Skip to content

This page describes Dictum 0.5.

Getting started

This page takes you from the download to your first dictation. Most of the time goes into downloading the speech model.

What you need

  • Windows 11 or Linux on a 64-bit x86 computer.
  • A processor with AVX2: Intel Haswell, AMD Zen or newer. Dictum does not start on an older one.
  • A microphone.
  • Disk space for a speech model: about 265 MB for the model the setup recommends for the processor, about 875 MB for the one it recommends for a graphics card.
  • A graphics card is optional. Whisper uses one through Vulkan when its driver supports it, and the processor otherwise. The setup recommends the larger model for a graphics card, unless the card has less than 2 GB of memory.

On Linux you also need:

  • glibc 2.38 or newer for hotkeys on X11: Ubuntu 24.04, Fedora 39, current Arch or newer. The X11 hotkeys also use the libraries libXtst, libXt, libXrandr and libxkbcommon, which a desktop install already has.
  • A clipboard tool to paste the text: wl-clipboard on Wayland, xclip or xsel on X11.
  • On Wayland, xdg-desktop-portal with your desktop's portal backend. KDE Plasma and GNOME install theirs. Dictum asks the portal for the hotkey and for permission to press Ctrl+V: see Wayland and X11.
  • Optional:
    • xprop on X11, so that Dictum recognises a terminal and pastes there with Ctrl+Shift+V;
    • GNOME Keyring or KWallet, to keep API keys out of the config file;
    • ydotool or wtype, to press Ctrl+V on Wayland when the desktop has no RemoteDesktop portal.

Install the clipboard tools with your package manager:

# Debian, Ubuntu
sudo apt install wl-clipboard xclip
# Fedora
sudo dnf install wl-clipboard xclip
# Arch, CachyOS
sudo pacman -S wl-clipboard xclip

Download

Download Dictum from the releases page. Each release has these files, where <version> is the release's version, such as 0.5.1:

File What it is
dictum-<version>-win-x64.msi Windows 11: the installer. Take this one unless you want no installer.
dictum-<version>-win-x64.zip Windows 11: the same programs, without an installer.
dictum-<version>-linux-x64.AppImage Linux: one file that runs as it is.
dictum-<version>-linux-x64.tar.gz Linux: installs for your user with install.sh. The Arch package is built from it too.
PKGBUILD Arch Linux and CachyOS: builds a pacman package from the tarball.
SHA256SUMS The SHA-256 sums of the four dictum-<version> packages.

Check the download

Before you install, you can check the file in two ways.

Its SHA-256 sum. In the folder you downloaded to, compare the file with its line in SHA256SUMS. On Linux, this prints OK after each file it finds:

sha256sum --check --ignore-missing SHA256SUMS

On Windows, this prints True when the MSI matches:

(Get-FileHash .\dictum-0.5.1-win-x64.msi).Hash -eq ((Select-String 'win-x64.msi' .\SHA256SUMS).Line -split ' ')[0]

Its VirusTotal report. The notes of each release on the releases page have a VirusTotal section with a link for each package. VirusTotal scans a file with dozens of antivirus engines, and the report shows what each of them found. Dictum is not signed yet and is little known, so one or two engines may flag it as suspicious while the rest find nothing.

Install

The commands on this page use version 0.5.1. Put in the version you downloaded.

The installer

  1. Open dictum-<version>-win-x64.msi. It asks no questions.
  2. If Windows says "Windows protected your PC", click More info, then Run anyway. See If Windows blocks Dictum.
  3. The installer needs no administrator rights. It installs Dictum for your user only, into %LOCALAPPDATA%\Programs\Dictum, and adds Dictum to the Start menu.
  4. When it is done, Dictum starts and opens its setup. Go on with the first run.

The zip, without an installer

  1. Extract dictum-<version>-win-x64.zip. It holds one folder, dictum-<version>-win-x64.
  2. Move that folder to a place that stays put and give it a name without the version, for example C:\Tools\Dictum. Not your Downloads folder: start at login remembers where the programs are, and an update goes into the same folder.
  3. Run dictum-gui.exe from that folder. Dictum opens its setup.

Pick one of the three.

AppImage: one file, and nothing is installed.

  1. Move the file to a place that stays put, give it a name without the version, and make it executable:

    mkdir -p ~/Applications
mv ~/Downloads/dictum-0.5.1-linux-x64.AppImage ~/Applications/Dictum.AppImage
chmod +x ~/Applications/Dictum.AppImage

  2. Run it. Dictum opens its setup.

    ~/Applications/Dictum.AppImage

Start at login remembers this path, and an update replaces the file under the same name. The command line is inside the AppImage too: put dictum after the file name, as in ~/Applications/Dictum.AppImage dictum status.

Tarball, for your user: installs into ~/.local/lib/dictum.

  1. Unpack the tarball and run its install.sh:

    cd ~/Downloads
tar -xzf dictum-0.5.1-linux-x64.tar.gz
./dictum-0.5.1-linux-x64/install.sh

  2. install.sh copies Dictum to ~/.local/lib/dictum, links dictum and dictum-gui into ~/.local/bin and adds Dictum to the application menu. It needs no root. Afterwards you can delete the tarball and the unpacked folder.

  3. Start Dictum from the application menu. It opens its setup.

Arch Linux, CachyOS: a pacman package, installed into /opt/dictum.

  1. Put PKGBUILD and dictum-<version>-linux-x64.tar.gz from the same release in one folder.

  2. In that folder, run:

    makepkg -si

    makepkg checks the tarball against the sum in PKGBUILD, builds the package dictum-bin and installs it with pacman, which asks for your password.

  3. Start Dictum from the application menu. It opens its setup.

If Windows blocks Dictum

Dictum's installer and programs are not signed yet, so Windows warns about them:

  • SmartScreen says "Windows protected your PC" the first time you open the MSI or dictum-gui.exe. Click More info, then Run anyway.
  • Smart App Control blocks apps that are not signed and has no exception for a single app. While it is on, Dictum cannot be installed or run. Windows Security → App & browser control → Smart App Control settings shows whether it is on.

Troubleshooting has more.

First run: the setup wizard

The first time Dictum starts, it opens Set up Dictum, a wizard of eight steps: Languages, Speech model, Text clean-up, Your words, Start at login, Downloads, Downloading and Ready. You pick the languages you dictate in, and Dictum recommends a speech model for your graphics card or processor. Clean-up with a language model and your own words are optional. Nothing is downloaded until you confirm the list of files, and each file is checked against its SHA-256 sum. You can close the window while the files download: the download goes on, and continues after a restart. The setup wizard reference explains every step.

If you close the wizard before the end, it keeps your answers: choose Settings in the tray menu to open it again at the same step. Closed while its files download, it comes back by itself when they are in.

Dictum then lives in the tray, by the clock. Windows hides new tray icons behind the ^ arrow, and the Ready step shows how to keep Dictum's icon in view, so that you can see when it records. On GNOME, the icon needs the "AppIndicator and KStatusNotifierItem Support" extension.

Your first dictation

  1. Click where the text should go: a text editor, a browser's address bar, a chat.
  2. Hold Ctrl+Alt+Space. Dictum plays a short tick, and a small overlay at the bottom of the screen shows that it is listening.
  3. Speak.
  4. Release the keys. A moment later, the text appears where the cursor is.

The overlay while Dictum listens The overlay while Dictum listens

The Ready step of the wizard has a box, Try dictation, to try this before you close it. The first dictation after Dictum starts waits until the speech model is in memory.

If no text appears, run dictum doctor (on Windows by its full path, see The command line) and see Troubleshooting. Dictum cannot paste into an app that runs as administrator: see Troubleshooting.

Wayland and X11

On Linux, how Dictum gets the hotkey and pastes depends on your session. To see which one you are in, run:

echo $XDG_SESSION_TYPE
  • X11: the hotkeys and pasting work without any setup.
  • Wayland: an app may not watch the keyboard or press keys in other apps by itself, so Dictum asks the desktop portal, xdg-desktop-portal:
    • for the hotkey, through the GlobalShortcuts portal, which KDE Plasma 6, GNOME 48 and newer, and Hyprland have. The desktop asks you once to allow the shortcut.
    • for pressing Ctrl+V, through the RemoteDesktop portal. The desktop asks you once whether Dictum may control the keyboard, and Dictum remembers the answer. Without that portal, Dictum uses ydotool when it is installed. On Sway and Hyprland the order is different: wtype first, then ydotool, then the portal.

  • No GlobalShortcuts portal, as on GNOME before 48: bind a key in your desktop's keyboard settings to the command dictum toggle. Press it once to start and once more to stop and paste. In Sway, you can hold the key to talk:

    bindsym --no-repeat Ctrl+Alt+space exec dictum start
bindsym --release Ctrl+Alt+space exec dictum stop

  • Terminals on Wayland: Dictum cannot see which app has the focus, so it pastes with Ctrl+V everywhere. A terminal wants Ctrl+Shift+V: set Paste shortcut in a profile of its own, in Settings → Profiles (see Settings).

dictum doctor lists what works, and what to install or allow when something does not.

Set up without the desktop app

If you prefer the terminal, or your desktop has no tray, set Dictum up with dictum init instead of the wizard:

dictum init

It asks the same questions one after another: your languages, the speech model, clean-up with a language model, your words, and whether to start Dictum at login. Then it lists the files to download, and downloads them after you agree.

Then run Dictum without the desktop app, in the foreground. Ctrl+C stops it.

dictum daemon

dictum daemon needs a desktop session for the hotkeys and pasting: Windows, or Wayland or X11 on Linux. For a script, dictum init --yes takes the options you give, such as --language "uk, en", and the recommendations for the rest; dictum init --help lists them.

The command line

On Linux, the tarball and the Arch package put dictum on your PATH. If ~/.local/bin is not on it, install.sh says so. With the AppImage, put dictum after the file name: ~/Applications/Dictum.AppImage dictum doctor.

On Windows, dictum.exe sits next to dictum-gui.exe and is not on PATH. After the MSI, run it by its full path in PowerShell:

& "$env:LOCALAPPDATA\Programs\Dictum\dictum.exe" doctor

After the zip, use your folder instead, for example & "C:\Tools\Dictum\dictum.exe" doctor.

To type plain dictum, add the folder to your Path:

  1. Open Start, type "environment" and choose Edit environment variables for your account.
  2. Under User variables, select Path and click Edit.
  3. Click New, enter %LOCALAPPDATA%\Programs\Dictum (or your zip folder), and click OK in both windows.

  4. Open a new PowerShell window and check:

    if (Get-Command dictum -ErrorAction SilentlyContinue) { 'dictum is on PATH' } else { 'dictum is not on PATH' }

Start at login

The wizard's Start at login step turns it on. Later, use Start Dictum at login in Settings → General. Dictum then starts its tray app when you log in, and the tray app starts the daemon. Without the desktop app, dictum daemon --install-autostart starts the daemon alone at login.

With the zip or the AppImage, the login item remembers where Dictum is. If you move it, turn start at login off and on again. Start at login has the details.

Next steps