I listen to a lot of podcasts, especially while driving, which makes it hard to take notes. This notebook is a WIP environment to build a podcast summarizer. I will then feed these notes into my Obsidian vault.
Stack:
- Optional: ytdl to download the audio from YouTube
- OpenAI Whisper to transcribe it
- LangChain to split up the text and feed it to davinci
- Get the summary
- Bonus: Create a Twitter thread for social sharing
Instructions:
- Download the audio from YT with the ytdl cell by putting it in the URL list, or copy your audio file in
./tmp - Run the cells
- Find raw transcript in
podcast-raw-transcripts - Find clean transcript in
podcast-clean-transcripts
TODOs:
- Add Obsidian integration