hayeong0/DDDM-VC ? reverse-engineered prompt

Build me a voice conversion project that can take a source speaker recording and a target speaker recording, then convert the source voice to sound like the target while keeping the speech content intact. I want to be able to train it on my own dataset if I have paired wav files and F0 features, and also run inference with the provided sample commands and checkpoints.

Please wire up the preprocessing so it can extract and use F0 features, load the file lists for train and test, and save converted audio into an output folder. It should also support the pretrained model setup from the README, including the main model, vocoder, and F0 VQ VAE checkpoint.

If anything is unclear, check the current docs online and make the workflow easy to run from the command line.