First published: 2020-06-17. Last updated: 2020-06-20.
Podcasts are one of my favourite ways to take in ideas and information. It is so convenient to be able to listen to them while doing other tasks and I find that arguments and facts are made all the more salient when delivered by an engaging podcaster.
However, podcasts have several downsides:
It is difficult to search through a podcast to find a particular section or phrase
It is difficult to highlight or take notes that refer to a particular section of a podcast
Reading is generally faster than listening, especially when re-reading or scanning through text
The obvious solution is get the best of both mediums by generating transcripts of podcasts. Having the transcripts of my favourite podcast episodes would make it much easier to reference and links together ideas from the different podcasts and facilitate the synthesis of overall thoughts on a topic covered in the podcasts.
Notes, links and tags on transcripts can make it easier to learn from podcasts
Some podcast players have started offering podcast functionality to some extent:
However, as far as I can tell, no podcast application is offering transcription of any podcast that a user wants. In all likelihood, any podcast application that is offering transcriptions will do so only for the most popular podcasts due the cost of executing the speech-to-text transcription itself. Furthermore, even if a podcast player were to offer this feature, I would want transcriptions to be available to all, not just to users who have access to a particular platform and not have access be dependant on the app provider tracking and profiling the user.
My "minimum-viable" vision for Podscribe is simple. Users enter a podcast title and an episode name and then click a button to have the podcast transcribed. They can they copy-paste the transcript into whatever text editor or note taking app that they prefer.
Speaker identification (diarization)
Export to note app (Google Docs, Notion, Evernote, etc)
Allow for file upload
Technically speaking, the largest difficulty in making Podscribe a reality is the actual task of converting the audio into text. Luckily, speech-to-text technology has progressed quite a bit and all of the major cloud providers (AWS, Azure, Google Cloud) are offering auto transcription for approximately $1 USD per hour of audio. Financing only the podcasts that I listen to and want to analyze further should be manageable for my personal piggy bank 🐖 however to transcribe any podcast that anyone wants, a source of funding would need to be found. However, on the bright-side, once a transcript has been generated, it could be served to any user that wants it!
A nifty approach could be to ask folks who want to see a podcast transcribed to pay some amount towards the transcription and then the transcription could be generated once the cost is reached. (Sounds like an opportunity for blockchain 👀)
Podcast creators probably retain most of the legal control over their content, so it is possible that they would want to be compensated for any revenue generated by PodScribe (though there are no plans for that currently) or may even not want the transcript of their podcast made available.