Speech to Text on Ubuntu: Your Free & Offline Options in 2026

The state of speech-to-text on Linux

For years, "voice recognition on Linux" meant cobbling scripts together. The major commercial dictation apps — Dragon, Superwhisper, Wispr Flow — skip Linux entirely. Old engines like IBM ViaVoice are long dead. What changed everything was OpenAI's Whisper: a free, offline, highly accurate model that anyone can run locally. The remaining question is how much setup you're willing to do.

Your Ubuntu speech-to-text options compared

Nerd Dictation (Vosk-based)

Best for: tinkerers who want free, open-source, and don't mind the terminal.

• Free and open source; uses Vosk for offline recognition.
• Lightweight and scriptable, good for simple dictation on X11.
• Limits: command-line setup, lower accuracy than Whisper-Large, and clunky on Wayland.

Whisper, the DIY way

Best for: developers who want maximum accuracy and total control.

You can run Whisper (or whisper.cpp / faster-whisper) directly and script transcription. You'll get state-of-the-art accuracy for free — but you maintain the glue: capturing the mic, triggering on a hotkey, and injecting text into the focused window. For how the engines themselves work, see our offline speech recognition explainer.

Desktop environment built-ins

GNOME and KDE offer accessibility and input features, and IBus-based input methods exist, but none provide the "hold a key, speak, text appears anywhere" experience out of the box. They're worth knowing about but rarely the answer for daily dictation.

AirTypes — offline and system-wide, no scripting

Best for: anyone who wants Whisper accuracy without building a pipeline.

AirTypes is one of the few polished dictation apps that treats Linux as a first-class platform. It packages Whisper into a native app and handles the hard parts — mic capture, hotkeys, and text injection on both X11 and Wayland — for you.

• Native Linux builds — .deb, .rpm, and AppImage for Ubuntu, Debian, Fedora, Arch, and more.
• 100% offline — Whisper runs locally; no audio leaves your machine.
• System-wide — dictate into terminals, IDEs, browsers, docs, and chat apps with one hotkey.
• Wayland-aware — automatically uses clipboard-paste injection where direct typing is restricted.
• My Agent (BYOK AI) — optionally route speech through your own AI key and type the reply at the cursor.

Ready to install? Don't repeat the legwork here — follow the step-by-step Linux setup guide (Ubuntu, Fedora, Arch), or the Whisper-on-Linux deep dive. For the platform overview, see offline voice to text for Linux.

Ubuntu specifics: audio and Wayland

• Audio: modern Ubuntu uses PipeWire (older releases use PulseAudio). Whisper-based tools work with either; PipeWire generally gives lower latency.
• Display server: Ubuntu defaults to Wayland on GNOME. Direct keyboard injection is restricted there, so the best tools fall back to clipboard-paste mode — which is exactly what AirTypes does automatically.
• Microphone check: run arecord -l to confirm your mic is detected before troubleshooting an app.

The full Ubuntu/Fedora/Arch walkthrough — packages, permissions, and hotkeys — lives in the Linux setup guide.

Which should you pick?

Pick a free DIY tool (Nerd Dictation / Whisper scripts) if you:

• Are comfortable in the terminal and want zero cost
• Only need basic dictation, mostly on X11

Pick AirTypes if you:

• Want Whisper accuracy with no setup or scripting
• Need system-wide typing that works on Wayland too
• Want offline privacy plus optional BYOK AI routing