v0.2.1

Lepramim v0.2.1

Lepramim is a local text-to-speech app for Linux: highlight text anywhere — a PDF, a web page, an editor — press Meta+R, and hear it read aloud. Everything runs on your machine; no accounts, no cloud, no telemetry.

Highlights

• One-hotkey reading — Meta+R to speak the current selection, Meta+P to pause/resume; global shortcuts on KDE, bindable via the org.lepramim.App bus service elsewhere
• Private, local speech — Kokoro-82M neural voices run entirely on your hardware
• 54 voices in 9 languages — English (US/UK), Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese, Mandarin, with a 0.50×–2.00× speed slider
• Made for real documents — strips citations and Markdown, expands abbreviations, reads numbers and math symbols as words
• Single-file install — one AppImage; speech models (~340 MB) download once on first launch and are checksum-verified
• Tray-first design — status at a glance, start/stop from the menu, optional "start with desktop", and an optional floating playback overlay

Notes

• Requires a Linux desktop with a system-tray host (StatusNotifier); X11 and Wayland are supported. On Wayland, one-keypress capture needs ydotool, wtype, xdotool, or dotool — otherwise copy first, then press Meta+R.
• CPU-only by default in the AppImage; CUDA is available in source builds (--backend cuda12).
• See the README for setup, troubleshooting, and file locations.

Download: Lepramim-*-x86_64.AppImage below — make it executable and run it.