Flüstern

WhisperFlow-style voice dictation for Linux

A lightweight voice-to-text tool that uses Groq's Whisper API for transcription and LLM for formatting. Works with any X11/Wayland window manager.

Features

• Fast transcription using Groq's Whisper Large V3 Turbo
• Smart formatting with automatic punctuation and capitalization
• Modern Slint GUI - Beautiful, fast, cross-platform UI
• CPU-optimized - No more high CPU usage after closing
• System tray icon with status indicators
• 50+ languages including Hindi, Arabic, Chinese, and more
• Microphone selection via tray menu
• Keyboard shortcut toggle (start/stop recording)
• Pastes directly into any focused text field
• Recording history with corrections and timing data

Demo

[Press shortcut] → Recording...
[Speak] "hello world this is a test"
[Press shortcut] → Processing...
[Output] "Hello world, this is a test."

Requirements

• Linux with X11 or Wayland
• PipeWire or PulseAudio
• Rust/Cargo (for building GUI)
• Groq API key (free tier available)

Installation

git clone https://github.com/chukfinley/fluistern.git
cd fluistern
./install.sh

The installer will:

1. Check and install dependencies (if needed)
2. Build the modern Slint GUI (Release mode, optimized)
3. Copy files to ~/.local/share/fluistern/
4. Create fluistern command in ~/.local/bin/
5. Set up systemd service
6. Prompt for Groq API key

After install, you can delete the cloned folder.

Dependencies

The installer will check for these:

Required:

• cargo / rust - Rust toolchain for building GUI
• yad - tray icon
• xdotool - simulating paste
• xclip - clipboard access
• ffmpeg - audio compression
• jq - JSON parsing
• curl - API calls
• pw-record (PipeWire)
• sqlite3 - database for recording history

On Ubuntu/Debian:

sudo apt install cargo yad xdotool xclip ffmpeg jq curl pipewire sqlite3

On Arch/Manjaro:

sudo pacman -S rust yad xdotool xclip ffmpeg jq curl pipewire sqlite3

Configuration

Get Groq API Key

1. Go to console.groq.com/keys
2. Create a free account
3. Generate an API key
4. The installer will prompt you, or edit ~/.local/share/fluistern/.env

Config Options

Edit ~/.local/share/fluistern/.env:

GROQ_API_KEY="your-key"      # Required
LANGUAGE="de"                 # Optional: en, de, es, fr, hi, etc.
MIC_SOURCE=""                 # Optional: specific mic (use tray menu to select)
NOTIFICATIONS="true"          # Show notifications (true/false)
TRAY_ICON="true"              # Show tray icon (true/false)

Add Keybinding

Add a keybinding in your WM config to run fluistern:

WM/DE	Config	Example
sxhkd	~/.config/sxhkd/sxhkdrc	super + r fluistern
Hyprland	~/.config/hypr/hyprland.conf	bind = SUPER, R, exec, fluistern
i3/sway	~/.config/i3/config	bindsym $mod+r exec fluistern
dwm	config.h	{ MODKEY, XK_r, spawn, SHCMD("fluistern") }

Usage

1. Start the daemon: systemctl --user start fluistern
2. Press your keybind to start recording
3. Speak
4. Press keybind again to stop and transcribe
5. Text appears in your focused window

Tray Menu (right-click)

• Toggle Recording - Start/stop recording
• Einstellungen & Historie - Open GUI for settings, logs, and recording history
• Select Microphone - Choose input device
• Select Language - Set transcription language (50+ languages)
• Quit - Stop the daemon

GUI Features

Open the modern Slint-based GUI from the tray menu to access:

• 📝 Recording History - View all recordings with expandable cards showing Whisper raw output, LLM formatted output, and correction fields
• ✏️ Corrections - Teach the system by correcting misheard words
• 🪲 Debug Logs - Real-time logs with auto-refresh for troubleshooting
• ⚙️ Settings - Configure API key, language, microphone, system prompt

The GUI features a clean, modern design with:

• Tabbed interface for easy navigation
• Expandable recording cards
• Real-time log monitoring (only refreshes when file changes - CPU optimized!)
• Professional blue color scheme

Supported Languages

Auto-detect, English, German, Spanish, French, Hindi, Italian, Portuguese, Dutch, Polish, Russian, Chinese, Japanese, Korean, Arabic, Turkish, Vietnamese, Thai, Indonesian, Ukrainian, Czech, Greek, Hebrew, Hungarian, Swedish, Danish, Finnish, Norwegian, Romanian, Bengali, Tamil, Telugu, Urdu, Persian, and many more.

How It Works

Keybind → Start recording (16kHz mono WAV)
Keybind → Stop → Compress to Opus (~30x smaller)
       → Upload to Groq Whisper API
       → Format with Groq LLM (punctuation, caps)
       → Pastes into focused window (Ctrl+V)

Files

After installation:

~/.local/share/fluistern/
├── fluistern-gui           # Rust GUI binary (settings, history, logs)
├── voice-input.sh          # Main toggle script
├── voice-input-daemon.sh   # Tray daemon
├── select-mic.sh           # Microphone selector
├── select-language.sh      # Language selector
├── .env                    # Your config (API key, language, mic)
├── history.db              # SQLite database for recordings
└── icons/                  # Tray icons

~/.local/bin/fluistern      # Symlink to run the tool

Commands

fluistern                           # Toggle recording
systemctl --user start fluistern    # Start daemon
systemctl --user stop fluistern     # Stop daemon
systemctl --user status fluistern   # Check status
journalctl --user -u fluistern -f   # View logs

Uninstall

systemctl --user stop fluistern
systemctl --user disable fluistern
rm -rf ~/.local/share/fluistern ~/.local/bin/fluistern
rm ~/.config/systemd/user/fluistern.service

Troubleshooting

No tray icon

• Install yad: sudo pacman -S yad
• Make sure you have a system tray

Transcription errors

• Set specific language in tray menu (auto-detect can miss)
• Speak clearly, reduce background noise
• Check microphone selection

Slow processing

• Mostly upload time to Groq API
• Keep recordings under 30 seconds for <5s processing

License

MIT

Credits

• Groq - Fast Whisper API
• OpenAI Whisper - Speech recognition model