Why I want podcast transcriptions
I often find myself listening to podcasts about homelessness, addiction, or other topics that relate to my work. A lot of times, I want to highlight parts of these podcasts to share in conversations or to chew on later to sharpen my thinking on the subject. To do that I prefer having the text available.
Easy and costly options
Just pay for Snipd or Matter
- Snipd is a podcast app that works fine, but it’s not cheap. On iOS, you’re looking at $12 a month.
- Matter is a read-it-later app that you can send your podcasts to. It’ll take a bit more effort than Snipd, but it’s easier on the wallet, starting at $8 a month.
Just pay for Reader and hope the podcast is on YouTube
Reader by Readwise is great for many reasons. It’s my go-to for read-it-later and general online-knowledge-management help. When it comes to podcast transcripts, they can pull in YouTube videos and grab the transcript. So if your podcast has a YouTube version, you’re golden - just import it to Reader and you’re set. Reader will set you back $9 a month on iOS. On Reddit, they’ve hinted that podcast transcription support is coming down the pipeline, so fingers crossed for that.
Using Reader with YouTube podcasts is by far my preferred method.
Kind of cheap, little scary, but useful!
If a podcast isn’t on YouTube, here’s my backup plan:
Get the transcript
Option 1: Use an app like MacWhisper, Aiko, or Vibe to get a free transcription of the audio file. You’ll end up with a txt file, but heads up - it won’t be neatly formatted into paragraphs. The punctuation is often good but not perfect. If you’re cool with reading it as-is, you can stop here. If you want it more readable, we’ll need to rope in another AI to polish it up.
Option 2: Use the Apple Podcast transcription. Just highlight all of it, right-click > services > open.
Run the script to create an epub
The transcript is accurate, but it’s not formatted well enough to read easily. So I run a Python script to clean it up. That sounded intimidating to me at first, but it’s actually pretty simple:
- Make sure Python is installed on your computer. If it’s not, go ahead and install it.
- I’ve got a script that Claude 3.5 Sonnet helped me create, you can find it on my GitHub. Download it and make sure to add your own Claude API key where it says [insert API key] at the beginning.
- In your terminal, navigate to where you put the script, drop the transcript .txt file there (name it something like podcast.txt), create an output folder, and then run the script. Make sure you’ve got the dependencies installed first. On my Mac, I run the script with this command, but specify the input file name and output directory name:
python3 podcast-transcript.py [input_file] [output_directory] - The script should do its thing and then ask you what title and author you want for the epub.
- You’ll end up with 2 usable outputs: combined_context.txt and whatever you named your epub. If you get stuck with any of this like I did, asking an AI is your best bet! I like getting an epub because it plays nice with Reader.
