Claude Voice — Gaming Companion

A voice-activated AI assistant powered by Claude, with wake word detection, push-to-talk, local speech-to-text (Whisper), and local text-to-speech (Piper).

Prerequisites

• Python 3.10+
• Piper TTS installed and on your PATH
• A working microphone and speakers/headphones
• An Anthropic API key
• Linux (uses evdev for push-to-talk key detection)

Setup

1. Clone the repo and open it in VS Code / Cursor:

   git clone https://github.com/pulsipd/claude-voice.git
   cd claude-voice
   code .   # or: cursor .

2. Create a virtual environment and install dependencies:

   Open a terminal in VS Code/Cursor (`Ctrl+``) and run:

   python3 -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt

3. Set your Anthropic API key:

   export ANTHROPIC_API_KEY="sk-ant-..."

   To make it permanent, add that line to your ~/.bashrc or ~/.zshrc.

4. Install Piper TTS:

   pip install piper-tts

   Or follow the install instructions at https://github.com/rhasspy/piper.

Running

With the virtual environment activated:

python3 claude-voice.py

On first run it will automatically download:

• A Whisper speech-to-text model (~150 MB)
• A Piper voice (~30 MB)

These are cached in the models/ directory.

Usage

The app has two input modes:

• Voice mode (default) — Say the wake word (default: "Hey Jarvis") to start listening. It stops recording after detecting silence.
• Push-to-talk mode — Hold the Scroll Lock key to record, release to send.

You can also type questions directly into the terminal.

Commands

Command	Description
/help	Show all commands
/quit	Exit the app
/clear	Clear conversation history
/model [name]	Show or set Claude model (sonnet, opus, haiku)
/mode [voice|ptt]	Switch between voice and push-to-talk
/wake [model]	List or change wake word model
/voice [name]	List or switch TTS voice
/voice download <name>	Download a new Piper voice
/devices	List audio input/output devices
/input <id>	Set audio input device
/output <id>	Set audio output device

Available Voices

Run /voice to see all options. Some highlights:

Name	Description
lessac	US female, clear and neutral
amy	US female, warm and friendly
joe	US male, casual and upbeat
alan	British male, deep and composed
jenny	British female, expressive
ryan	US male, broadcaster style

Download with /voice download <name>, then switch with /voice <name>.

Troubleshooting

• "ANTHROPIC_API_KEY not set" — Make sure you exported the key in the same terminal session.
• "piper not found" — Install with pip install piper-tts or check it's on your PATH.
• No audio input — Run /devices to list devices, then /input <id> to pick the right mic.
• Wake word not working — Try /mode ptt as a fallback or /wake to see available models.
• Permission error on key detection — Push-to-talk uses evdev and may need your user in the input group: sudo usermod -aG input $USER (then log out/in).

License

MIT