Offline on Linux

Dictate into any app. Keep the audio home.

System-wide voice typing with local models. X11 and Wayland. One install command.

Install VocalinuxSourcemacOS / Windows
Offline
Local engines
Display
X11 + Wayland
Default
whisper.cpp
License
GPL-3.0
Vocalinux tray overlay showing live speech transcription on Linux
Live transcription overlayMore shots

Install in one command

Interactive installer detects hardware, lets you pick an engine, and wires the desktop app.

Ubuntu · Fedora · Debian · Arch · openSUSE

install.sh --interactive
$ curl -fsSL \
  https://raw.githubusercontent.com/jatinkrmalik/vocalinux/main/install.sh \
  -o /tmp/vl.sh && \
bash /tmp/vl.sh --interactive
Then launch: vocalinuxCompare enginesDistro notes

Built for real Linux desktops

Privacy, injection reliability, and engines that match your hardware, not a cloud dashboard with a mic icon.

App screenshots
Vocalinux settings dialog showing speech engine options
Engine and model controls in the settings GUI

Stays on your machine

Local engines process audio on-device. No cloud upload, no telemetry, no account.

Offline privacy

Works where you type

Terminals, browsers, IDEs, office apps, and any focused text field on X11 or Wayland.

Use cases

Fast by default

whisper.cpp with Vulkan on AMD, Intel, and NVIDIA. Tiny model is about 74MB.

GPU acceleration

Your shortcuts

Toggle or push-to-talk. Bind modifiers the way your hands already work.

Shortcuts

Desktop reliability

Suspend recovery, IBus hardening, layout preservation, non-ASCII injection fallbacks.

Reliability notes

Neural silence filter

Silero VAD drops empty buffers before recognition, with amplitude fallback if needed.

Voice activity detection
Vocalinux system tray icon and menu
Tray status while you work

The Linux voice gap, closed

macOS and Windows shipped system dictation years ago. Linux users got fragments. Vocalinux is the full desktop path: tray, hotkeys, injection, and local models.

  • No cloud dependency for local engines
  • Works across apps, not one editor
  • Guided install, not a research project
  • Open source you can audit

From install to first sentence

  1. 1

    Run the installer

    One curl. It pulls dependencies, models, and desktop integration.

  2. 2

    Pick an engine

    Default whisper.cpp, or Whisper, VOSK, or a remote server you control.

  3. 3

    Hold or toggle

    Activate the shortcut, speak, and text lands in the focused field.

What the installer does

  • Installs system dependencies
  • Creates an isolated virtual environment
  • Downloads speech models for your choice
  • Sets up desktop integration and PATH
  • Creates an application launcher

Uninstall

uninstall.sh
$ curl -fsSL \
  https://raw.githubusercontent.com/jatinkrmalik/vocalinux/main/uninstall.sh \
  -o /tmp/vul.sh && \
bash /tmp/vul.sh

System requirements

OS
Ubuntu, Debian, Fedora, Arch, openSUSE, or equivalent
Runtime
Python 3.9+ with GTK 3 / PyGObject
Memory
4GB minimum; 8GB+ for larger models
Display
Microphone plus X11 or Wayland session
Disk
~200MB for the default whisper.cpp setup

Guides by distribution

Distro-specific notes for Ubuntu, Fedora, and Arch, including post-install checks.

Ubuntu

Ubuntu 22.04+ for GNOME, KDE, X11, and Wayland.

Read guide

Fedora

Fedora Workstation setup with distro-specific notes.

Read guide

Arch

Arch, Manjaro, and EndeavourOS with injection checks.

Read guide

Speech engines

Local engines process audio on-device. Remote API is optional for a server you control.

whisper.cpp

Default

C++ Whisper with Vulkan. Fast install, multi-vendor GPU.

  • About 1–2 min default setup
  • AMD / Intel / NVIDIA via Vulkan
  • Tiny model ~74MB

Whisper

Original OpenAI PyTorch path for NVIDIA CUDA workflows.

  • Same model family accuracy
  • CUDA when you already live in PyTorch
  • Larger install footprint

VOSK

Small footprint for older machines and tight RAM budgets.

  • CPU-friendly streaming
  • Models around ~40MB
  • Great on modest hardware

Remote API

Offload to a server you trust while keeping desktop injection local.

  • OpenAI-compatible endpoints
  • whisper.cpp server support
  • Local VAD still applies
Full engine comparison

FAQ

Is Vocalinux really 100% offline?

Yes for local engines: whisper.cpp, Whisper, and VOSK process speech on your machine. Remote API is optional and only talks to servers you configure. Offline details.

Does Vocalinux collect usage telemetry?

No. The installed app does not send usage telemetry, analytics events, or background usage pings.

Which Linux distributions are supported?

Ubuntu 22.04+, Debian 11+, Fedora 39+, Arch, openSUSE Tumbleweed, and most modern desktops on X11 or Wayland. Install guides.

How do I switch between speech engines?

Settings dialog or CLI: --engine whisper_cpp, whisper, vosk, or remote_api. Compare engines.

Can Vocalinux use a remote transcription server?

Yes. OpenAI-compatible Whisper servers and the whisper.cpp server endpoint under Settings → Advanced → Remote Server. Remote API guide.

What happens when I close my laptop lid?

v0.10.1+ recovers speech recognition and shortcuts after suspend/resume. Reliability notes.

Does Vocalinux preserve my keyboard layout?

Yes. v0.10.1+ keeps your XKB layout when activating IBus.

Is Vocalinux free?

Yes. GPL-3.0, no premium tiers. About the project.

Voca on other platforms

Related projects for macOS and Windows. This site is about the Linux app.

VocaMac

Beta

Native macOS menu bar app. Offline voice-to-text with WhisperKit and CoreML.

WebsiteGitHub

VocaLinux

You are here

System tray app with whisper.cpp, Vulkan, and full offline support on Linux.

InstallGitHub

VocaWin

Planned

Windows tray app with an offline-first design. Still in planning.

WebsiteGitHub

Install and try a real dictation session

Free under GPL-3.0. Local engines by default. Star the repo if it helps your setup.

Install VocalinuxStar on GitHub