Best Linux Dictation Apps: Speech to Text on Wayland & X11

Quick answer: Our pick for most Linux users is Blip AI: it types cleaned-up text into any app on Wayland and X11 from one AppImage, with no GPU, model download or API key. Pro adds on-device models; the same app runs on Mac, Windows and phones. For free, fully offline speech to text, use Vocalinux, then OpenWhispr or Handy.

We build Blip AI, one of the apps on this list, so weigh our picks accordingly. Prices, platforms and licences were checked on each vendor's or project's own site on 30 October 2026, and Superwhisper's re-checked on 5 November 2026. Where another app is the better choice, we say so. Using Mac or Windows too? See best dictation software in 2026.

Search for a dictation or speech to text app and you'll probably land on Wispr Flow, Superwhisper, Willow Voice or Aqua Voice. Their download pages offer Mac, Windows and phones, and nothing for Linux; Superwhisper's Linux build has its own page and needs Wayland. That leaves Linux users with two families of tools: open-source projects that run speech models locally, and a small number of commercial apps that ship a Linux build. Ubuntu now adds a third, Canonical's opt-in Myna preview. This guide covers all three, and speech to text on Ubuntu goes deeper on that distro. Linux Mint users should read speech to text on Linux Mint.

What's the best speech to text for Linux?

This guide covers apps that type what you say into whatever window has focus. Our pick for most people is Blip AI, and we put it first because it does all four of these:

1. Polished text without a GPU, model download or API key. Blip Cloud is the default, so even a thin laptop gets filler words removed and punctuation fixed.
2. Wayland and X11 from one AppImage, built for Ubuntu/Debian, Fedora and Arch on x86-64.
3. Offline when you need it, with on-device Whisper models on Pro.
4. The same app on Mac, Windows, iPhone and Android.

If you want free, open source and fully offline instead, pick Vocalinux, then OpenWhispr or Handy. The rest depends on what you need:

Why is dictation on Linux harder than on Mac or Windows?

Two things make Linux different.

Wayland blocks global input. On X11, an app can read a global hotkey and type into any window with xdotool. Wayland deliberately restricts both for security. Dictation apps have to use desktop portals, ydotool, wtype, dotool or compositor-specific workarounds, so many tools only partly work on GNOME or KDE under Wayland. Read any "Wayland support" claim as "Wayland support on the compositors the developer tested."

Local models need hardware. Whisper's large models want a GPU. The small models that run comfortably on a laptop CPU are noticeably less accurate, especially for non-English speech and technical vocabulary. Even Superwhisper's own FAQ says "offline models only run really well on Apple Silicon macs." Cloud transcription sidesteps the hardware problem, at the cost of needing a connection.

Which Linux dictation app cleans up your text without a GPU?

Blip AI is our pick here. Blip Cloud is the default, so a laptop with no GPU doesn't have to run a speech model, there's no API key to set up, and you still get filler words removed and punctuation fixed. OpenWhispr can do the same through OpenWhispr Cloud or your own API key. Spokenly does too, on Pro or with your own API keys. Handy and Speed of Sound transcribe on your machine, and can send the text to a cloud AI model for cleanup if you add a key. Superwhisper's Linux build (Linux page) also has cloud dictation, but needs Wayland with GNOME, KDE Plasma or Hyprland. Blip AI and OpenWhispr, checked 30 October 2026:

Blip AI is the one we'd pick if you want the cloud to do the work: Blip Cloud is the default, and on it text appears about 700 ms after you let go of Ctrl + Win. It also drafts replies when you start with "Hey Blip", has on-device models on Pro, and has apps for Mac, Windows, iPhone and Android. OpenWhispr suits you better if you want open source, or free local models alongside the cloud.

Does Ubuntu have built-in dictation? Canonical's Myna preview

Not by default, but Canonical now has an opt-in one. On 12 October 2026 it published Myna, a speech-to-text app for the Ubuntu desktop, as a technology preview. Per that announcement (checked 31 October 2026):

• Versions and desktop: Ubuntu 24.04 LTS, 26.04 LTS and 26.10, on GNOME using Wayland. Canonical's post names GNOME on Wayland only and doesn't mention X11 or KDE.
• Offline: once a model is downloaded, no internet connection is needed for transcription, and audio isn't uploaded or stored.
• Typing: you press a hotkey and speak, and Myna transcribes directly into the application you're using.
• Models: Parakeet, Whisper and FunASR, among the engines it supports.
• Install: sudo add-apt-repository ppa:canonical-desktop-team/myna, then sudo apt install myna-config, and finish setup in the Myna settings app.

Try Myna if you run Ubuntu on GNOME with Wayland, want offline dictation from Canonical, and don't mind a preview that Canonical says is still under active development. Canonical's page doesn't say whether Myna removes filler words or fixes punctuation, so check that yourself. Pick Blip AI instead if you use X11 or a distro other than Ubuntu (Fedora, Arch or Debian), also dictate on a Mac, Windows PC or phone, or want cleaned-up text from Blip Cloud without downloading a model.

How we picked

• System-wide typing: the app should put text into whatever window has focus. Every app below can, though Speech Note is built around its own notes window.
• Wayland and X11: we list what each project documents, not our own test on every compositor.
• Offline or cloud: both are valid; we say which each app does.
• Cleanup: whether filler words are removed and punctuation fixed, or you get a verbatim transcript.
• Price and maintenance: free or paid, and whether the project is still active in 2026.
• Order: Blip AI comes first because it meets the four criteria at the top of this guide. Then come the free, offline picks in the order we'd try them (Vocalinux, OpenWhispr, Handy), then the other apps.

The best dictation apps for Linux

Blip AI: our pick for most Linux users

Format: AppImage (x86-64) · Display: Wayland and X11 · Built for: Ubuntu/Debian, Fedora, Arch · Price: free plan with 2,000 words a month, no card; Pro $15/month or $12/month billed yearly

Blip AI works the same way on Linux as on Mac and Windows: hold Ctrl + Win, speak, let go, and cleaned-up text appears in whatever window has focus, including terminals and editors. On Blip Cloud, the default, text appears about 700 ms after you let go of the key. Filler words go, and punctuation and formatting are fixed. Blip gets technical terms like Vercel, Supabase and Claude right, and you can add your own names and jargon to its custom dictionary. Start with "Hey Blip" and Action Mode drafts the reply you describe instead of transcribing word for word.

Blip supports both cloud and on-device models. Transcription runs on Blip Cloud by default, so a thin laptop gets the same accuracy as a workstation, without you managing API keys. There's also an on-device option on Pro: download a Whisper model (Base, Small, Medium or Large Turbo) and Blip transcribes and cleans up text offline, with the audio staying on your machine.

Screenshot of blipai.app, taken October 24, 2026.

The honest limits: Blip is proprietary, the Linux AppImage is x86-64 only (on ARM, look at Vocalinux, Handy, Spokenly's .deb and .rpm, or Superwhisper), and the cloud path needs a connection.

Download Blip AI for Linux. On X11 it runs as is; on Wayland it needs some setup, covered below.

Vocalinux: best free, offline option

Format: install script, AppImage (x86-64 or ARM64), .deb, .rpm, AUR, Snap or PyPI · Display: X11 and Wayland · Price: free (AGPL-3.0)

Vocalinux is a free, open-source dictation app built for Linux. It uses whisper.cpp by default and can also run Whisper, Faster Whisper, Vosk or Parakeet, all locally by default, so nothing leaves your machine unless you set up its optional remote API. It types into whichever app has focus on both X11 and Wayland. The trade-offs: accuracy depends on the model your hardware can run, and you get a verbatim transcript with voice commands like "new line", not AI cleanup (you can hook in your own script to rewrite the text). Install and Wayland steps are in Vocalinux on Linux.

Screenshot of github.com, taken October 24, 2026.

OpenWhispr: open source with AI cleanup

Format: AppImage, .deb, .rpm or .tar.gz (x86-64; no ARM build) · Display: X11; Wayland needs setup · Price: free (MIT); Pro $80 per user per year

OpenWhispr runs Whisper, Parakeet or Cohere Transcribe locally, or cloud transcription through OpenWhispr Cloud or your own API key. Unlike most open-source tools, it has a real cleanup step, on by default, that removes filler words and fixes grammar and punctuation. You choose the model that does it: OpenWhispr Cloud, your own OpenAI, Anthropic, Gemini or Groq key, or a local model. A user in our Discord suggested it, and it's under active development.

On Wayland, OpenWhispr's Linux guide has you join the input group so its paste helper can type, and it falls back to wtype on Sway and Hyprland or the ydotool daemon on GNOME and KDE. The free plan includes unlimited local models and your own keys, plus 2,000 words a week of OpenWhispr Cloud; Pro adds unlimited cloud transcription (pricing, checked 30 October 2026). Packages, Wayland setup and hotkeys are in OpenWhispr on Linux.

Screenshot of github.com, taken October 24, 2026.

Handy: the simplest local option

Format: AppImage, .deb or .rpm, each for x86-64 and ARM64 (v0.9.8 release) · Display: X11; limited Wayland support · Price: free (MIT)

Handy is a small, cross-platform, offline dictation app. Press a shortcut, speak, and the text appears in the active window. It runs Whisper and Parakeet models locally, and you can add your own Whisper models. Filler-word removal is on by default, and AI post-processing is an opt-in setting that uses a provider such as OpenAI, Anthropic, OpenRouter or Groq, or a custom endpoint. Mac and Windows builds exist too (setup and Wayland details: Handy on Linux). Handy's README calls its Wayland support limited: you need wtype or dotool (or ydotool on Ubuntu 26.04), and you set the global shortcut in your desktop settings.

Screenshot of github.com, taken October 24, 2026.

Voxtype: built for Wayland first

Format: AUR, .deb, .rpm or a standalone binary · Display: Wayland (Hyprland, Niri, Sway, River, GNOME, KDE) and X11 · Price: free (MIT)

Voxtype is a Rust push-to-talk dictation tool that's local by default, with nine transcription engines including Whisper, Parakeet and Moonshine, plus optional remote Whisper servers. It can pipe each transcription through an external command, such as a local LLM, to remove filler words or fix grammar. You set it up in the terminal rather than a settings window. If you run a tiling Wayland setup and like configuring things, start here. For install commands and a compositor hotkey for each desktop, see Voxtype on Linux.

Screenshot of github.com, taken October 24, 2026.

Speech Note: offline notes, with voice typing

Format: Flatpak · Display: X11; Wayland needs ydotool · Price: free (MPL-2.0)

Speech Note is a notes app first, with many offline speech engines, languages, text-to-speech and translation. It's the easiest install here, one click on Flathub. It can also act as a voice typing tool: its "insert into active window" action, bound to a global shortcut, types into whatever window has focus. Its README says that on Wayland this needs the ydotool daemon, and global shortcuts need a recent KDE Plasma or GNOME.

Screenshot of flathub.org, taken October 24, 2026.

Speed of Sound: portal-based, on Flathub

Format: Flatpak, Snap, AppImage, .deb or .rpm · Display: X11 and Wayland via desktop portals · Price: free (MIT)

Speed of Sound transcribes on your device with Whisper, Parakeet or Canary models and types into the focused app through desktop portals, which is the cleanest way to support both display protocols. Optional text polishing can use a local model through Ollama or llama.cpp, or a cloud provider. It's younger and smaller than the others here, and quieter lately: its latest release (v0.14.0) came out in May 2026 and its last commit in July 2026.

Screenshot of github.com, taken October 24, 2026.

nerd-dictation: for people who script their desktop

Format: Python script · Display: X11 via xdotool; Wayland via ydotool, dotool or wtype · Price: free (GPL-3.0)

nerd-dictation has no GUI. You bind a key to start and stop it, it runs the Vosk engine offline, and it types into the focused window. You can rewrite its output with a Python config file. It hasn't had a commit since October 2025, so treat it as finished rather than actively developed. Install steps for Ubuntu and Fedora, which VOSK model to pick and how to make it type on Wayland are in our nerd-dictation guide. For how Whisper models fit into Linux dictation and which size to run, see Whisper dictation on Linux.

If you're coming from Talon Voice, which is voice control rather than dictation, see Talon Voice alternatives for Linux dictation.

Screenshot of github.com, taken October 24, 2026.

Superwhisper: official on Linux, Wayland only

Format: one-line installer (.tar.zst archive) for x86-64 or ARM · Display: Wayland only, with GNOME, KDE Plasma or Hyprland · Price: plans page lists Pro at $8.49/month and doesn't mention Linux (checked 5 November 2026)

Superwhisper's Linux page (checked 5 November 2026) lists Ubuntu 24.04+, Debian 13+, Fedora 40+, Linux Mint 22+, Arch, CachyOS, openSUSE Tumbleweed and Omarchy, on x86-64 and ARM, with a one-line installer, and says it "Needs Wayland (Hyprland, KDE Plasma or GNOME) and systemd", so it isn't an option in an X11 session. Version 1.2.0 came out on 3 November. The builds started in September 2026 as Omarchy previews on GitHub, whose README still calls them prerelease software. It's worth a look if you're on one of those Wayland desktops or on ARM. More on Superwhisper for Linux.

Screenshot of github.com, taken October 24, 2026.

Spokenly: commercial, with a free local tier

Format: .deb or .rpm (x86-64 and ARM64), AppImage (x86-64) · Display: X11 and Wayland · Price: free local tier; Local+ $29 one-time; Pro $12/month billed yearly (checked 30 October 2026)

Spokenly is a commercial dictation app with a native Linux build. Its free tier gives unlimited on-device transcription with local Whisper and Parakeet models, with no account. Pro adds cloud engines, or you can bring your own API keys, and its AI text processing removes filler words and fixes punctuation. Spokenly says it supports X11 and Wayland but notes that shortcuts and text insertion vary by compositor, so test your setup. Prices are from Spokenly's pricing page. It also runs on Mac, Windows and iPhone.

Screenshot of spokenly.app, taken October 24, 2026.

Talon: voice control, not just dictation

Format: binary · Display: X11 only · Price: free download; Talon+ on Patreon

Talon is for controlling the whole desktop by voice, including writing code. It's the deepest tool here and has the steepest learning curve; dictation is a subset of what it does. Talon doesn't support Wayland, and its changelog says version 1.0 is planned to be the last public release for Linux/X11, with ongoing Linux support moving to the paid Talon+ tier.

Screenshot of talonvoice.com, taken October 24, 2026.

Linux dictation apps compared

Voxtype vs Handy: which should you use?

Voxtype and Handy are both free, MIT-licensed and run speech models on your own machine. The difference is how you set them up and which desktops they suit. Both READMEs were checked on 30 October 2026.

Pick Handy if you want an app you can install and start using without reading docs, you're on X11, or on GNOME or KDE and don't mind adding a shortcut and ydotool, or you also work on a Mac or a Windows PC. Its filler-word removal works out of the box.

Pick Voxtype if you run a tiling Wayland compositor such as Hyprland, Sway or Niri, you're happy editing keybindings, or you want more engines and a status indicator in Waybar.

Both give you close to a verbatim transcript unless you wire up a language model yourself. If you want filler words and punctuation cleaned up and replies drafted without running or paying for a model of your own, that's what Blip AI for Linux does.

Do Wispr Flow, Superwhisper, Willow or Aqua Voice work on Linux?

Mostly no. These are the apps people search for most. Three have no Linux version; Superwhisper has one that needs Wayland:

• Wispr Flow: No. Its help center lists Mac, Windows, iPhone and Android, and Linux as not supported. The workarounds and what to use instead are in Wispr Flow for Linux.
• Superwhisper: Mac, Windows, iOS and Android, plus Linux: its Linux page (checked 5 November 2026) lists Ubuntu 24.04+, Debian 13+, Fedora 40+, Mint 22+, Arch and more, and says it needs Wayland with GNOME, KDE Plasma or Hyprland. The GitHub README still calls the builds prerelease. Superwhisper on Linux · Blip AI vs Superwhisper
• Willow Voice: Mac, Windows and iPhone, with Android coming soon. Linux isn't mentioned on its site. Willow Voice on Linux
• Aqua Voice: Mac, Windows, iPhone and Android. Its blog points Linux users to command-line tools. Aqua Voice on Linux · Blip AI vs Aqua Voice

Looking for Dragon NaturallySpeaking? It's Windows-only: see Dragon NaturallySpeaking on Linux. Asking about Typeless? See Typeless for Linux. Spokenly, VoiceInk or MacWhisper? See Spokenly, VoiceInk and MacWhisper on Linux.

How to install Blip AI on Linux

On X11, the AppImage works out of the box: make it executable and run it. On Wayland there are four steps, about two minutes, so Blip can type into other windows:

1. Install the tools. Ubuntu/Debian: sudo apt install wl-clipboard wtype. Fedora: sudo dnf install wl-clipboard wtype. Arch: sudo pacman -Syu wl-clipboard wtype. wtype types on Hyprland and Sway; you don't need ydotool.
2. Join the input group for push-to-talk: sudo usermod -aG input $USER. Then fully log out and back in, and check with groups | grep input.
3. On GNOME or KDE, allow Blip to paste. The first time you dictate, your desktop asks whether Blip can control your keyboard; allow it. On KDE Plasma 6 the paste doesn't always go through, but your words are on the clipboard, so press Ctrl + V.
4. Run the AppImage. Make sure ~/.local/bin is on your PATH, then run chmod +x Blip-AI-*.AppImage and ./Blip-AI-*.AppImage.

Allow microphone access when asked. If you move the AppImage later, re-run step 4. The Blip AI for Linux page has the download and the same steps with copy buttons.

How to choose a Linux dictation app

Ask three questions.

Can your machine run a local model well? With a recent GPU or a fast CPU, the open-source tools are free and private, so try Vocalinux, OpenWhispr or Handy first. On a thin laptop or an older desktop, cloud transcription will be more accurate than the small local models you can run, and Blip's free plan lets you check that without paying.

Do you want a verbatim transcript or finished text? Vocalinux, Speech Note and nerd-dictation type what you said. OpenWhispr, Blip AI and Superwhisper clean it up, and Handy, Voxtype and Speed of Sound can if you add a model.

Do you only use Linux? If yes, a Linux-only open-source tool is fine. If you also work on a Mac, a Windows PC or your phone, one app and one account across all of them saves setting things up twice. Blip AI and Superwhisper both cover Linux, Mac, Windows, iOS and Android. Superwhisper's Linux build needs Wayland with GNOME, KDE Plasma or Hyprland, so X11 users are ruled out (its README still calls the builds prerelease); Blip's Linux AppImage covers Ubuntu/Debian, Fedora and Arch on Wayland and X11. Spokenly covers Linux, Mac, Windows and iPhone.

For more options beyond Linux, see our Wispr Flow alternatives comparison.

FAQ

What's the best speech to text app for Linux? For most Linux users, we recommend Blip AI: Blip Cloud removes filler words and fixes punctuation without a GPU, Pro adds offline Whisper models, and the same app runs on Mac, Windows, iPhone and Android. If you'd rather have free, open-source dictation that runs fully offline, start with Vocalinux, then try OpenWhispr or Handy.

What's the best free speech to text app for Linux? For fully offline and open source, Vocalinux. If you also want AI cleanup of filler words, OpenWhispr. For cloud accuracy without a GPU, Blip AI's free plan gives you 2,000 words a month with no card.

Does Ubuntu have built-in dictation? Not by default. Canonical's Myna, released 12 October 2026 as a technology preview, is an opt-in app you install from a PPA. It runs on Ubuntu 24.04, 26.04 and 26.10 with GNOME on Wayland and works offline once a model is downloaded.

Is Voxtype or Handy better? Handy is easier: a desktop app with a settings window that also runs on Mac and Windows, with filler-word removal on by default. Voxtype suits tiling Wayland compositors such as Hyprland and Sway, offers nine local engines and is set up in the terminal. Both are free and run offline.

Does Wispr Flow work on Linux? No. Wispr Flow's help center lists Linux as not supported; the workarounds and what to use instead are in Wispr Flow for Linux.

Does Superwhisper have a Linux version? Yes. Superwhisper's Linux page (checked 5 November 2026) lists Ubuntu 24.04+, Debian 13+, Fedora 40+, Linux Mint 22+, Arch, CachyOS, openSUSE Tumbleweed and Omarchy, on x86-64 and ARM, and says it needs Wayland (GNOME, KDE Plasma or Hyprland) and systemd. Its GitHub README still calls the builds prerelease.

Does dictation work on Wayland? Yes, but most apps need a helper tool to type into other windows. Speed of Sound types through desktop portals, and so does Blip AI on GNOME and KDE (it uses wtype on Hyprland and Sway). Vocalinux, Voxtype, Handy, OpenWhispr, Speech Note and nerd-dictation use tools such as wtype, ydotool or dotool, and some ask you to join the input group. Superwhisper's Linux build needs Wayland with GNOME, KDE Plasma or Hyprland; Talon is X11 only.

Can I dictate into a terminal on Linux? Yes. Any system-wide dictation app on this list types into the focused window, including a terminal. Blip AI sends the terminal's own paste shortcut, so it works in Konsole, Kitty and Alacritty.

Is there an offline dictation app for Linux? Yes. Vocalinux, OpenWhispr, Handy, Voxtype, Speech Note, Speed of Sound, nerd-dictation and Spokenly's free tier all run locally. Blip AI Pro also has an on-device option alongside its default cloud transcription. For Mac and Windows too, see offline dictation apps.

Last checked 30 October 2026. If a vendor has shipped a Linux build since then, tell us and we'll update this page.