100% Private & Offline

Private, Offline
Speech-to-Text
for Linux.

A stable, simple, CPU-only dictation tool powered by OpenAI Whisper. Type directly into any app using your voice, with zero internet connection required.

install.sh
$ curl -sSL https://raw.githubusercontent.com/CS-1313/TuxVox/main/install.sh | bash
Copied!
Ubuntu Debian Fedora Arch Kali
TuxVox TuxVox

The easiest way to type using your voice on Linux. Just press a hotkey, speak, and the text magically appears in any window you're using...|

What is TuxVox?

TuxVox lets you quickly dictate text and paste it into any application. It uses OpenAI Whisper running entirely on your local hardware.

100% Offline

No data transmission or collection. After the initial one-time model download, TuxVox works entirely offline permanently.

CPU-Only Power

Highly optimized Whisper models run smoothly on standard processors. No expensive GPU or complex CUDA drivers needed.

Open Source

Licensed under AGPLv3. Transparent and completely free of telemetry, trackers, or any fees.

Core Features

Everything you need for effortless local dictation.

One-Click Recording

Single button to start and stop your microphone. No complicated menus.

Typewriter Effect

Transcribed text flows in word-by-word with a smooth, readable streaming effect.

Save Transcripts Locally

Store a copy of your transcription history to a .txt file for safe, private reference.

Clean-Slate Engine

Transcription engine resets between runs, preventing memory leaks or state bleed.

Smart Recommendation

Automatically detects your hardware RAM and suggests the optimal Whisper model.

Privacy First

Audio is processed locally and never stored. Complete data sovereignty.

GNOME Native

Built with GTK4 and libadwaita for a native, seamless Linux desktop experience.

Auto-Copy

Quickly copy all transcribed text to clipboard with a single click.

🧪 Experimental Feature

Global Hotkey & Inline Typing

Dictate directly into any application. Press a global hotkey, speak, and the text is typed directly into your current window—like magic.

  • Wayland & X11 Support: Bypasses display servers using evdev and /dev/uinput for universal compatibility.
  • Kernel-level injection: Keystrokes are injected at the kernel level, landing exactly where your focus is.
  • Audio Feedback: Clean, distraction-free audio chimes let you know when recording starts, stops, and types.
Global Hotkey
Ctrl + Shift + L
Start Chime
Dictation is now recording

Choose Your Model

TuxVox supports the full range of OpenAI Whisper models. Pick the right balance of speed and accuracy for your hardware.

Model Speed Accuracy Required RAM
Tiny ⚡ Fastest ★★☆☆☆ ~150 MB
Base 🚀 Fast ★★★☆☆ ~300 MB
Small Recommended 🔄 Balanced ★★★★☆ ~500 MB
Medium 🐢 Slower ★★★★★ ~1.5 GB
Large / V3 🐌 Slow ★★★★★+ ~3.0 GB

Frequently Asked Questions

Is an internet connection required?
No, TuxVox is 100% offline. The only time it requires network access is when you select a model for the first time, to download the AI weights. After that, it functions entirely offline indefinitely.
Why does it install in a virtual environment (venv)?
Modern Linux distributions like Ubuntu 24.04+ and Fedora block system-wide pip installs to prevent conflicts. A virtual environment keeps TuxVox's dependencies isolated and stable.
Does Experimental Mode work on Wayland?
Yes! Unlike many tools that break on Wayland, TuxVox uses direct kernel interfaces (evdev and /dev/uinput) to capture hotkeys and inject text directly, bypassing the display server entirely.
What if my Global Hotkey does nothing?
Ensure you are in the input group. You must run the setup script (sudo bash ~/TuxVox/scripts/setup-uinput.sh) and reboot your computer for group changes to take effect.
I hear the chime, but no text is typed.
This usually means /dev/uinput is not writable. Make sure you ran the setup-uinput.sh script to set up the udev rules, and verify the target window has focus before you press the hotkey.
How do I uninstall TuxVox?
Simply delete the ~/TuxVox application folder and your configuration in ~/.config/tuxvox. It leaves no other traces on your system aside from the optional udev rules.