souzatharsis/podcastfy ? reverse-engineered prompt

Build me an open source Python tool that turns things like websites, PDFs, images, YouTube links, or just a topic into a natural sounding podcast style audio conversation.

I want to be able to paste in one or more sources, choose the language and vibe, and get back an audio file plus a transcript. It should feel like a mini NotebookLM for podcasts, but open source and easy to use from Python and from the command line. Please make it support both short episodes and longer ones, and let me customize the voices, style, and structure of the conversation.

If it helps, look up the current docs for the best API settings and text to speech options. It would be great if the project also includes a simple way to run it locally or through a small API, and if it can work with modern LLM and voice providers without too much setup.