The maker behind CosyVoice.org
I’m building CosyVoice.org to turn powerful multilingual speech research into a tool people can open, understand, and use.
Who's behind this?
I’m the maker who designs, builds, and maintains this site. I shape each part around one goal: making voice generation more approachable and easier to put to work.
Why did I build it?
I wanted to close the gap between an impressive open-source model and audio people can actually use. Runtime setup, voice and parameter choices, and turning an experiment into a repeatable workflow should not stand between an idea and its voice.
What was harder than expected?
One lesson kept coming back: successful synthesis is not the same as a workflow that feels reliable. Voice selection, multilingual text, cloning inputs, generation time, and failures all need to make sense, with a clear next step at every point.
What am I proudest of?
I’m proud that anyone can try the core experience before creating an account, then move from a short experiment to more advanced voice workflows without first learning the infrastructure behind them.
What's the mission?
My mission is to make expressive, high-quality speech tools easier to explore and use across languages, while staying clear about what comes from open-source research and what this service adds.
Open source at the core
This service is built around the open-source CosyVoice model, released under Apache-2.0. I did not create the underlying model. CosyVoice.org provides the hosted access, product workflows, operational care, and practical enhancements around it.
Let’s keep the conversation open
Have a question, feedback, or a problem with the service? Email me directly or send a message.