AI Speech to Text for AI Agency

Hold SHIFT to speak, release to type. Fast, accurate voice-to-text using Groq Whisper API.

Features

• Hold SHIFT to record - Release to transcribe and type
• Fast transcription - Groq Whisper API (~0.5s)
• High accuracy - Whisper large-v3-turbo model
• System tray icon - Right-click for settings
• Auto-hide widget - Only shows when recording
• Easy settings - Mic picker, API key input in UI
• Privacy focused - Audio processed by Groq, not stored

New in v1.2.0

• Accounting Mode - Converts spoken numbers to digits (e.g., "one hundred twenty three" → "123")
• Comma Formatting - Optional commas in large numbers (e.g., "1,234,567")
• Casual Mode - Lowercase output with informal punctuation
• Filter Words - Block unwanted phrases (e.g., "thank you" when nothing said)
• Blue Theme - Modern dark blue UI
• Custom Hotkeys - Change push-to-talk key in settings
• Emoji Voice Commands - Say "happy emoji" to insert 😊 (100+ emojis supported)

Two Versions Available

Version	File	Description
Full	VoiceType.exe	All features, system tray, emoji support
Lite	VoiceTypeLite.exe	Optimized for older/slower computers

Lite Version Differences

• Uses distil-whisper-large-v3-en model (faster)
• No system tray icon (less memory)
• No emoji conversion
• Simpler UI
• Smaller audio chunks
• Shares settings with Full version

Installation

Option 1: Pre-built Executables

Windows:

• Download VoiceType.exe (or VoiceTypeLite.exe for older computers) from the dist folder
• Double-click to run (no installation needed)

macOS:

• Download VoiceType.pkg
• Double-click it
• Click "Continue" then "Install"
• Done! VoiceType is in your Applications folder

Option 2: Run from Source

1. Clone the repo

   git clone https://github.com/yourusername/ai-speech-to-text.git
   cd ai-speech-to-text

2. Create virtual environment

   python -m venv venv

3. Activate venv

   • Windows: venv\Scripts\activate
   • Mac/Linux: source venv/bin/activate

4. Install dependencies

   pip install -r requirements.txt

5. Run

   python voice_type.py

   Or on Windows, just double-click run.bat

Building from Source

Windows

pip install pyinstaller
pyinstaller build/VoiceType.spec --noconfirm
# Or for Lite:
pyinstaller build/VoiceTypeLite.spec --noconfirm

The executable will be created at dist/VoiceType.exe

macOS

chmod +x build/build-mac.sh
./build/build-mac.sh

This creates:

• dist/VoiceType.app - The application bundle
• dist/VoiceType.pkg - PKG installer (share this file)

Note: You must build on the target platform. Windows builds only work on Windows, Mac builds only work on Mac.

Setup

1. Get a free API key from Groq Console
2. Right-click tray icon → Settings (or settings open automatically on first run)
3. Paste your API key
4. Select your microphone
5. Configure features (Accounting Mode, Casual Mode, Filter Words)
6. Click Save

Usage

1. Place cursor where you want text
2. Hold SHIFT and speak (widget appears)
3. Release SHIFT to transcribe
4. Text appears at cursor position
5. Widget auto-hides after 2 seconds

Features in Detail

Accounting Mode

When enabled, converts spoken number words to digits:

• "one" → "1"
• "twenty five" → "25"
• "one hundred" → "100"

Comma Formatting

When enabled with Accounting Mode, adds commas to large numbers:

• "1000000" → "1,000,000"

Casual Mode

When enabled, outputs lowercase text with informal punctuation:

• No capitalization
• Periods removed
• Multiple punctuation reduced

Filter Words

Block unwanted phrases from being typed. Useful for blocking:

• "thank you" (common hallucination when nothing said)
• "thanks"
• Any custom words

Enter as comma-separated list in settings.

Tips

• Speak clearly for best results
• Works in any app (IDEs, browsers, editors)
• Right-click tray icon for settings anytime (Full version only)
• Say "emoji" after an emoji name to insert it (Full version only)

Emoji Support (Full Version Only)

Speak emoji names to insert actual emojis! Just say the emoji name followed by "emoji":

Say This	Get This
"happy emoji"	😊
"sad emoji"	😢
"angry emoji"	😠
"laughing emoji"	😂
"heart emoji"	❤️
"fire emoji"	🔥
"thumbs up emoji"	👍
"thinking emoji"	🤔
"party emoji"	🎉
"rocket emoji"	🚀

Examples:

• "That's awesome fire emoji" → "That's awesome 🔥"
• "Great job thumbs up emoji" → "Great job 👍"
• "I'm confused thinking emoji" → "I'm confused 🤔"

Over 100+ emojis supported including emotions, animals, food, gestures, and more!

Requirements

• Python 3.8+
• Microphone
• Internet connection
• Groq API key (free tier available)

Security

• API keys are stored locally in ~/.voice-type-config.json
• No data is sent anywhere except Groq API for transcription
• Audio is processed in real-time and not saved to disk permanently

Version History

• v1.2.0 - Accounting mode, casual mode, filter words, blue theme, Lite version
• v1.1.0 - Emoji support, custom hotkeys
• v1.0.0 - Initial release

License

MIT