Podcast Agent

Inspiration

when I go to the gym, I queue up podcasts from the 400 podcasts that I follow on spotify. I thought an agent could sort out the signal from the noise; give me more coverage and create discovery opportunities.

What it does

given a list of podcasts and some learning goals, the agent will download episodes, transcribe them, semantically query and present clips.

How we built it

Python, OpenAI Whisper, Listen Notes, Supabase and GPT Index library, WhisperX for Python.

Challenges we ran into

We only got about 70% of the way to the demo. Its just a matter of more hacking.

Accomplishments that we're proud of

The pipeline is working great and we have transcribed 10 episodes from my current listening queue and currently indexing for semantic search!

What we learned

Getting our hands on these tools for a practical application, and getting the pieces connected was great. We closed the gap on some questions we had about transcription, indexing and speech/text alignment.