# QwenAudio/CosyVoice

Multi-lingual large voice generation model, providing inference, training and deployment full-stack ability.

Repository: https://github.com/QwenAudio/CosyVoice
Canonical: https://ross.abutalabs.com/products/cosyvoice
Homepage: https://funaudiollm.github.io/cosyvoice3
Language: Python
License: Apache-2.0
License Family: permissive
Topics: audio-generation, gpt-4o, text-to-speech, tts, cantonese, chatbot, chatgpt, chinese, english, fine-grained, fine-tuning, japanese, korean, multi-lingual, natural-language-generation, python, cosyvoice, cross-lingual, voice-cloning
Last push: 2026-05-25T18:15:40+00:00

## Health v2 (maintenance only)
Score: 61/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 84, release rhythm 35, longevity 56
- inputs: {"age_days": 791, "days_push": 100, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 22925, forks 2637 (observed 2026-08-28T04:11:34.222166+00:00)

## What it is
CosyVoice is a multilingual large voice generation model (TTS) with full-stack inference, training, and deployment support. It offers zero-shot voice cloning, streaming synthesis with ~150ms latency, and instruction-based control across 9 languages and 18+ Chinese dialects.

## Use cases
- clone a voice from a short audio sample
- generate speech in Chinese, English, Japanese, Korean and other languages
- build a streaming text-to-speech service with low latency
- fine-tune a TTS model on custom speaker data
- control emotion, speed, and dialect of synthesized speech via instructions
- synthesize speech with correct pronunciation of numbers and symbols

## When to choose
- you need production-grade multilingual or cross-lingual TTS with voice cloning
- you want streaming audio output with very low latency
- you need Chinese dialect support or pronunciation inpainting
- you want to train or fine-tune your own voice generation model

## When to avoid
- you need a lightweight offline TTS without GPU resources
- you only need simple pre-recorded voice playback
- you require languages outside its supported set
- you want a fully managed cloud TTS API rather than self-hosted models

## Facets
- artifact type: library
- maturity: active
- function: tts, speech-recognition, machine-learning, llm-inference, audio-processing
- domain: speech-processing, artificial-intelligence, large-language-models
- platform: python, cross-platform
- tags: voice-cloning, text-to-speech, multilingual, zero-shot-synthesis, streaming-tts, fine-tuning, voice-generation, chinese-dialects, natural-language-processing, audio, gpu, linux

## Member repositories
- QwenAudio/CosyVoice (main) score 61

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:34.222166+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 2026-08-29T16:56:50.496774+00:00, confidence not recorded.
  - readme: https://github.com/QwenAudio/CosyVoice (fetched 2026-08-28T04:11:34.222166+00:00, sha 5e932c1da217)
- Data as of 2026-08-30T08:39:29.467469+00:00.