ElevenLabs Voice Personas v2.1

Comprehensive voice synthesis toolkit using ElevenLabs API.

🚀 First Run - Setup Wizard

When you first use this skill (no config.json exists), run the interactive setup wizard:

python3 scripts/setup.py

The wizard will guide you through:

1. API Key - Enter your ElevenLabs API key (required)
2. Default Voice - Choose from popular voices (Rachel, Adam, Bella, etc.)
3. Language - Set your preferred language (32 supported)
4. Audio Quality - Standard or high quality output
5. Cost Tracking - Enable usage and cost monitoring
6. Budget Limit - Optional monthly spending cap

🔒 Privacy: Your API key is stored locally in config.json only. It never leaves your machine and is automatically excluded from git via .gitignore.

To reconfigure at any time, simply run the setup wizard again.

---

✨ Features

• 18 Voice Personas - Carefully curated voices for different use cases
• 32 Languages - Multi-language synthesis with the multilingual v2 model
• Streaming Mode - Real-time audio output as it generates
• Sound Effects (SFX) - AI-generated sound effects from text prompts
• Batch Processing - Process multiple texts in one go
• Cost Tracking - Monitor character usage and estimated costs
• Voice Design - Create custom voices from descriptions
• Pronunciation Dictionary - Custom word pronunciation rules
• OpenClaw Integration - Works with OpenClaw's built-in TTS

---

🎙️ Available Voices

Voice	Accent	Gender	Persona	Best For
rachel	🇺🇸 US	female	warm	Conversations, tutorials
adam	🇺🇸 US	male	narrator	Documentaries, audiobooks
bella	🇺🇸 US	female	professional	Business, presentations
brian	🇺🇸 US	male	comforting	Meditation, calm content
george	🇬🇧 UK	male	storyteller	Audiobooks, storytelling
alice	🇬🇧 UK	female	educator	Tutorials, explanations
callum	🇺🇸 US	male	trickster	Playful, gaming
charlie	🇦🇺 AU	male	energetic	Sports, motivation
jessica	🇺🇸 US	female	playful	Social media, casual
lily	🇬🇧 UK	female	actress	Drama, elegant content
matilda	🇺🇸 US	female	professional	Corporate, news
river	🇺🇸 US	neutral	neutral	Inclusive, informative
roger	🇺🇸 US	male	casual	Podcasts, relaxed
daniel	🇬🇧 UK	male	broadcaster	News, announcements
eric	🇺🇸 US	male	trustworthy	Business, corporate
chris	🇺🇸 US	male	friendly	Tutorials, approachable
will	🇺🇸 US	male	optimist	Motivation, uplifting
liam	🇺🇸 US	male	social	YouTube, social media

🎯 Quick Presets

• default → rachel (warm, friendly)
• narrator → adam (documentaries)
• professional → matilda (corporate)
• storyteller → george (audiobooks)
• educator → alice (tutorials)
• calm → brian (meditation)
• energetic → liam (social media)
• trustworthy → eric (business)
• neutral → river (inclusive)
• british → george
• australian → charlie
• broadcaster → daniel (news)

---

🌍 Supported Languages (32)

The multilingual v2 model supports these languages:

Code	Language	Code	Language
en	English	pl	Polish
de	German	nl	Dutch
es	Spanish	sv	Swedish
fr	French	da	Danish
it	Italian	fi	Finnish
pt	Portuguese	no	Norwegian
ru	Russian	tr	Turkish
uk	Ukrainian	cs	Czech
ja	Japanese	sk	Slovak
ko	Korean	hu	Hungarian
zh	Chinese	ro	Romanian
ar	Arabic	bg	Bulgarian
hi	Hindi	hr	Croatian
ta	Tamil	el	Greek
id	Indonesian	ms	Malay
vi	Vietnamese	th	Thai

# Synthesize in German
python3 tts.py --text "Guten Tag!" --voice rachel --lang de

# Synthesize in French
python3 tts.py --text "Bonjour le monde!" --voice adam --lang fr

# List all languages
python3 tts.py --languages

---

💻 CLI Usage

Basic Text-to-Speech

# List all voices
python3 scripts/tts.py --list

# Generate speech
python3 scripts/tts.py --text "Hello world" --voice rachel --output hello.mp3

# Use a preset
python3 scripts/tts.py --text "Breaking news..." --voice broadcaster --output news.mp3

# Multi-language
python3 scripts/tts.py --text "Bonjour!" --voice rachel --lang fr --output french.mp3

Streaming Mode

Generate audio with real-time streaming (good for long texts):

# Stream audio as it generates
python3 scripts/tts.py --text "This is a long story..." --voice adam --stream

# Streaming with custom output
python3 scripts/tts.py --text "Chapter one..." --voice george --stream --output chapter1.mp3

Batch Processing

Process multiple texts from a file:

# From newline-separated text file
python3 scripts/tts.py --batch texts.txt --voice rachel --output-dir ./audio

# From JSON file
python3 scripts/tts.py --batch batch.json --output-dir ./output

JSON batch format:

[
  {"text": "First line", "voice": "rachel", "output": "line1.mp3"},
  {"text": "Second line", "voice": "adam", "output": "line2.mp3"},
  {"text": "Third line"}
]

Simple text format (one per line):

Hello, this is the first sentence.
This is the second sentence.
And this is the third.

Usage Statistics

# Show usage stats and cost estimates
python3 scripts/tts.py --stats

# Reset statistics
python3 scripts/tts.py --reset-stats

---

🎵 Sound Effects (SFX)

Generate AI-powered sound effects from text descriptions:

# Generate a sound effect
python3 scripts/sfx.py --prompt "Thunder rumbling in the distance"

# With specific duration (0.5-22 seconds)
python3 scripts/sfx.py --prompt "Cat meowing" --duration 3 --output cat.mp3

# Adjust prompt influence (0.0-1.0)
python3 scripts/sfx.py --prompt "Footsteps on gravel" --influence 0.5

# Batch SFX generation
python3 scripts/sfx.py --batch sounds.json --output-dir ./sfx

# Show prompt examples
python3 scripts/sfx.py --examples

Example prompts:

• "Thunder rumbling in the distance"
• "Cat purring contentedly"
• "Typing on a mechanical keyboard"
• "Spaceship engine humming"
• "Coffee shop background chatter"

---

🎨 Voice Design

Create custom voices from text descriptions:

# Basic voice design
python3 scripts/voice-design.py --gender female --age middle_aged --accent american \
  --description "A warm, motherly voice"

# With custom preview text
python3 scripts/voice-design.py --gender male --age young --accent british \
  --text "Welcome to the adventure!" --output preview.mp3

# Save to your ElevenLabs library
python3 scripts/voice-design.py --gender female --age young --accent american \
  --description "Energetic podcast host" --save "MyHost"

# List all design options
python3 scripts/voice-design.py --options

Voice Design Options:

Option	Values
Gender	male, female, neutral
Age	young, middle_aged, old
Accent	american, british, african, australian, indian, latin, middle_eastern, scandinavian, eastern_european
Accent Strength	0.3-2.0 (subtle to strong)

---

📖 Pronunciation Dictionary

Customize how words are pronounced:

Edit pronunciations.json:

{
  "rules": [
    {
      "word": "OpenClaw",
      "replacement": "Open Claw",
      "comment": "Pronounce as two words"
    },
    {
      "word": "API",
      "replacement": "A P I",
      "comment": "Spell out acronym"
    }
  ]
}

Usage:

# Pronunciations are applied automatically
python3 scripts/tts.py --text "The OpenClaw API is great" --voice rachel

# Disable pronunciations
python3 scripts/tts.py --text "The API is great" --voice rachel --no-pronunciations

---

💰 Cost Tracking

The skill tracks your character usage and estimates costs:

python3 scripts/tts.py --stats

Output:

📊 ElevenLabs Usage Statistics

  Total Characters: 15,230
  Total Requests:   42
  Since:            2024-01-15

💰 Estimated Costs:
  Starter    $4.57 ($0.30/1k chars)
  Creator    $3.66 ($0.24/1k chars)
  Pro        $2.74 ($0.18/1k chars)
  Scale      $1.68 ($0.11/1k chars)

---

🤖 OpenClaw TTS Integration

Using with OpenClaw's Built-in TTS

OpenClaw has built-in TTS support that can use ElevenLabs. Configure in ~/.openclaw/openclaw.json:

{
  "tts": {
    "enabled": true,
    "provider": "elevenlabs",
    "elevenlabs": {
      "apiKey": "your-api-key-here",
      "voice": "rachel",
      "model": "eleven_multilingual_v2"
    }
  }
}

Triggering TTS in Chat

In OpenClaw conversations:

• Use /tts on to enable automatic TTS
• Use the tts tool directly for one-off speech
• Request "read this aloud" or "speak this"

Using Skill Scripts from OpenClaw

# OpenClaw can run these scripts directly
exec python3 /path/to/skills/elevenlabs-voices/scripts/tts.py --text "Hello" --voice rachel

---

⚙️ Configuration

The scripts look for API key in this order:

1. ELEVEN_API_KEY or ELEVENLABS_API_KEY environment variable
2. OpenClaw config (~/.openclaw/openclaw.json → tts.elevenlabs.apiKey)
3. Skill-local .env file

Create .env file:

echo 'ELEVEN_API_KEY=your-key-here' > .env

---

🎛️ Voice Settings

Each voice has tuned settings for optimal output:

Setting	Range	Description
stability	0.0-1.0	Higher = consistent, lower = expressive
similarity_boost	0.0-1.0	How closely to match original voice
style	0.0-1.0	Exaggeration of speaking style

---

📝 Triggers

• "use {voice_name} voice"
• "speak as {persona}"
• "list voices"
• "voice settings"
• "generate sound effect"
• "design a voice"

---

📁 Files

elevenlabs-voices/
├── SKILL.md              # This documentation
├── README.md             # Quick start guide
├── config.json           # Your local config (created by setup, in .gitignore)
├── voices.json           # Voice definitions & settings
├── pronunciations.json   # Custom pronunciation rules
├── examples.md           # Detailed usage examples
├── scripts/
│   ├── setup.py          # Interactive setup wizard
│   ├── tts.py            # Main TTS script
│   ├── sfx.py            # Sound effects generator
│   └── voice-design.py   # Voice design tool
└── references/
    └── voice-guide.md    # Voice selection guide

---

🔗 Links

• ElevenLabs
• API Documentation
• Voice Library
• Sound Effects API
• Voice Design API

---

📋 Changelog

v2.1.0

• Added interactive setup wizard (scripts/setup.py)
• Onboarding guides through API key, voice, language, quality, and budget settings
• Config stored locally in config.json (added to .gitignore)
• Professional, privacy-focused setup experience

v2.0.0

• Added 32 language support with --lang parameter
• Added streaming mode with --stream flag
• Added sound effects generation (sfx.py)
• Added batch processing with --batch flag
• Added cost tracking with --stats flag
• Added voice design tool (voice-design.py)
• Added pronunciation dictionary support
• Added OpenClaw TTS integration documentation
• Improved error handling and progress output