For the past few months, I've been working on a listening app, and I'd like to take you along for the journey — from where it all started to where it's going.
Through this series, I'll share the ideas, challenges, mistakes, and small victories that have shaped Orateur.
But before we dive into the technology, let's go back to the very beginning.
How did the idea come about?
Like many people, I subscribe to several newsletters and receive dozens of interesting articles every week. Every time I see one, I tell myself:
"I'll read it later."
So I open it in a new browser tab and move on with my day. The problem is... "later" almost never comes.
When I finally return to that tab, the article is often long, a notification pops up, I get distracted, or I simply lose the motivation to start reading.
Eventually, I close the tab without having read a single paragraph. After a few weeks, my browser turns into a graveyard of forgotten tabs. I'm pretty sure I'm not the only one this happens to.
Another challenge: articles in English
There was another issue as well. Many of the articles I wanted to read were written in English. Although I can read and understand English fairly well, spending twenty or thirty minutes reading a technical article requires much more mental effort than reading in French. After a while, it becomes tiring, and I often end up giving up.
So I started asking myself a simple question:
What if I could simply listen to these articles while doing something else?
The spark
With the rise of AI, one day I had the idea of asking it to read an article I was genuinely interested in. That very article eventually became the demo article available on the Orateur website.
Being Claude Code, it naturally went into full over-engineering mode. All that because at the time it couldn't read(no TTS feature built yet), so it built an actual web app to read that specific article. I quickly realized it could become much more than a fun experiment. So I began automating the entire workflow.
The goal was simple: paste an article's URL, fall back to copying and pasting its text whenever extraction wasn't possible, or import the content directly from PDF, Markdown, or plain text files. At that point, I was no longer building a proof of concept. I was building Orateur.
One idea led to another...
The project quickly kept growing. I no longer wanted to simply listen to articles. I wanted the whole experience to feel effortless. One URL, one click, a few seconds later, your article was ready to listen to. Little by little, that small web app evolved into a real application
What I'm most proud of
Surprisingly, importing articles isn't the feature I'm most proud of.
The real milestone was integrating an AI speech model capable of producing voices that sound far more natural than the default voices available on our phones and computers. And there's something even better.
It works completely offline.
Once the models are downloaded and installed, you only need an internet connection to fetch an article. Everything else---from text processing to playback---runs locally on your device. That was incredibly important to me.
I wanted Orateur to remain useful anywhere, even with a slow, expensive, or completely unavailable internet connection.
Give Orateur a try
If you're curious, you can already try the web version of Orateur:
https://orateur.digitalekim.net
The mobile app and a browser extension are currently in development.
What's next?
This is only the beginning. In this first article, I shared the story behind why I started building Orateur. But the most interesting part is still ahead.
In the next episode, We'll talk about how I went from the robotic text-to-speech voices built into our devices to AI-powered voices that sound surprisingly natural—while still running entirely offline.
Trust me... getting there was far from easy.