openbmb/voxcpm ? reverse-engineered prompt

Build me a clean web app for VoxCPM2 that lets me type text and turn it into realistic speech. I want the main flow to be simple, with options for multilingual text to speech, voice design from a short description, and voice cloning from a reference clip. It should let me play the result in the browser and download the audio, and the output should sound like high quality studio audio.

Make the interface easy for non technical users, with clear labels, a few helpful examples, and sensible defaults. If it makes sense, include a basic advanced section for things like generation strength, speed, and number of steps. Also add support for choosing a language and uploading a reference voice sample when cloning.

If you need current setup details or model usage guidance, look up the latest docs online first, then wire it into the app so it works end to end.