← Back to Discover

TTS WebUI

Open-source interface you run locally that brings many speech, music and audio models together in one place, with an extension system and an OpenAI-compatible speech endpoint.

Text to speechMusic generation
Visit website ↗

What it does

TTS WebUI is an open-source application you run on your own machine. It combines a Gradio backend with a React front end and collects many audio models behind a single interface, including Bark, Tortoise, Kokoro, MusicGen, Stable Audio, RVC and Demucs. Models that are not installed by default are added as extensions. It can be installed with the TTS WebUI Ignition installer (a winget command is given for Windows), a legacy installer script, a manual Python setup or a Docker image. The base installation is around 10.7 GB and each model needs a further 2 to 8 GB. Once running, the Gradio server listens on port 7770 and the React UI on port 3000. Extensions are Python packages that can be installed from inside the interface, and a community catalogue lists extensions made by others. An OpenAI-compatible speech endpoint lets clients such as Silly Tavern and OpenWebUI use the voices it generates. The code is MIT licensed, but model weights carry their own licences. The page notes that the MusicGen and AudioGen weights are CC BY-NC 4.0, which is a non-commercial licence.

Capabilities

  • One interface for many speech, music and audio models
  • Extension system for adding models that are not installed by default
  • OpenAI-compatible speech endpoint for Silly Tavern, OpenWebUI and similar clients
  • Installer, manual setup and Docker image options
  • Community extension catalogue

When to use it

  • You want to try several open speech and music models without setting each one up separately
  • You need a local text-to-speech endpoint for chat front ends such as Silly Tavern or OpenWebUI
  • You are comfortable with a local install that needs several gigabytes of disk per model