is part of the setup
  • Pinyin correction is aimed at Chinese character pronunciation

FAQ

What controls pronunciation and pauses?

The material lists pinyin correction for Chinese characters and punctuation-based pause control.

What does the documented setup involve?

The steps cover cloning the GitHub repository, installing PyTorch and other dependencies, and preparing audio data.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review