# fishaudio/fish-speech

SOTA Open Source TTS

Repository: https://github.com/fishaudio/fish-speech
Canonical: https://ross.abutalabs.com/products/fish-speech
Homepage: https://speech.fish.audio
Language: Python
License: NOASSERTION
License Family: other
Topics: llama, transformer, tts, valle, vits, vqgan, vqvae
Last push: 2026-08-22T08:55:50+00:00

## Health v2 (maintenance only)
Score: 74/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 40, longevity 75
- inputs: {"age_days": 1058, "days_push": 11, "days_rel": 459, "gap_med": 29.5, "n_releases_24m": 7}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 32413, forks 2797 (observed 2026-08-28T04:11:57.549326+00:00)

## What it is
Fish Speech is an open-source state-of-the-art text-to-speech system (Fish Audio S2) trained on over 10 million hours of audio across ~50 languages, using a Dual-Autoregressive architecture with reinforcement learning alignment. It supports voice cloning, inline prosody/emotion control via natural-language tags, multi-speaker and multi-turn generation, with CLI, WebUI, and API server deployment options.

## Use cases
- generate natural speech from text
- clone a voice from a short audio sample
- build a multilingual TTS API server
- add expressive emotional narration with tags like [laugh] and [whispers]
- generate multi-speaker dialogue audio
- self-host a text-to-speech service with Docker

## When to choose
- you need state-of-the-art TTS quality with low word error rates
- you want fine-grained emotion and prosody control in generated speech
- you need multilingual speech generation across ~50 languages
- you can run inference on a 24GB GPU and want self-hosted deployment

## When to avoid
- you need a permissively licensed model for unrestricted commercial use (custom research license applies)
- you only have CPU-only or low-VRAM hardware
- you need real-time on-device TTS for mobile or embedded devices

## Facets
- artifact type: library
- maturity: active
- function: tts, speech-recognition, machine-learning, llm-inference, audio-processing
- domain: speech-processing, deep-learning, artificial-intelligence
- platform: python, cross-platform
- tags: voice-cloning, text-to-speech, zero-shot-tts, multilingual, emotion-control, vqgan, autoregressive, audio, linux, docker, gpu

## Member repositories
- fishaudio/fish-speech (main) score 74

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:57.549326+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:52:06.719683+00:00, confidence not recorded.
  - readme: https://github.com/fishaudio/fish-speech (fetched 2026-08-28T04:11:57.549326+00:00, sha 45e8bd53481c)
  - homepage: https://speech.fish.audio (fetched 2026-08-29T07:48:57.785630+00:00, sha 40aaf2d34adb)
  - site_page: https://speech.fish.audio/install (fetched 2026-08-29T07:48:57.788853+00:00, sha 6f9940204184)
- Data as of 2026-08-30T08:39:29.467469+00:00.