Voice dictation that actually runs on Linux
Wispr Flow and Aqua Voice have no Linux client, and Dragon’s desktop product is Windows-only. If you got here after checking all of them, this is the short version: VibeVoice ships a .deb for Debian and Ubuntu, and the same binary runs on Fedora and Arch.
Is there a dictation tool that actually runs on Linux?
Almost none. Wispr Flow and MacWhisper have no Linux build, and Dragon’s desktop product is Windows-only. The two that do run are Talon Voice, which costs nothing and is enormously capable but needs a command set installed before it recognises anything, and only runs on X11, and VibeVoice, which ships a .deb for Debian and Ubuntu, and the same binary runs on Fedora and Arch and works straight away.
Otter.ai opens in a Linux browser but never types into your applications, so it does not solve the dictation problem. If you want full hands-free control of the machine rather than dictation, Talon is the better tool and we say so on its comparison page.
Which dictation apps support Linux?
| App | Linux client | Notes |
|---|---|---|
| VibeVoice | Yes | Native client, same features as Windows and macOS |
| Wispr Flow | No | Windows, macOS, iOS and Android — no Linux client |
| Superwhisper | No | macOS, Windows and iOS — no Linux |
| Dragon | No | Windows only |
| Otter.ai | Browser | Web app, but it does not type into your applications |
| Talon Voice | Yes (X11) | Cross-platform, but Linux is X11 only and you install a command set first |
Talon is the other real Linux option and is a serious tool — see our Talon comparison for where it is the better choice.
A compiled binary, not a browser
It sits in the system tray, not in a tab. There is no runtime to install alongside it, and nothing has to stay open for dictation to keep working.
Text while you talk
Text arrives punctuated, about half a second after you stop, and the transcription is not done by your laptop — so next to no fan noise and next to no battery cost while you dictate.
X11 and Wayland
A global hotkey you choose inserts text into the focused field on both display servers — terminals, editors, browsers and GTK or Qt applications alike.
Installing on Linux
The client ships as a .deb package for Debian and Ubuntu. Download it from the download page and install it with your package manager:
sudo apt install ./vibevoice-*-amd64.deb vibevoice-client # first run prints a device code
Confirm the device code once in the browser and the client holds its own key from then on. On Fedora, Arch and anything else without dpkg, unpack the .deb or run the binary directly — there is no distribution-specific package yet, and pretending otherwise would waste your afternoon.
Once the package is installed
Two steps, and neither is a compiler.
Link it to your account
The client shows a device code once; you confirm it in the browser. There is no key file to copy around and no config to edit by hand.
Hold your hotkey and speak
Text lands in whatever window has focus, on X11 and on Wayland alike. Release the key and read it back before you send anything.
What it does not do
- It needs a network connection. Recognition runs on our servers, so there is no offline mode and no local model — if the audio must stay on your machine, this is the wrong tool.
- It types; it does not control the desktop. No voice commands for windows, menus or the cursor. Talon Voice is the tool for that, and on Linux it is genuinely good.
- Some applications refuse synthetic input. Games that grab the input device, remote-desktop and virtual-desktop sessions, and hardened enterprise clients do not accept programmatic keystrokes, and the client cannot type into them.
- Packaging is Debian and Ubuntu only. There is no RPM, no AUR entry and no Flatpak yet; on other distributions you unpack the .deb or run the binary yourself.
- There is no custom vocabulary. Package names, internal service names and unusual proper nouns come back as ordinary English and have to be corrected by hand.
Common questions
Which dictation tools actually support Linux?+
Three, realistically. Talon Voice runs on Linux and is a serious piece of software, but it is a framework you configure in Python rather than an application you install. VibeVoice ships a native client that types into the focused field on both X11 and Wayland. Wispr Flow, MacWhisper and Dragon NaturallySpeaking have no Linux build at all; Otter.ai opens in a browser but never types into your applications.
Does it work on Wayland as well as X11?+
Yes, on both. Synthetic input is handled differently on the two, and a tool built only for X11 silently does nothing in a Wayland session.
Which distributions are supported?+
The package is a .deb, so Debian, Ubuntu and their derivatives install it directly. On Fedora, Arch and others the binary runs but there is no native package yet — no RPM, no AUR entry and no Flatpak. That is a genuine gap rather than an oversight.
Does it need a GPU?+
No. Transcription happens on our servers, so the client is doing almost nothing: it captures audio, streams it, and types the result. A ten-year-old ThinkPad performs the same as a workstation, which is the practical advantage of not running the model locally.
Does it run offline?+
No. Audio is streamed for transcription, so a connection is required. If offline operation is a hard requirement on Linux, something you run and host yourself is the honest answer — we do not offer one.
What does it cost?+
30 minutes a month free with no card required, then €3 a month for Pro. One account covers the Linux, macOS and Windows clients, uploaded recordings and the WhatsApp bot.
The same client runs on Windows 11 and macOS. Wanting hands-free control of the desktop rather than dictation? VibeVoice vs Talon Voice says why Talon is the better tool for that.
Where to go next
Last updated . Competitor figures checked August 2026: prices, platforms, and each vendor’s own description of when its text is finished, which is their account rather than our measurement. All of it changes, so check theirs before buying.