Dank Material Shell (DMS) Whisper Plugin

July 7, 2026 · View on GitHub

⚠️ Deprecated / retired — this plugin is no longer maintained. Use Whisperer instead. The archived code here still works.

A Dank Material Shell plugin that brings the power of Whisper AI transcription directly to your status bar.

Record voice notes with a single click (or via IPC keybinds), have them transcribed in the background using Whisper, copied automatically to your clipboard, and cleanly backed up locally (both the audio and the generated text).

Features

  • Minimalist Design: A teal pre-Hispanic conch icon that blinks and highlights when recording, seamlessly integrating with your DMS theme.
  • Silent Workflow: Notifies you of recording start and successful transcription using native system notifications.
  • IPC Controlled: Easily integrate it into your window manager (Hyprland, Niri, Sway, etc.) via global keybinds.
  • Organized History: Saves separate timestamped audio and text files, and maintains a centralized chronological markdown log.

Requirements and Dependencies

Before installing the plugin, ensure you have the necessary recording utilities, clipboard managers, and the Whisper engine installed on your system.

Arch Linux (CachyOS, EndeavourOS, etc.)

  1. Install system utilities:

    sudo pacman -S alsa-utils wl-clipboard ffmpeg

  2. Install a Whisper backend:

    # OpenAI Whisper CLI
sudo pacman -S python-pipx
pipx install openai-whisper

# faster-whisper Python backend used by the plugin fallback
sudo pacman -S python-faster-whisper

# Optional: faster-whisper-compatible CLI wrapper
pipx install whisper-ctranslate2

# Optional: whisper.cpp CLI (GPU-accelerated via Vulkan, works on any GPU)
# CachyOS ships a prebuilt binary in its repos:
sudo pacman -S whisper-cpp-vulkan
# On plain Arch without those repos, build from the AUR instead:
# yay -S whisper.cpp

Ubuntu / Debian

  1. Install system utilities and pipx:

    sudo apt update
sudo apt install alsa-utils wl-clipboard ffmpeg pipx

  2. Install a Whisper backend:

    # OpenAI Whisper CLI
pipx install openai-whisper

# Recommended faster-whisper-compatible CLI wrapper
pipx install whisper-ctranslate2

# Alternative Python fallback if your distribution allows user installs
python3 -m pip install --user faster-whisper

# Optional: whisper.cpp CLI
# Install from your distro packages or build from https://github.com/ggerganov/whisper.cpp

Fedora

  1. Install system utilities and pipx:

    sudo dnf install alsa-utils wl-clipboard ffmpeg pipx

  2. Install a Whisper backend:

    # OpenAI Whisper CLI
pipx install openai-whisper

# Recommended faster-whisper-compatible CLI wrapper
pipx install whisper-ctranslate2

# Alternative Python fallback if your distribution allows user installs
python3 -m pip install --user faster-whisper

# Optional: whisper.cpp CLI
sudo dnf install whisper.cpp

(Note: alsa-utils provides the arecord tool, wl-clipboard provides wl-copy, and ffmpeg is required by the openai-whisper and whisper-ctranslate2 backends to decode the recorded audio. Make sure ~/.local/bin is in your $PATH environment variable so CLI backends installed via pipx are recognized. For the Python fallback, python3 -c "import faster_whisper" must work in the shell environment that launches DMS. For whisper.cpp, set the command and model path in plugin settings.)

Verify the backend installation

Run the checks for the backend you plan to use:

# OpenAI Whisper backend
command -v whisper

# faster-whisper CLI backend, optional
command -v whisper-ctranslate2

# faster-whisper Python fallback
python3 -c "import faster_whisper; print(faster_whisper.__version__)"

# whisper.cpp backend, optional
command -v whisper-cli

Plugin Installation

  1. Clone this repository into your local Dank Material Shell plugins folder:

    git clone https://github.com/arqueon/dms-whisper.git ~/.config/DankMaterialShell/plugins/dms-whisper

  2. Make the main script executable:

    chmod +x ~/.config/DankMaterialShell/plugins/dms-whisper/whisper-action.sh

  3. Open Dank Material Shell settings (usually by pressing Mod + ,), go to the Plugins tab, click the Scan button, and enable the plugin named DMS Whisper.

  4. To fully load the plugin and its IPC handler for the first time, restart your DMS environment:

    dms restart

Usage

Via Graphical Interface (Dankbar)

Simply left-click the pre-Hispanic conch icon located on your top or vertical panel. The icon will blink and highlight to indicate that recording is active. Click it again to stop capturing audio and start the automatic transcription.

Via IPC Commands (Keybinds)

The plugin registers a command in the Dank Material Shell IPC bus that allows you to toggle the recording state without relying on mouse clicks.

You can bind this command in your compositor's configuration file:

Hyprland (hyprland.conf):

# Bind to SUPER + W
bind = SUPER, W, exec, dms ipc whisper toggle

Niri (config.kdl):

// Bind to Mod + W
binds {
    Mod+W { spawn "dms" "ipc" "whisper" "toggle"; }
}

Where are my notes and audio files saved?

The plugin automatically organizes everything in your Documents folder:

  • Base Directory: ~/Documents/Whisper/
  • Original Audio: Whisper_YYYY-MM-DD_HH-MM-SS.wav
  • Raw Extracted Text: Whisper_YYYY-MM-DD_HH-MM-SS.txt
  • Global Record (The Log): WhisperNotes.md. This is a living document where all your transcriptions are sequentially appended as bullet points with their exact date and time.

Advanced Customization

Use the DMS plugin settings to change the Whisper backend, model, output directory, language, prompt context, or translation behavior.

The plugin supports three backend choices:

  • openai-whisper: default whisper command.
  • faster-whisper: default whisper-ctranslate2 command. If that CLI is not available, the plugin falls back to the installed faster_whisper Python package.
  • whisper.cpp: default whisper-cli command, plus a required local model path. If your build exposes main or another binary name, change it in settings.

By default, Whisper auto-detects the spoken language. Common model sizes, ordered by precision and weight, are: tiny, base, small, medium, large.

If transcription fails, check the last backend error at:

cat "${XDG_RUNTIME_DIR:-/tmp}/dms-whisper/last-error.log"