Your own custom conversational AI podcast with two hosts that you can interrupt to ask questions in real time

Unlike traditional text-to-speech, what if instead of one AI reading your document aloud, you had two AI hosts having a natural conversation about it?

For example: Host: guides the discussion and asks questions and Expert: explains concepts, gives intuition, examples and analogies.

The interesting part is that you can interrupt the conversation at any point and ask your own questions, such as:

- "Give me an example."

- "Can you explain that again?"

- "Compare this with another approach."

The podcast then continues from where it stopped.

There could be two retrieval modes:

📄 Document only — answers come strictly from the uploaded document.

🌐 Document + Web — answers combine the uploaded document with relevant current web information.

I found myself wishing for something like this while reading long research papers, design docs, RFCs and study material. Listening is often easier than staring at a screen for hours.

I ended up building a prototype called PaperPod to explore the idea.

I am not promoting, I genuinely want feedback from people who read a lot of documentation.

PaperPod: https://paper-pod-one.vercel.app/

If anyone is curious, I'm happy to share the demo and explain some of the implementation (LLM orchestration, retrieval, voice generation, and real-time Q&A).