Advanced voice dictation app with AI-powered formatting, command mode, and auto-paste
A modern, feature-rich voice-to-text app built with Flutter. Optimized for Windows with support for macOS and Linux. Uses Groq's Whisper API for transcription and LLaMA 3.3 for intelligent text formatting.
- π€ Voice Dictation - Record and transcribe speech to text instantly
- β¨οΈ Global Keyboard Shortcuts - Start/stop recording from anywhere with Ctrl+Shift+R (Windows/Linux) or Cmd+Shift+R (macOS)
- π Notes History - All transcriptions are automatically saved with full editing capabilities
- π Search & Filter - Quickly find notes with powerful search
- βοΈ Rich Note Editor - Edit, favorite, pin, and organize your notes
- π Custom Dictionary - Add corrections to auto-fix common transcription mistakes
- π Auto-Paste - Automatically pastes transcription into active window (Windows)
- π― Command Mode - Use voice commands to manipulate text:
- "Fasse zusammen" / "Summarize" - Summarize text
- "Γbersetze nach Englisch" / "Translate to English" - Translate
- "LΓΆsche den letzten Satz" / "Delete last sentence" - Delete
- "KΓΌrze das" / "Shorten" - Make text shorter
- "Erweitere das" / "Expand" - Add more details
- "Mach das formal" / "Make formal" - Rewrite formally
- "Korrigiere" / "Fix grammar" - Fix grammar mistakes
- π§Ή Filler Word Filtering - Automatically removes "vielen Dank", "Γ€h", "um", etc.
- π Note Pinning - Pin important notes to the top
- β Favorites - Mark notes as favorites for quick access
- π€ AI Formatting - Automatic punctuation, capitalization, and paragraph breaks
- π£οΈ Voice Commands - "Absatz" for paragraphs, "in AnfΓΌhrungszeichen" for quotes, etc.
- π Multi-Language - Supports German, English, Spanish, French, Italian, and Portuguese
- π Clipboard Integration - Automatically copies formatted text to clipboard
- π¨ Modern UI - WhisperFlow-inspired interface with sidebar navigation
- π Dark Mode - Automatic dark/light theme support
- Flutter SDK - Download and install
- Visual Studio 2022 with "Desktop development with C++" workload
- Groq API Key - Free at console.groq.com
Easiest way - just double-click build.bat!
Or manually:
flutter pub get
flutter build windows --releaseThe executable will be at build\windows\x64\runner\Release\fluistern_app.exe
π See WINDOWS_BUILD.md for detailed Windows build instructions
- Launch the app
- Go to Settings (βοΈ icon in sidebar)
- Enter your Groq API key
- Select your preferred language
- Configure features (Auto-Paste, Command Mode, etc.)
- Click "Save Settings"
- Click the blue microphone button (or press Ctrl+Shift+R)
- Speak your text
- Click the red stop button when done
- The app will:
- Transcribe your speech using Whisper
- Format it with AI
- Apply corrections from your dictionary
- Filter out filler words (if enabled)
- Save as a note
- Copy to clipboard (or auto-paste if enabled)
Say commands during or after dictation:
German Commands:
- "Fasse zusammen" - Summarize the text
- "Γbersetze nach Englisch" - Translate to English
- "LΓΆsche den letzten Satz" - Delete last sentence
- "KΓΌrze das" - Make shorter
- "Erweitere das" - Add more details
- "Mach das formal" - Rewrite formally
- "Mach das informell" - Rewrite casually
- "Korrigiere" - Fix grammar
English Commands:
- "Summarize" - Summarize the text
- "Translate to German" - Translate to German
- "Delete last sentence" - Delete last sentence
- "Shorten" - Make shorter
- "Expand" - Add more details
- "Make formal" - Rewrite formally
- "Make informal" - Rewrite casually
- "Fix grammar" - Fix grammar
- "Absatz" / "Paragraph" - Insert paragraph break
- "in AnfΓΌhrungszeichen" - Quote the previous word/phrase
- "Komma" - Insert comma
- "Punkt" - Insert period
- "Fragezeichen" - Insert question mark
- "Ausrufezeichen" - Insert exclamation mark
When enabled in Settings, transcriptions are automatically pasted into your active window:
- Have any app open (Word, Slack, Email, etc.)
- Press Ctrl+Shift+R to record
- Speak and stop recording
- Text is automatically pasted into your app!
- Search - Type in the search bar to find notes
- Edit - Click any note to view and edit
- Pin - Click the pin icon to pin notes to the top
- Favorite - Click the star icon to mark as favorite
- Delete - Click the trash icon to delete a note
- Copy - Click the copy icon to copy note content
- Go to Dictionary screen (π icon in sidebar)
- Click the + button to add a correction
- Enter the wrong text (e.g., "teh")
- Enter the correct text (e.g., "the")
- All future transcriptions will auto-correct!
See WINDOWS_BUILD.md for complete Windows instructions
Quick build:
flutter pub get
flutter build windows --releaseOr use the build script:
build.batSystem Dependencies (required for building):
# Ubuntu/Debian/Linux Mint
sudo apt-get update
sudo apt-get install build-essential lld libkeybinder-3.0-dev
# Arch Linux
sudo pacman -S base-devel lld keybinder3
# Fedora
sudo dnf install gcc-c++ lld keybinder3-develNote: If you don't want to use global hotkeys on Linux (e.g., your window manager doesn't support keybinder3), you can disable them in Settings after building.
Build:
flutter pub get
flutter build linux --releaseBinary will be in build/linux/x64/release/bundle/
flutter pub get
flutter build macos --releaseNote: macOS builds are supported but untested. Auto-paste is not yet implemented for macOS.
- Groq API Key - Your API key from console.groq.com
- Language - Select transcription language (German, English, Spanish, French, Italian, Portuguese)
- Auto-Paste - Automatically paste into active window (Windows only)
- Command Mode - Enable voice commands for text manipulation
- Filler Word Filtering - Remove common filler words from transcriptions
- Enable Global Hotkeys - Allow Ctrl+Shift+R from anywhere
- Disable this on Linux if keybinder3 is not available
Recording β Whisper API β LLM Formatting β Corrections β Filler Filtering β Command Mode β Auto-Paste
- Record: Audio is recorded in m4a format (16kHz, mono)
- Transcribe: Groq's Whisper API converts speech to text
- Format: LLaMA 3.3 70B adds punctuation, capitalization, and handles voice commands
- Correct: Custom dictionary applies your corrections
- Filter: Remove filler words (if enabled)
- Command: Execute voice commands (if enabled)
- Paste: Auto-paste into active window (if enabled) or copy to clipboard
lib/
βββ main.dart # App entry point
βββ screens/
β βββ main_layout.dart # Navigation sidebar
β βββ notes_screen.dart # Main recording & notes list
β βββ note_detail_screen.dart # Note editor
β βββ dictionary_screen.dart # Corrections management
β βββ settings_screen.dart # Settings & configuration
βββ services/
β βββ audio_service.dart # Audio recording
β βββ groq_api_service.dart # Whisper & LLM integration
β βββ storage_service.dart # Settings persistence
β βββ database_service.dart # SQLite database
β βββ filler_filter_service.dart # Filler word removal
β βββ auto_paste_service.dart # Auto-paste functionality
β βββ command_mode_service.dart # Voice command processing
βββ models/
β βββ note.dart # Note data model
β βββ correction.dart # Correction data model
βββ providers/
βββ app_provider.dart # State management
- Build fails: Make sure Visual Studio 2022 is installed with C++ tools
- App doesn't start: Run from command prompt to see error messages
- Auto-paste doesn't work: Make sure Auto-Paste is enabled in Settings
- CMake error about keybinder-3.0: Install libkeybinder-3.0-dev (see Installation section)
- ld.lld not found: Install lld and build-essential packages
- Hotkeys don't work: Disable global hotkeys in Settings
- "API key not set" error: Go to Settings and enter your Groq API key
- Transcription is inaccurate: Try speaking more clearly, or adjust microphone settings
- Commands not working: Make sure Command Mode is enabled in Settings
- macOS auto-paste support
- Android version
- Custom hotkey configuration
- Export notes to various formats (PDF, Markdown, etc.)
- Voice text selection ("mark last sentence", "copy that", etc.)
- Cloud sync
- Multiple language support in single session
- Offline mode with local Whisper
MIT
- Groq - Fast Whisper API & LLaMA inference
- Flutter - Cross-platform framework
- Inspired by WhisperFlow
- fluistern - Original Linux-only version (Bash/Python)
Made with β€οΈ for productive voice dictation