First project: Speakerine
Turning your reading list into a personalised podcast.
I almost forgot about introducing my first project! 🤦♂️
I have started working on Speakerine: an application to turn your reading list into a personalised podcast.
This is what its landing page looks like right now: speakerine.app
The concept
The idea is quite simple:
Through my days, I often stumble upon articles I want to read. Since I usually don’t have the time to read them right away, I keep the tab opened “for later”… but later never comes
On the other hand, I have multiple occasions when I could listen to some content instead of reading it.
While commuting to work, exercising, cooking, doing the dishes or my laundry, while cleaning, … there are actually many opportunities to be listening to content while our hands are busy with routine tasks.
So I thought: why not create a Pocket-like app that allows me to easily transform any article into its audio transcription, and organise that into a playlist?
I kept this idea in my ideas list for a while, and since it kept creeping back to me as something I felt would be nice to do and valuable, I decided to build it!
I thought about many different features that could make it a daily companion for podcast aficionados, but I guess I’ll have to do it one step at a time :)
The name
Why Speakerine?
I wanted an original name which could be directly related to the text-to-speech nature of the project. I think Speak/Speaker in “Speakerine” works well for that, even though the word Speakerine itself might not be very well known internationally.
It’s actually the word “Speaker”, turned to a feminine form using the French suffix -ine. Speakerines were continuity announcers on the French television stations, appearing between programs, during unwanted interruptions or at the end of the day.
What I already know
My target is a mobile app – it seems to me that would be the best way to make it a daily tool that can be used on the go.
I’ll build the MVP as a web application though, as it’ll allow me to go faster to something I can put in your (and my ^^) hands!
With that objective in mind, I’ll start with Ruby on Rails as a framework to get things done quickly.
I know what the main technical challenges are, and (mostly) how to solve them:
Challenge #1: given a URL, extract the meaningful content
This is what the “reader mode” of your browser does, and some of those readers are open source. I might have to customise the output of such tools to better prepare it for Speech-to-Text processing.
Content behind paywalls or login/password will be a specific challenge I’ll have to figure out.
Challenge #2: turn this content into a quality audio transcription
Whisper or similar models are the most obvious approach
Using a language detector will probably be useful as well, to inform the model of the language it should use
What needs figuring out
There are so many things I still need to figure out, but the two most impacting areas I believe are:
Marketing & distribution:
I need to find where I can reach people who would be interested in the app, and how to target them
I also need to figure out SEO/ASO: what keywords will bring traffic to my app’s website or App Store listing?
Pricing:
Can I do something else than subscription, like credits or pay-as-you-go?
I’m thinking it will highly depend on:
Speech-to-Text performance and what kind of server I’ll need to rent to make it work efficiently
How varied are transcribed articles in terms of length per article and per user
Next steps
The other day I’ve installed both Ruby on Rails and RubyMine on my computer, for the first time in like… 12 years 😱
I’m pretty excited to go back to RoR honestly, but after that I was like: “now what?”. And I figured I had not yet thought through what I wanted an initial prototype or MVP to look like.
So I closed RubyMine, got a pen and paper out, and jotted down some initial wireframes of the minimal web app to get things running.
This is what I’ll work on for Techtrails’ next issue!
This project sounds cool, have you found customers who would be interested? Can the AI read the text in a way that sounds as interesting as a podcast?