Claude Code Voice Integration

Add voice-to-text capabilities to Claude Code using OpenAI Whisper for speech recognition.

Features

• 🎤 Push-to-Talk Recording - Customizable hotkey (default: Right Shift)
• 📋 Auto-Clipboard Copy - Transcribed text automatically copied for easy pasting
• 🎯 Local Processing - Complete privacy with local Whisper transcription
• ⚡ Non-Intrusive - Claude Code runs normally with voice as background assistant
• 🔇 Smart Silence Detection - Auto-stops recording after silence
• 📊 Real-time Audio Levels - Visual feedback while recording

Installation

Requirements

• Python 3.9+
• macOS/Linux
• Microphone access
• Claude Code CLI installed

System Dependencies

macOS

brew install portaudio ffmpeg

Ubuntu/Debian

sudo apt-get update
sudo apt-get install portaudio19-dev python3-pyaudio ffmpeg

Setup

1. Create virtual environment and install dependencies:

   uv venv claude-voice-env
   uv pip install -r requirements.txt --python claude-voice-env/bin/python

2. Configure voice settings:

   ./claude-voice --configure-voice

Global Installation

chmod +x install.sh
# Add ~/.local/bin to PATH
./install.sh

Uninstall

chmod +x uninstall.sh
./uninstall.sh

Usage

Start Session

claude-voice

Voice Input Workflow

1. Start Recording: Hold your PTT key (default: Right Shift)
2. Speak: Talk naturally, audio levels show in real-time
3. Stop Recording: Release PTT key or stay silent (auto-stops after 1.5s)
4. Auto-copy: Transcribed text copied to clipboard automatically
5. Paste: Press Cmd+V (or Ctrl+V) in Claude

🎤 Speak → 📋 Auto-copy → Cmd+V Paste → 🚀 Claude responds

Example Session

claude-voice

Claude Code with Voice Support
========================================
🎙️ Voice input enabled! Press right_shift to speak
Starting Claude Code...

# Hold Right Shift and say: "analyze the main.py file"
🎤 Recording... (Release key or stay silent to stop)
  Level: ████████████░░░░░░░░░░░░░░░░░░ 2847

🎤 Voice transcribed: analyze the main.py file
📋 Copied to clipboard! Just paste into Claude (Cmd+V)

# Just paste (Cmd+V) into Claude and it processes normally

Configuration Options

Push-to-Talk Keys

• right_shift - Default, convenient single-key PTT ⭐
• f13 - If you have function keys beyond F12
• caps_lock - Repurpose caps lock
• left_shift - Alternative shift key

Whisper Models

• tiny - Fastest, lowest accuracy (39M parameters)
• base - Good balance (74M parameters) ⭐ Recommended
• small - Better accuracy (244M parameters)
• medium - High accuracy (769M parameters)
• large - Best accuracy (1550M parameters)

Silence Settings

• Duration: How long to wait for silence before auto-stopping
  • 0.5s - Quick stops, sensitive to pauses
  • 1.5s - Default, allows natural pauses ⭐
  • 3.0s - Slower stops, good for thinking pauses
• Threshold: Audio level sensitivity (lower = more sensitive)
  • 50 - Very sensitive (quiet environments)
  • 100 - Balanced ⭐ Recommended
  • 200 - Less sensitive (noisy environments)
  • 500 - Much less sensitive (very noisy)

Tips & Best Practices

• Speak Clearly: Enunciate commands clearly
• Quiet Environment: Reduces false transcriptions
• Consistent Distance: Keep consistent distance from microphone
• Command Patterns: Use consistent phrasing for common commands
• Shortcuts: Create aliases for frequently used voice commands

Privacy & Security

This implementation provides complete privacy and security:

• All audio processing happens locally on your machine
• No data is sent to external servers
• Complete privacy of your voice and code
• Perfect for sensitive development environments