project-voice

name	project-voice
description	Generate unique voice hits for project completion sounds using 11 Labs API. Use when user asks to "create a project voice", "generate a voice hit", "make a completion sound", "set up project audio", or mentions voice generation for project notifications.

Project Voice Skill

Generate short, fun voice hits using 11 Labs API (V3 model with emotional tags) and configure project-specific completion sounds.

When to Use

• User invokes /project-voice
• User asks to create a unique project sound
• User wants to distinguish between multiple Claude Code sessions
• User mentions 11 Labs voice generation for completion notifications

Workflow

Step 1: Gather Context

Either:

• User provides text: Use their phrase directly
• Auto-generate: Analyze project context to create a fun phrase

For auto-generation, check:

1. Project folder name
2. README.md (if exists)
3. package.json name/description (if exists)

Generate a 2-5 word phrase that's fun, silly, and project-themed.

Step 2: Add Emotional Tag

Add an appropriate emotional tag from the V3 model. See references/emotional-tags.md for options.

Examples:

• [excited] Pixels assembled!
• [whisper] Peace arrived.
• [triumphant] Blastoff complete!

Step 3: Select Voice

Choose a voice from the catalog. See references/voice-catalog.md for options.

• Random selection for variety
• Or match voice style to project vibe (warm, energetic, calm, etc.)

Step 4: Generate Audio

Run the generation script:

python3 ~/.claude/skills/project-voice/scripts/generate_voice.py \
  --text "[excited] Your phrase here!" \
  --output-dir "$(pwd)/.claude/sounds" \
  --voice-id "21m00Tcm4TlvDq8ikWAM"  # Optional, random if omitted

The script will:

• Auto-version files (voice_v1.mp3, voice_v2.mp3, etc.)
• Never overwrite existing files
• Output JSON with the created file path

Step 5: Update Project Hook

Create or update .claude/settings.local.json in the project root:

{
  "hooks": {
    "Stop": [
      {
        "matcher": "",
        "hooks": [
          {
            "type": "command",
            "command": "afplay -v 0.1 /absolute/path/to/.claude/sounds/voice_v1.mp3"
          }
        ]
      }
    ]
  }
}

Important: Use the absolute path from the script output.

API Key Setup

1. Get an API key from https://elevenlabs.io/
2. Add it to ~/.claude/skills/project-voice/.env:

   ELEVENLABS_API_KEY=your_key_here
   

Script Reference

generate_voice.py [OPTIONS]

Options:
  --text TEXT          Phrase to vocalize (with emotional tags)
  --voice-id ID        Specific voice ID (optional, random if omitted)
  --output-dir DIR     Where to save the audio file
  --list-voices        Show available voices and exit

Example Session

User: /project-voice

Claude: I'll create a unique voice hit for this project.

Looking at the project... it's called "basketball-tracker" with a sports theme.

Generating: "[excited] Hoops on fire!"

[Runs script, gets voice_v1.mp3]

Updated .claude/settings.local.json with Stop hook.

Your project now has its own completion sound! When Claude Code finishes
in this project, you'll hear "Hoops on fire!" to know it's done.

References

• references/voice-catalog.md - Available voices with IDs and styles
• references/emotional-tags.md - V3 emotional tag reference