Typr

Elegant dictation with press-and-hold recording powered by OpenAI's Whisper and GPT-4o-mini.

Features

• 🎙️ Press & Hold Recording: Natural keyboard shortcut workflow
• 🗣️ Smart Transcription: Local Whisper or OpenAI API transcription
• 🤖 Intelligent Processing: GPT-4o-mini processes "note to the editor" commands
• ⌨️ Seamless Typing: Cross-platform text insertion
• ⚙️ Simple Setup: JSON-based configuration
• 🔄 Process-Based: Elegant inter-process communication

Quick Start

1. Install

# Make sure Deno is installed
curl -fsSL https://deno.land/install.sh | sh

# Build and install
./build.sh

2. Configure

typr config  # Check current configuration

Add your OpenAI API key to ~/.typr-settings.json (auto-created on first run).

3. Setup Keyboard Shortcut

typr shortcuts

Follow the instructions to bind a keyboard shortcut to typr toggle.

4. Start Recording!

• Press your shortcut: Starts recording (you'll hear a beep)
• Release your shortcut: Stops, transcribes, and types the result

Commands

• typr config - Show current configuration and status
• typr shortcuts - Show keyboard shortcut setup instructions
• typr record - One-time recording and transcription
• typr toggle - Toggle recording (used by shortcuts)

System Requirements

For Audio Recording

• macOS: Built-in (uses AVFoundation)
• Linux: ffmpeg with PulseAudio support
• Windows: ffmpeg with DirectShow support

For Text Typing

• macOS: Built-in (uses AppleScript)
• Linux: xdotool package
• Windows: Built-in (uses PowerShell)

For Local Whisper (Optional)

pip install openai-whisper

Installation

Install Dependencies

Linux (Ubuntu/Debian):

sudo apt update
sudo apt install ffmpeg xdotool

Linux (Arch/Manjaro):

sudo pacman -S ffmpeg xdotool

macOS:

brew install ffmpeg

Windows:

# Install ffmpeg via chocolatey or download from https://ffmpeg.org/
choco install ffmpeg

Install Typr Globally

# After building
sudo mv ./dist/typr /usr/local/bin/
# or on Windows, add to PATH

Configuration

Settings are stored in ~/.typr-settings.json:

{
  "openAIKey": "your-api-key",
  "whisperPrompt": "Custom vocabulary context...",
  "llmPrompt": "Custom processing instructions...",
  "useLocalWhisper": false
}

Usage Examples

Basic Recording

typr record
# Speak into microphone, press Enter to stop
# Text will be transcribed and optionally typed

With Custom Instructions

Say "note to the editor" followed by instructions during recording:

"Please write a professional email about the quarterly results. Note to the editor: make it formal and include bullet points."

The GPT model will process your transcription according to your custom LLM prompt.

Interactive Mode

typr start
# Press 'r' to record, 'q' to quit

Differences from Desktop Version

The CLI version maintains all core functionality but with some changes:

✅ Preserved:

• OpenAI Whisper/GPT integration
• Local Whisper support
• Custom prompts and vocabulary
• Cross-platform text typing
• Same transcription quality

🔄 Changed:

• Process-based architecture instead of daemon mode
• Keyboard shortcut triggers process spawning
• Elegant signal-based inter-process communication
• Simplified setup with clear instructions

📦 Improved:

• No background processes or daemons
• Robust process coordination via PID files
• Natural press-and-hold workflow
• Better resource usage (only runs when needed)

Development

The CLI is implemented in a single TypeScript file (typr.ts) using Deno:

# Run directly with Deno
deno run --allow-all typr.ts setup

# Build binary
deno compile --allow-all --output dist/typr typr.ts

Troubleshooting

Audio Recording Issues

• Ensure ffmpeg is installed and in PATH
• Check microphone permissions
• Test with: ffmpeg -f [format] -i [input] -t 5 test.wav

Transcription Issues

• Verify OpenAI API key is valid
• For local Whisper: pip install openai-whisper
• Check internet connection for API calls

Text Typing Issues

• Linux: Install xdotool
• macOS: Grant accessibility permissions if prompted
• Windows: Run as administrator if needed

License

Same as original project.