OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Vits

Vits is an end-to-end text-to-speech model that synthesizes audio from text using variational inference, normalizing flows and adversarial learning. Its stochastic duration predictor supports varied speech rhythms and pitches for the same input. The Python/PyTorch repository includes inference and training examples, pretrained models, and dataset-specific examples for LJ Speech and VCTK.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Vits — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Vits when you need a documented text-to-speech model workflow, from pretrained inference to training on the supported LJ Speech or VCTK examples.

Who it is for

  • Developers exploring or using a PyTorch text-to-speech model.
  • Researchers or practitioners training a single-speaker model on LJ Speech or a multi-speaker model on VCTK.

Documented uses

  • Generate speech from text with a pretrained Vits model.
  • Train a single-speaker TTS model using LJ Speech.
  • Train a multi-speaker TTS model using VCTK.
  • Explore generated audio through the sample demo or interactive Colab notebook.
Before you choose

Requirements & limits

The repository is public and declares an MIT license. Its README links to pretrained models, an audio-sample demo and an interactive Colab notebook.

  • Training requires Python 3.6 or later, installed Python requirements and a dataset.
  • VCTK training instructions specify downsampling WAV files to 22050 Hz.
  • Building Monotonic Alignment Search is part of the documented prerequisites.
  • Preprocessed phonemes are provided for LJ Speech and VCTK; preprocessing is described for users’ own datasets.
  • The model’s stochastic duration predictor supports multiple possible rhythms and pitches for a text input.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare a Vits training setup

Set up the documented prerequisites and dataset links before following a training example.

Before you start

  • Python 3.6 or later
  • A cloned Vits repository and its Python requirements installed
  • LJ Speech for single-speaker training, or VCTK for multi-speaker training
  • For VCTK, downsampled WAV files at 22050 Hz

Steps

  1. Install the repository’s Python requirements; the README notes that espeak may need to be installed first.

  2. Download and extract LJ Speech or VCTK. For VCTK, use downsampled WAV files at 22050 Hz.

  3. Rename or create a link to the dataset WAV folder as DUMMY1 for LJ Speech or DUMMY2 for VCTK.

  4. Build Monotonic Alignment Search from the monotonic_align directory using the README’s documented build command.

Expected outcome

The documented Python dependencies, dataset location and alignment component are prepared for the training stage.

Guide limitations

  • This guide covers setup only; the repository lists separate training commands for LJ Speech and VCTK.
  • Users with their own datasets must run preprocessing; the README says preprocessing is not needed for the supplied LJ Speech and VCTK phoneme files.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
jaywalnut310/vits · Official homepage title · retained in its source language
How we review sourcesSuggest a correction