fmiotello/dpai ? reverse-engineered prompt

Build me a Python project for deep prior audio inpainting.

I want a simple tool that can take an audio file with missing or damaged parts and reconstruct the gaps as naturally as possible. Make it usable from a script or small app, with clear inputs for a source audio file, a masked section, and an output file. Include whatever training or inference code is needed, plus a basic example so I can try it quickly.

If it helps, look up current docs online for any audio processing libraries or best practices you need. Keep the code clean, easy to run, and give me a minimal way to test the model on a sample audio clip.