supertone-oss-archive/supertonic ? reverse-engineered prompt

Build me a simple local text to speech app using this project so I can type or paste text and hear it spoken on my device, without sending anything to the cloud.

I want it to support multiple languages, let me choose the language or use automatic handling when I am not sure, and include the expressive tags like laugh and breath if they are available. Keep it fast and lightweight, and make the audio export clean and ready to play back right away.

If there are good examples already in the repo, use them as the basis and wire up the end to end flow so I can run it locally with the archived model files. A small demo interface would be great too, as long as it is simple and works on a normal machine. Look up current docs online if you need to.