- Home
- Text-to-speech tools
- CosyVoice
CosyVoice
What it is
CosyVoice is an open-source multilingual text-to-speech system with capabilities for inference, training, and deployment. Fun-CosyVoice 3.0 is designed for zero-shot multilingual speech synthesis across nine languages and more than 18 Chinese dialects or accents, including multilingual and cross-lingual voice cloning. Users can guide language, dialect, emotion, speed, and volume, and adjust pronunciation with Chinese Pinyin or English CMU phonemes. It handles numbers, special symbols, and varied text formats without a traditional frontend module. Text-in, audio-out streaming is supported, with stated latency as low as 150 ms. The repository documents Conda installation using Python 3.10, pretrained model downloads, and Docker deployment with gRPC or FastAPI, plus NVIDIA Triton and TensorRT-LLM acceleration. FastAPI and gRPC servers and clients are included in the deployment instructions. The software is Apache License 2.0 licensed; the repository characterizes its content as intended for academic use and to demonstrate technical capabilities. The documented optional ttsfrd normalization wheel is for Linux x86_64.
For and against
For
- Apache License 2.0 repository.
- Supports voice cloning and nine languages.
- Text and audio streaming has latency stated as low as 150 ms.
- Deployment options include Docker, gRPC, and FastAPI.
Against
- Documented installation uses Python 3.10.
- Optional ttsfrd wheel is specified for Linux x86_64.
- Repository describes intended use as academic and demonstrative.
The verdict
CosyVoice offers multilingual synthesis, pronunciation controls, voice cloning, and deployment paths for self-managed use. Check its documented runtime requirements and stated intended use against your project before adopting it.
Who it suits
CosyVoice may suit developers or researchers seeking a self-managed speech synthesis system with API and deployment options. The repository frames its content for academic purposes and demonstration of technical capabilities.
Plans and prices
Open-source software
Apache-2.0 licensed repository · self-managed installation and deployment
Read 3 Oct 2026 · github.com
On the record
- API access
- Yesgithub.com · read 20 Sep 2026
- API interfaces
- The deployment instructions include FastAPI and gRPC servers and clients.github.com · read 3 Oct 2026
- Commercial use
- Yesgithub.com · read 20 Sep 2026
- Deployment options
- The repository documents Docker deployment with gRPC or FastAPI, as well as NVIDIA Triton and TensorRT-LLM acceleration.github.com · read 3 Oct 2026
- Export formats
- WAVgithub.com · read 20 Sep 2026
- Installation
- The repository documents installation with Conda and Python 3.10, and offers pretrained model downloads through ModelScope or Hugging Face.github.com · read 3 Oct 2026
- Intended use
- The repository says its content is for academic purposes and to demonstrate technical capabilities.github.com · read 3 Oct 2026
- Languages and dialects
- Version 3.0 covers nine languages and more than 18 Chinese dialects or accents, with multilingual and cross-lingual zero-shot voice cloning.github.com · read 3 Oct 2026
- License
- The repository is licensed under Apache License 2.0.github.com · read 3 Oct 2026
- Pronunciation control
- The system supports pronunciation inpainting with Chinese Pinyin and English CMU phonemes.github.com · read 3 Oct 2026
- Purpose
- CosyVoice is a multilingual text-to-speech system that provides inference, training, and deployment capabilities.github.com · read 3 Oct 2026
- Runtime requirements
- The documented installation uses Python 3.10; the optional ttsfrd normalization wheel is specified for Linux x86_64.github.com · read 3 Oct 2026
- Streaming
- It supports text-in and audio-out streaming, with latency stated as low as 150 ms.github.com · read 3 Oct 2026
- Support
- The maintainers direct users to GitHub Issues and an official Dingding chat group for discussion.github.com · read 3 Oct 2026
- Text normalization
- It supports reading numbers, special symbols, and varied text formats without a traditional frontend module.github.com · read 3 Oct 2026
- vLLM compatibility
- The repository says CosyVoice 2 and 3 support vLLM 0.11.x or newer and vLLM 0.9.0, while versions between those releases are untested.github.com · read 3 Oct 2026
- Voice cloning
- Yesgithub.com · read 20 Sep 2026
- Voice instructions
- Users can provide instructions for language, dialect, emotion, speed, and volume.github.com · read 3 Oct 2026
