Paper to Podcast 2.0

Drop a link, and it turns into a two-voice chat that makes clear what it’s about and why it matters.

Comment

Inspiration

The inspiration comes from my difficulty understanding papers. It’s difficult to digest information quickly at first. We should at least know what it’s trying to solve and its real-world impact.

What it does

Paper to Podcast 2.0 transforms any research paper or article URL into a natural, conversational podcast featuring two AI voices discussing the content. Users simply:

  1. Click the circle and paste a paper HTML URL
  2. Wait while AI processes the content then will extracting key points, simplifying jargon, and creating natural dialogue
  3. Listen instantly

The app automatically:

  • Fetches and processes academic papers HTML URL
  • Simplifies complex technical content into understandable language
  • Generates a dynamic script with natural back-and-forth dialogue
  • Creates high-quality audio using distinct AI voices
  • Auto-plays the finished podcast for immediate listening

How we built it

We use Figma Make for fast prototyping and runtime testing. Once the core functionality is stable, we refine the code with other AI editors like Warp and Codex. My main approach is heavy iteration and reworking the code repeatedly until the app behaves exactly as we envision.

Challenges we ran into

  1. Latency. AI responses are tightly bound to infrastructure performance, and with limited resources, this was a bottleneck for further development.
  2. Longer papers are hard to process. We still don't know how to solve it, but it's our priority in the future.

Accomplishments that we're proud of

The app’s core workflow is already functional, even though there’s still a long list of improvements to make.

What we learned

Vibe coding requires practice to achieve best results quickly. We also realized better hardware is essential for faster local development, especially when working with large open-source models like gpt-oss-20b or gpt-oss-120b.

What's next for Paper to Podcast 2.0

We’re planning major upgrades for faster and more accurate conversions, user accounts, a database layer, audio streaming, improved evaluation metrics, and eventually pricing, if everything goes well.

Built With

Share this project:

Updates