Listen to anything you can highlight.
Select text. Press Meta+R. Hear it in a natural neural voice.
Local text-to-speech for Linux — no accounts, no cloud, no telemetry.
Get Lepramim · Why Lepramim · Quick start · Everyday use · Voices · Help
Turn a long article into a listening session, work through a paper, or hear your own writing back. Lepramim reads the selection from the app you're already using: a PDF reader, browser, editor, or anywhere you can highlight text.
| What you want | What Lepramim gives you |
|---|---|
| Listen without switching apps | One shortcut to speak the selection, another to pause or resume. |
| Keep your text private | Kokoro-82M generates speech on your hardware. After the one-time model download, reading works offline. |
| Make dense text easier to hear | Optional cleanup removes citations and Markdown, expands abbreviations, and reads numbers and math symbols as words. |
| Find a voice that fits | 54 voices, 9 language options, and playback speed from 0.50× to 2.00×. |
| Keep controls close | A tray icon shows engine status; an optional floating overlay puts pause, sentence navigation, and stop over any window. |
| Start listening quickly | One portable executable, models downloaded from the app, and optional startup at login. |
-
Download the latest
Lepramim-*-x86_64-portable.runfrom Releases. -
Make it executable and run it:
chmod +x Lepramim-*-x86_64-portable.run ./Lepramim-*-x86_64-portable.run
-
On first launch, a welcome window downloads the speech models (~340 MB) with a progress bar. You can Continue (the download runs in the background and Lepramim becomes available as soon as it finishes) or Skip and download later from the Models tab.
-
Highlight a sentence in another app and press Meta+R on KDE Plasma. On other desktops, bind your preferred shortcuts first. Use Meta+P to pause or continue.
That's the whole install — no system services, no configuration files to edit. Downloads are verified by checksum, and models are kept between app updates.
Opening Lepramim a second time never starts a duplicate: the running copy just brings its control window forward.
- Listen: highlight text anywhere, then press Meta+R (or right-click the tray icon and choose Speak highlighted selection). On Wayland, Lepramim briefly presses Ctrl+C for you to capture the highlight; your existing clipboard content is never spoken by mistake — before-and-after clipboard snapshots are compared. If nothing is highlighted, a notification tells you to select text first.
- Pause / resume: Meta+P, the tray menu, or the floating overlay.
- Stop: the tray menu or the overlay's stop button.
- Tray at a glance: dim blue means stopped, breathing blue means warming up, solid green means speaking (it stays green while paused so you know where you left off).
- Change how it sounds: open the control window from the tray to pick a voice and language, adjust speed, audition voices with Test voice, and toggle the preprocessor cleanup options.
An always-on-top bar shows the current sentence with previous, pause, next, and stop buttons. It is off by default to keep Lepramim discreet; turn it on in the control window.
| Shortcut | Action |
|---|---|
| Meta+R | Speak highlighted selection |
| Meta+P | Pause / resume playback |
On KDE Plasma the shortcuts are registered system-wide and work everywhere.
On other desktops, Lepramim exposes an org.lepramim.App service on the
session bus (SpeakSelection / Toggle) so you can bind keys of your choice
in your keyboard settings.
Wayland note: capturing a highlight with one keypress needs one of
ydotool, wtype, xdotool, or dotool on your system (the portable file bundles
xclip/wl-clipboard for the clipboard itself). Without a key-injection
tool, just copy the text yourself (Ctrl+C) and press Meta+R — Lepramim
reads the clipboard content.
Choose a voice in the control window, adjust the speed, and use Test voice to hear a sample before reading. The bundled catalog covers:
| Language option | Voices |
|---|---|
| English (American) | 20 |
| English (British) | 8 |
| Spanish | 3 |
| French | 1 |
| Hindi | 4 |
| Italian | 2 |
| Japanese | 5 |
| Portuguese (Brazilian) | 3 |
| Chinese (Mandarin) | 8 |
See the full voice catalog and model details for voice IDs, download locations, and model licensing.
- A Linux desktop with a system-tray host (StatusNotifier). The app refuses to start without one and tells you so.
- A graphical session (X11 or Wayland).
- The x86_64 portable release targets glibc 2.35 or newer.
Qt, audio libraries, and clipboard helpers are bundled in the portable file. On Wayland, automatic selection capture also needs a key-injection tool; copying the text yourself works without one.
- No tray icon: your desktop has no StatusNotifier host running. Start one (e.g. Plasma's system tray, GNOME's AppIndicator/KStatusNotifierItem extension, or your bar's StatusNotifier module) and relaunch.
- Meta+R does nothing: on KDE, check that no other action grabbed Meta+R.
Elsewhere, bind your preferred keys to the
org.lepramim.Appbus service in keyboard settings. On Wayland without a key-injection tool, copy first. - “Select text first” notification: nothing was highlighted and the clipboard held no new text. Highlight (or copy) some text and try again.
- No sound: check your system volume and output device, then stop and restart the engine from the tray menu.
- Speech never starts after an update: open the Models tab in the
control window. If a file shows missing, press Download missing models.
If trouble persists, quit the app, delete
~/.cache/lepramim/models, and relaunch — the welcome window will offer the download again. - Something else? The engine log lives at
$XDG_RUNTIME_DIR/lepramim/daemon.log; include the relevant lines when asking for help.
Lepramim works out of the box. Every setting is stored as plain text at
~/.config/lepramim/config.toml, and the control window edits it for you.
Speech models live in ~/.cache/lepramim/models/ — outside the app on
purpose, so reinstalling or updating Lepramim never re-downloads them.
- Lepramim is MIT-licensed — see
LICENSE. - The Kokoro-82M model and voice pack downloaded by the app are Apache-2.0
(details in
docs/models.md); Lepramim uses them unmodified.
If you want to build Lepramim yourself, see
docs/building.md for system dependencies and the build,
package, and smoke-test commands.
Have an idea or found a bug? Open an issue with your desktop environment and the relevant engine log. If Lepramim helps you read more comfortably, a star helps others find it.