Open source · KDE Plasma · Wayland

Hold. Speak. Release.

Push-to-talk dictation for Linux. What you say is typed into the window you are in. No app to open, no window to switch to.

InstallView on GitHub
Message

Write a message…

or holdCtrlAltD
The real HUD, drawn by the same code İmlâ uses on your desktop. The voice here is simulated.
Why

Dictation should be part of the system, not another app.

I built İmlâ for myself. There are plenty of dictation apps, but each one is a window to open and keep in mind. İmlâ runs as a small user service: a global shortcut, a short sound and a small HUD at the top of the screen while it listens. When you are not speaking, it is not there.

Muhammed Emin Akalan
Transcription

On your machine, or in the cloud.

On your machine

0.9 s
for 27 seconds of speech on an RTX 5060 Ti

whisper.cpp is built in. The model runs on your GPU through Vulkan and falls back to the CPU. Nothing leaves your computer.

Whisper large-v3-turbo · 547 MB · offline

In the cloud

Any model
that accepts audio on OpenRouter

Use any audio-capable model on OpenRouter with your own key. The default is GPT Audio Mini, the best price and accuracy in our Turkish dictation test.

OpenRouter · your API key · pay per use
Details

Small things that matter every day.

Hold or tap
Hold the shortcut while you talk, or tap once to start and again to stop. The default is Ctrl+Alt+D.
Any language
The language is detected per clip and mixed speech is kept as spoken. You can also pin one.
Spelling hints
Give it the names and terms you use, and they come out spelled right.
Your clipboard stays
İmlâ pastes through the clipboard, then puts back what was there before.
A HUD that follows your voice
A field of dots at the top of the screen while it listens, a soft flow while it writes. Short tones on start and stop.
Quiet when idle
No window, no tray app. The HUD draws nothing until you press the shortcut.
Install

Arch Linux, KDE Plasma 6, Wayland.

The package is built in Docker, so no Rust or C++ toolchain lands on your machine. You need Docker and base-devel.

  1. 01Build and install the package
  2. 02Choose a provider, model and shortcut
bash
$ git clone https://github.com/muhammedeminakalan/imla && cd imla
$ make package
$ sudo pacman -U packaging/arch/imla-*-x86_64.pkg.tar.zst
$ imla setup

Change anything later with imla setup. Check it with imla status.

Other distributions are not packaged yet. The source is on GitHub. GitHub