OMALEXIA 
█████  █   █   ███   █      █████  █   █  █████   ███
█   █  ██ ██  █   █  █      █       █ █     █    █   █
█   █  █ █ █  █████  █      ████     █      █    █████
█   █  █   █  █   █  █      █       █ █     █    █   █
█████  █   █  █   █  █████  █████  █   █  █████  █   █

Beta A dyslexia-first profile for Omarchy Linux · by Michael de By | The Fresh Office

Linux that reads to you, listens to you, and stays calm.

We are building the best operating system for dyslexic people, and for everyone whose eyes get tired before their day does.

Omalexia layers three things on top of Omarchy: a voice that reads anything, ears that type what you say, and a screen that gets out of the way. All of it runs on your own machine. Nothing leaves it.

Install the beta from GitHub →
F9 talk
F10 listen
Super+R reading mode

The goal

Make the best operating system for dyslexics and alike: not an accessibility setting buried in a menu, but the default.

Most computers treat reading as free and speaking as a feature. For a dyslexic person it is the other way round: decoding text costs effort every minute of the day, and the voice is the fast path. Omalexia starts from that. Hearing text, saying text and seeing text calmly are first-class, one key each, and they work in every app, the terminal included.

Credit where it is due

Omarchy already does half the work.

Before we added a single line, the Omarchy team had already built something that helps dyslexic people every day. Tiling windows mean nothing overlaps and nothing hides: every window has its place, the keyboard moves you between them, and the screen stays tidy without effort. One consistent theme runs through every app. There is no clutter to decode and no hunting for the window you lost.

That focus is not only calmer, it is more fun. Omarchy is fast, opinionated and beautifully put together, and it makes the computer feel like a tool you own rather than a place you get lost in. Thank you, DHH and everyone who builds Omarchy. Omalexia exists because there was such a good foundation to build on.

What we hold important

Four things, in this order.

Local first

▶ The best text to speech

Highlight anything, or just point at it, and press F10: you hear it within a tenth of a second, and the word being spoken is highlighted in the text while you listen. Web pages, mail, PDFs, code, a screenshot of a slide: if it is on the screen it can be read. The language is detected from the text itself, and you can add a voice for any language you read.

  • Neural voices running on the laptop, no cloud, no account
  • The same voice everywhere: browser reader mode, screen reader, any app that talks
  • Stop, speed up, slow down without looking
Local first

● Dictation that just works

Hold F9, say it, let go: the words land where your cursor is, with punctuation, in Dutch or English, in a terminal or a text box alike.

  • A tick you can hear when it starts and stops listening
  • Your own word list: “fresh office” becomes “The Fresh Office”
  • Filler words gone, first letter capitalised, nothing sent anywhere
Calm by default

▭ A screen you can read when tired

Letters that wobble at 22:00 are not a character flaw; they are a layout problem. When you want to keep working we have you covered: lines of a readable length, the window you are reading centred and the rest dimmed, a warm paper tint instead of glare, and letterforms that stay apart.

  • Reading mode on one key, narrow single windows on wide monitors
  • A distinct, plain typeface as the desktop default; dyslexia fonts available, never forced
  • Bigger text everywhere with one setting, reduced motion when you want it
In the open

▤ Choices you can check

Every choice is measured before it is made and written down afterwards: which voice starts fastest on a CPU, which fonts the research actually supports, why line length matters more than font shape. As we build, the how and the why land here.

  • Built on Omarchy, installed as one opt-in profile
  • Everything reversible with one command
  • Source, notes and numbers on GitHub

How we work

Experience first, then optimise.

Measured, not assumed

A voice that takes a second to start is a voice you stop using. We time everything on the machines we actually carry and pick what is instant, then offer the richer options as a choice.

One key, everywhere

No app-specific plugins to remember. Reading, dictation and reading mode are keys on the keyboard and buttons in the bar, and they work in the same way in a browser, an editor and a terminal.

Yours to keep

Speech in both directions runs on your own computer. That is faster, it works on a train, and what you read and say stays with you.

Open source, open door

Help make this the best accessibility add-on for Omarchy.

Omalexia is open source under the MIT licence, and everyone who wants to help is welcome: dyslexic readers who can tell us what actually works, people who know voices and speech models, people who care about fonts and layout, translators, testers, and anyone who just wants Omarchy to be kinder to more people. Ideas, bug reports and pull requests all count.

The code, the notes and the numbers are in one repository. Start with an issue if you are unsure; start with a pull request if you are not.

Read the code → Open an issue →

Try it

Now in beta testing.

Omalexia has left the announcement stage: it runs all day on our own machines, and it is ready for more hands. Beta means the core works, the edges are still rough, and your bug reports decide what gets fixed first. On an Omarchy machine:

$ git clone https://github.com/thefreshoffice/omalexia.git
$ cd omalexia && ./install.sh

Safe to re-run; omalexia off puts the visual changes back.

Installation instructions on GitHub → Report what breaks →

More as we build

What lands here next.

  • NowBeta testing: Omalexia runs daily on our own machines. Install it from GitHub and tell us what breaks.
  • NewThe speech research: every engine we measured, what a native listener said, and why Dutch is the hard case.
  • NewThe changelog: what changed, and when, in plain words.
  • NextHow it works: the read-aloud daemon, the dictation pipeline, the reading layout, and what each key does.