RVC-Project/Retrieval-based-Voice-Conversion-WebUI ? reverse-engineered prompt

Build me an easy voice conversion app with a simple web interface, so someone can train a custom voice model from a small amount of clean speech, ideally around 10 minutes or a bit less, and then use it to change spoken or sung audio into that voice. I want it to handle both training and inference in one place, with options to upload audio, pick a pitch extraction method, and get good results without obvious voice leakage. It would be great if it also supports real time voice changing with low latency, plus vocal and instrumental separation for preparing datasets or covers.

Please make it practical for regular users on Windows first, but keep Linux and Mac in mind too. It should work with different kinds of GPUs when possible, including Nvidia and other common options. Include the extra model downloads and ffmpeg setup the app needs, sensible launch scripts, and a basic API if the repo already supports that. If anything is unclear, look up the current docs online and wire it up so it actually runs end to end.