Repolode
Submit
fishaudio

fishaudio/fish-speech

Open-source text-to-speech system with voice cloning, natural-language emotion control and multi-speaker generation, built on LLM-style models.

Open-source alternative to:ElevenLabs

View on GitHubWebsite
Stars
33,309
Forks
2,894
Open issues
21
Last commit
6 days ago

What is it?

Fish Speech is an open-source text-to-speech project from Fish Audio. It turns text into speech and can clone a voice from a short reference clip, roughly 10 to 30 seconds for the S2-Pro model. Inline tags such as [whisper], [angry] or [laughing] control emotion and prosody within the text, and speaker IDs let one text contain several characters without generating each speaker separately.

The project ships two model generations. Drama 3 is the newer flagship, aimed at multi-character scenes and long-form content, and adds multi-turn revision and local editing of a single word within existing audio. S2-Pro is the previous generation, described as small enough for a consumer GPU and covering about a hundred languages. Both share the same command line and API server, and the model type is detected from the checkpoint's config.

The repository documents installation, weight download, Docker deployment, and inference through the command line, an API server, or a WebUI. Model weights are published on Hugging Face, and the codebase and weights are released under the Fish Audio Community License, with the S2-Pro weights under a separate research license.

Key features

  • 01Voice cloning from a short reference clip
  • 02Inline natural-language tags control emotion and prosody
  • 03Multi-speaker generation within one text
  • 04Local word-level editing of existing audio (Drama 3)
  • 05Streaming output for long text
  • 06Command line, API server and WebUI inference

Use cases

  • Generating voiceovers or audiobook narration with expressive delivery
  • Producing multi-character scripted scenes or dialogue audio
  • Building a self-hosted TTS backend exposed through an API
  • Experimenting with voice cloning and speech synthesis research

Quick start

  1. 1

    Read the installation guide for environment setup, weight download and Docker deployment

  2. 2

    Read the inference guide for command line, API server and WebUI usage

Things to know

  • The codebase and model weights are under the Fish Audio Community License, not a standard open-source license.
  • The S2-Pro weights are under the Fish Audio Research License, which may restrict commercial use.
  • Setup depends on external documentation and separately downloaded model weights.

Topics

#text-to-speech#voice-cloning#speech-synthesis#transformer#llm#multilingual#audio#llama#tts#valle

FAQ

What is fish-speech?
Open-source text-to-speech system with voice cloning, natural-language emotion control and multi-speaker generation, built on LLM-style models. Fish Speech is an open-source text-to-speech project from Fish Audio. It turns text into speech and can clone a voice from a short reference clip, roughly 10 to 30 seconds for the S2-Pro model. Inline tags such as [whisper], [angry] or [laughing] control emotion and prosody within the text, and speaker IDs let one text contain several characters without generating each speaker separately.
Is fish-speech free to use? What license does it have?
fish-speech is open source under the Other license. Check the license terms before using it in commercial products.
Is fish-speech still maintained?
Actively maintained. The last commit to fish-speech was on Oct 14, 2026.
What can I use fish-speech for?
Generating voiceovers or audiobook narration with expressive delivery Producing multi-character scripted scenes or dialogue audio Building a self-hosted TTS backend exposed through an API Experimenting with voice cloning and speech synthesis research
How do I install fish-speech?
Read the installation guide for environment setup, weight download and Docker deployment Read the inference guide for command line, API server and WebUI usage
Where can I get fish-speech?
The source code and installation instructions are on GitHub: https://github.com/fishaudio/fish-speech

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects