ClawVox

Transform your OpenClaw assistant into a professional voice production studio with ClawVox - powered by ElevenLabs.

Quick Reference

Setup

1. Get your API key from elevenlabs.io/app/settings/api-keys
2. Configure in ~/.openclaw/openclaw.json:

{
  skills: {
    entries: {
      "clawvox": {
        apiKey: "YOUR_ELEVENLABS_API_KEY",
        config: {
          defaultVoice: "Rachel",
          defaultModel: "eleven_turbo_v2_5",
          outputDir: "~/.openclaw/audio"
        }
      }
    }
  }
}

Or set the environment variable:

export ELEVENLABS_API_KEY="your_api_key_here"

Voice Generation (TTS)

Basic Text-to-Speech

# Quick speak with default voice (Rachel)
{baseDir}/scripts/speak.sh 'Hello, I am your personal AI assistant.'

# Specify voice by name
{baseDir}/scripts/speak.sh --voice Adam 'Hello from Adam'

# Save to file
{baseDir}/scripts/speak.sh --out ~/audio/greeting.mp3 'Welcome to the show'

# Use specific model
{baseDir}/scripts/speak.sh --model eleven_multilingual_v2 'Bonjour'

# Adjust voice settings
{baseDir}/scripts/speak.sh --stability 0.5 --similarity 0.8 'Expressive speech'

# Adjust speed
{baseDir}/scripts/speak.sh --speed 1.2 'Faster speech'

# Use multilingual model for other languages
{baseDir}/scripts/speak.sh --model eleven_multilingual_v2 --voice Rachel 'Hola, que tal'
{baseDir}/scripts/speak.sh --model eleven_multilingual_v2 --voice Adam 'Guten Tag'

Voice Models

Available Voices

Premade voices: Rachel, Adam, Antoni, Bella, Domi, Elli, Josh, Sam, Callum, Charlie, George, Liam, Matilda, Alice, Bill, Brian, Chris, Daniel, Eric, Jessica, Laura, Lily, River, Roger, Sarah, Will

Long-Form Content

# Generate audio from text file
{baseDir}/scripts/speak.sh --input chapter.txt --voice "George" --out audiobook.mp3

Speech-to-Text (Transcription)

Basic Transcription

# Transcribe audio file
{baseDir}/scripts/transcribe.sh recording.mp3

# Save to file
{baseDir}/scripts/transcribe.sh --out transcript.txt audio.mp3

# Transcribe with language hint
{baseDir}/scripts/transcribe.sh --language es spanish_audio.mp3

# Include timestamps
{baseDir}/scripts/transcribe.sh --timestamps podcast.mp3

Supported Formats

• MP3, MP4, MPEG, MPGA, M4A, WAV, WebM
• Maximum file size: 100MB

Voice Cloning

Instant Voice Clone

# Clone from single sample (minimum 30 seconds recommended)
{baseDir}/scripts/clone.sh --name MyVoice recording.mp3

# Clone with description
{baseDir}/scripts/clone.sh --name BusinessVoice \
  --description 'Professional male voice' \
  sample.mp3

# Clone with labels
{baseDir}/scripts/clone.sh --name MyVoice \
  --labels '{"gender":"male","age":"adult"}' \
  sample.mp3

# Remove background noise during cloning
{baseDir}/scripts/clone.sh --name CleanVoice \
  --remove-bg-noise \
  sample.mp3

# Test cloned voice
{baseDir}/scripts/speak.sh --voice MyVoice 'Testing my cloned voice'

Voice Library Management

# List all available voices
{baseDir}/scripts/voices.sh list

# Get voice details
{baseDir}/scripts/voices.sh info --name Rachel
{baseDir}/scripts/voices.sh info --id 21m00Tcm4TlvDq8ikWAM

# Search voices (filter output with grep)
{baseDir}/scripts/voices.sh list | grep -i "female"

# Filter by category
{baseDir}/scripts/voices.sh list --category premade
{baseDir}/scripts/voices.sh list --category cloned

# Download voice preview
{baseDir}/scripts/voices.sh preview --name Rachel -o preview.mp3

# Delete custom voice
{baseDir}/scripts/voices.sh delete --id "voice_id"

Sound Effects

# Generate sound effect
{baseDir}/scripts/sfx.sh 'Heavy rain on a tin roof'

# With duration
{baseDir}/scripts/sfx.sh --duration 5 'Forest ambiance with birds'

# With prompt influence (higher = more accurate)
{baseDir}/scripts/sfx.sh --influence 0.8 'Sci-fi laser gun firing'

# Save to file
{baseDir}/scripts/sfx.sh --out effects/thunder.mp3 'Rolling thunder'

Note: Duration range is 0.5 to 22 seconds (rounded to nearest 0.5)

Voice Isolation

# Remove background noise and isolate voice
{baseDir}/scripts/isolate.sh noisy_recording.mp3

# Save to specific file
{baseDir}/scripts/isolate.sh --out clean_voice.mp3 meeting_recording.mp3

# Don't tag audio events
{baseDir}/scripts/isolate.sh --no-audio-events recording.mp3

Requirements:

• Minimum duration: 4.6 seconds
• Supported formats: MP3, WAV, M4A, OGG, FLAC

Dubbing (Multi-Language Translation)

# Dub audio to Spanish
{baseDir}/scripts/dub.sh --target es audio.mp3

# Dub with source language specified
{baseDir}/scripts/dub.sh --source en --target ja video.mp4

# Check dubbing status
{baseDir}/scripts/dub.sh --status --id "dubbing_id"

# Download dubbed audio
{baseDir}/scripts/dub.sh --download --id "dubbing_id" --out dubbed.mp3

Supported languages: en, es, fr, de, it, pt, pl, hi, ar, zh, ja, ko, nl, ru, tr, vi, sv, da, fi, cs, el, he, id, ms, no, ro, uk, hu, th

API Usage Examples

For direct API access, all scripts use curl under the hood:

# Direct TTS API call
curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/VOICE_ID" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello world", "model_id": "eleven_turbo_v2_5"}' \
  --output speech.mp3

Error Handling

All scripts provide helpful error messages:

• 401: Authentication failed - Check your API key
• 403: Permission denied - Your API key may not have access
• 429: Rate limit exceeded - Wait before trying again
• 500/502/503: ElevenLabs API issues - Try again later

Testing

Run the test suite to verify everything works:

{baseDir}/test.sh YOUR_API_KEY

Or with environment variable:

export ELEVENLABS_API_KEY="your_key"
{baseDir}/test.sh

Troubleshooting

Common Issues

1. "exec host not allowed (requested gateway)"

   • The skill needs to run commands in a sandbox environment
   • Configure OpenClaw to use sandbox: tools.exec.host: "sandbox"
   • Or enable sandboxing in your OpenClaw config
   • Alternative: Configure exec approvals for gateway host (see OpenClaw docs)

2. Parse errors with quotes or exclamation marks

   • Use single quotes instead of double quotes: 'Hello world' not "Hello world!"
   • Avoid exclamation marks (!) in text when using double quotes
   • For complex text, use the --input option with a file

3. "ELEVENLABS_API_KEY not set"

   • Ensure ELEVENLABS_API_KEY is set or configured in openclaw.json
   • Check that the API key is at least 20 characters long

4. "jq is required but not installed"

   • Install jq: apt-get install jq (Linux) or brew install jq (macOS)

5. "Rate limited"

   • Check your ElevenLabs plan quota at elevenlabs.io/app/usage
   • Free tier: ~10,000 characters/month

6. "Voice not found"

   • Use {baseDir}/scripts/voices.sh list to see available voices
   • Check if the voice ID is correct

7. "Dubbing failed"

   • Ensure source audio is clear and audible
   • Check supported language codes

8. "File too large"

   • Transcription: 100MB max
   • Dubbing: 500MB max
   • Voice cloning: 50MB per file

Debug Mode

# Enable verbose output
DEBUG=1 {baseDir}/scripts/speak.sh 'test'

# Show API request details
DEBUG=1 {baseDir}/scripts/transcribe.sh audio.mp3

Pricing Notes

ElevenLabs API pricing (approximate):

• Flash v2.5: ~$0.06/min
• Turbo v2.5: ~$0.06/min
• Multilingual v2: ~$0.12/min
• Voice cloning: Included in plan
• Sound effects: ~$0.02/generation
• Transcription: ~$0.02/min (Scribe v1)

Free tier: ~10,000 characters/month

Links

• ElevenLabs Dashboard
• API Documentation
• Voice Library
• Pricing