PaperPod - Turn any document into an interactive AI podcast

@arindemirevm1u You can share, download podcast audio and transcript script with proper speaker diarization (Host, Guest). Also there is live transcript support when the podcast audio plays. You can click on any line to jump in the audio.

Just Click on Show Transcript

From UI's perspective should I make this more clear?

@m_mertcan52889 

Thank you! That's exactly what I was aiming for.

My goal wasn't to build another TTS reader—it was to recreate the feeling of listening to a podcast where you can naturally pause and ask questions whenever something isn't clear.

If AI is going to change how we learn, I think conversations are a much better interface than static pages.

Would love to know what you'd listen to first—research papers, books, documentation, or something else?

@hayriye29240360 Thank you! Commute learning is actually one of the biggest use cases I had in mind.

There are so many documents we want to read but never find uninterrupted time for. My hope is that waiting in traffic, walking, or commuting can become productive learning time instead.

Is there a particular type of content you'd use it for during your commute?

@meral237026 
Thank you so much!

I found that most AI readers still feel like robotic narration. I wanted PaperPod to feel more like listening to a discussion where you're free to interrupt, ask follow-up questions, and immediately continue the conversation.

I'm really happy that came across in your experience.

@nazlbornovxsfy 

Thanks! 😊

That conversational feeling was one of the hardest parts to get right. We spent a lot of time refining the dialogue generation so it feels like two people genuinely explaining concepts instead of alternating paragraphs.

It's great to hear that it feels natural. That's exactly the experience I was hoping to create.

@tarkbjqy Really appreciate that, it means a lot coming from someone who actually tried it. :)

The biggest engineering challenge wasn't the AI itself, it was latency. I spent quite a bit of time optimizing the generation pipeline using adaptive concurrency, faster audio synthesis, smarter LLM routing, and retry mechanisms to get the total generation time close to a minute.

If you get a chance, I'd love for you to try it with something more challenging—a research paper or a document full of architecture diagrams—and let me know where it breaks. That's the kind of feedback I'm looking for.

@savas_teki53713 

Thank you! That was actually one of the core design goals.

Traditional text-to-speech often feels like someone reading paragraphs aloud. I wanted it to feel like you're listening to two people naturally explaining the material instead.

The interruption feature came from my own habit of constantly stopping while reading to ask "Wait... why does this work?"

Curious—what kind of documents would you personally use PaperPod for?

@ademhelvac59796 

That's an excellent suggestion.

The transcript already exists internally with speaker information, so adding timestamps and making it exportable is definitely feasible.

I also really like the idea of generating "conversation highlights" or key takeaways automatically after the podcast finishes. That would make revision much easier without replaying the entire episode.

Added to my roadmap. Thanks for the idea! 🚀