Integrate podcasts into your second brain with speech-to-text
A short experiment on getting the most out of podcasts.
As set out elsewhere, I recently preordered David McRaney’s How Minds Change after listening to three recent episodes of his You Are Not So Smart podcast on my treadmill as I tried getting myself into shape for hiking around Lapland (if you want my place at his workshop, check out my post).
But there was a problem. Since 2017 I’ve been getting the most value out of the best podcasts by using their transcripts during my annotation process, massively improving my retention of what I hear:
Listen & Learn: how to absorb podcast knowledge
I’ve been listening to podcasts for ten years. But something’s always nagged me. How much of what I hear do I actually…
But there were no transcripts for these episodes, so I experimented with speech-to-text engines to create my own. TL:DR; Welder’s excellent and free; Otter.ai’s got some amazing features, but it’s pricey; and just forget Google.
I started with Google’s transcription service. A mistake: while it’s almost free (processing two 50minute podcasts a month would cost ~$12 a year), Welder (below) is free, while Google’s transcripts (here’s a .txt export of episode 232 and a .csv of episode 230) were underwhelming compared to both Welder and Otter.ai. The interface isn’t easy for non-developers, either.
Otter.ai: wow! (also: pricey!)
Otter.ai (referral link) blew me away, so I made a short video of me using it to explore and annotate edition 232:
After uploading the .mp3 file I was presented with the transcription in the main pane. David’s words were already separated from those of his interviewee, Adam Grant (and his interviewee, Margaret Atwood), so in three clicks I assigned a name to each voice.
Clicking any word plays the audio from that point (pause or change playback speed using the player, bottom), with each word highlighted as it is spoken. Selecting a word or phrase also opens an inline editor allowing you to highlight, comment or create a ToDo about it. These annotations appear in the “Meeting Gems™” sidebar, right. You can also add an image, edit the text, copy it to your clipboard or share it.
Otter.ai’s main purpose is to help teams get more value out of meetings, so it integrates with your calendar and online videoconferencing tools, albeit only Google Meet in the free plan, which is limited to 3 uploads of max. 30 minutes each per month (hence my transcription only captured the first 30 minutes).
The free plan is therefore not generous enough to annotate podcasts, while the Pro plan ($100/y) is overkill: I don’t actually need “Meeting Gems”, as I’m annotating podcasts directly to my Hub using the MyHub bookmarklet.
tools like Otter.ai could transform audio into useful social objects
But Otter.ai blew my mind by showing me how tools like this could transform audio files into social objects. Imagine if podcasts came packaged in a public, Otter-like container, allowing everyone to make both private or public notes in the sidebar as they read and listen, and comment to other users’ comments. Communities would emerge around individual episodes, podcasters and entire channels, with hashtags helping users discover relevant content from new publishers. The feedback in these notes would help podcasters answer questions and identify new topics to explore, while users sharing their notes socially would expand the podcast’s audience.
All this is probably too ambitious to include in my upcoming book chapter on Personal and Social Knowledge Graphs, but what the hell, it’s going in ;)
PS Although it’s expensive, if enough people sign up for Otter.ai using my referral link, I’ll use my free month of Pro to transcribe some You Are Not So Smart episodes, which I’ll pass to David for publication on his site.
Welder/Veed: free and fit for purpose… for now
Welder is a complete package for podcasters, its Transcription Studio is functional and free… and it’s migrating to Veed.io.
Like Otter, Welder separates the different speakers, as well as letting you add timestamps. While it doesn’t have Otter.ai’s bells and whistles, the transcription itself seems pretty accurate, and it’s free. What’s not to like?
Judge for yourself: check out my exports (.txt, .srt) of episode 231, or explore the 231 transcript on Welder itself (requires free sign-up).
Update: I received an email literally as I was polishing this post. Welder is shutting down, and moving its users to www.veed.io/. The Welder transcription service is still free and will be integrated into Veed, which aims to improve on it and keep it free.
If you’re into podcasts, then you will probably enjoy the launch workshop of David McRaney’s new book. It’s open to those who pre-ordered it, but I can’t make it so I’m giving my place away via a Twitter competition.
I’m also writing a chapter of a book on Personal Knowledge Graphs — check out my “brain dump first draft” introduction, and help me make it better: