Multi-lingual large voice generation model, providing inference, training and deployment full-stack ability.
About this project
CosyVoice is an open-source speech and audio project on GitHub: Multi-lingual large voice generation model, providing inference, training and deployment full-stack ability. It covers speech and audio, the layer that turns text into voice, voice into text, or raw audio into something structured. It is maintained by QwenAudio and published as a public repository you can read, fork and run yourself.
On GitHub the repository is tagged audio-generation, cantonese, chatbot, chatgpt, chinese, cosyvoice, cross-lingual, english, fine-grained and fine-tuning, which is a fair summary of what it covers. Most deployments care about latency and voice quality, so the project usually ships both a batch path and a streaming or real-time one.
It is written mainly in Python and released under APACHE-2.0, a permissive license — commercial use is fine as long as copyright notices are kept. The last commit was on May 26, 2026, so development is moving, just not on a daily basis.
To try it, clone the repository and follow its README; that file is the canonical install guide. There is also a project page at https://funaudiollm.github.io/cosyvoice3 for documentation or a hosted demo. Because it runs on your own infrastructure, budget time for environment setup: dependencies, model weights and configuration usually take longer than the first successful run.
Key facts as recorded on GitHub: 24,435 stars; 2,789 forks; primary language Python; license APACHE-2.0; last commit on May 26, 2026; topics audio-generation, cantonese, chatbot, chatgpt, chinese, cosyvoice.