Voice Cloning

Create custom voices for your AI assistants using voice cloning technology

Voice cloning allows you to create custom voices from audio samples, giving your AI assistants unique and personalized voices that match your brand or specific use cases.

---

Overview

Burki Voice AI’s voice cloning feature enables you to:

• Upload Voice Samples: Upload high-quality audio recordings to create voice models
• Multi-Provider Support: Use ElevenLabs, Resemble AI, and other providers that support voice cloning
• Instant Voice Creation: Generate cloned voices ready for immediate use
• Voice Management: Organize, test, and manage your custom voices
• Usage Analytics: Track voice usage for billing and optimization

🎙️ Voice Sample Upload

Upload audio samples with validation and processing

🤖 AI Voice Training

Provider-powered voice training with quality optimization

📊 Usage Analytics

Track synthesis usage and voice performance

🔧 Easy Integration

Seamless integration with existing assistant configurations

---

Supported Providers

ElevenLabs

• Instant Voice Cloning: Create voices from single audio samples
• High Quality: Professional-grade voice synthesis
• Multiple Languages: Support for 29+ languages
• Quick Processing: Voices ready in seconds

Resemble AI

• Professional Training: Advanced voice training algorithms
• Custom Models: Highly personalized voice characteristics
• Unlimited Voices: Create as many voices as needed
• Enterprise Features: Advanced customization options

Future Providers

• Inworld AI: Coming soon with emotional voice cloning
• OpenAI: Voice cloning capabilities when available

---

Voice Sample Requirements

Audio Quality Guidelines

File Format Requirements

Supported Formats:

• MP3 (recommended)
• WAV (highest quality)
• FLAC (lossless)
• M4A/AAC
• OGG

Technical Specifications:

• Sample Rate: 22kHz or higher
• Bit Rate: 128kbps minimum
• Channels: Mono preferred, stereo acceptable
• File Size: Maximum 50MB

Recording Guidelines

Duration Requirements:

• Minimum: 10 seconds of clear speech
• Recommended: 30-60 seconds for better quality
• Maximum: 10 minutes (longer samples may not improve quality)

Content Guidelines:

• Clear Speech: No background noise or music
• Natural Tone: Conversational, not monotone
• Consistent Volume: Steady audio levels throughout
• Single Speaker: Only the target voice in the recording

Quality Tips

For Best Results:

1. Environment: Record in a quiet room with soft furnishings
2. Microphone: Use a quality microphone 6-12 inches from mouth
3. Content: Read varied sentences with different emotions
4. Consistency: Maintain the same speaking style throughout
5. Format: Save in WAV format for highest quality

---

Getting Started

Step 1: Upload Voice Sample

Navigate to your assistant’s configuration and open the Voice Cloning section:

1. Upload Audio File: Drag and drop or click to select your audio file
2. Add Metadata: Provide a name, description, and tags
3. Validation: System automatically validates audio quality
4. Processing: File is uploaded and prepared for cloning

Example Upload

const formData = new FormData();
formData.append('file', audioFile);
formData.append('name', 'Professional Voice');
formData.append('description', 'Clear, professional speaking voice');
formData.append('tags', 'professional, clear, business');

const response = await fetch('/assistants/123/voice-samples/upload', {
  method: 'POST',
  body: formData,
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY'
  }
});

Step 2: Create Cloned Voice

Once your sample is uploaded, create a cloned voice:

1. Select Provider: Choose ElevenLabs or Resemble AI
2. Configure Options: Set voice name, language, and quality settings
3. Initiate Cloning: Start the voice training process
4. Monitor Progress: Track cloning status in real-time

Example Voice Creation

const response = await fetch('/assistants/123/cloned-voices/create', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'Authorization': 'Bearer YOUR_API_KEY'
  },
  body: JSON.stringify({
    voice_sample_id: 456,
    provider: 'elevenlabs',
    name: 'Custom Professional Voice',
    language: 'en',
    enhance_quality: true
  })
});

Step 3: Use Cloned Voice

Once processing is complete, assign the voice to your assistant:

1. Voice Selection: Choose from your cloned voices
2. Testing: Preview the voice with sample text
3. Assignment: Set as the assistant’s default voice
4. Go Live: Start using the voice in live calls

---

Voice Management

Voice Library

Organizing Voices

Voice Categories:

• Brand Voices: Official company voices
• Character Voices: Specific personas or characters
• Language Variants: Same voice in different languages
• Seasonal/Campaign: Temporary or promotional voices

Tagging System:

• Use consistent tags for easy filtering
• Include language, gender, style descriptors
• Add use case tags (customer service, sales, etc.)

Voice Analytics

Usage Tracking:

• Synthesis Count: Number of times voice was used
• Duration Metrics: Total audio generated
• Cost Tracking: Provider usage and billing
• Performance: Quality scores and user feedback

Optimization Insights:

• Most/least used voices
• Cost per synthesis by provider
• Quality trends over time
• User preference patterns

Voice Testing

Test your cloned voices before deployment:

1. Text-to-Speech Preview: Enter sample text to hear the voice
2. Quality Assessment: Evaluate clarity, naturalness, and accuracy
3. Comparison Testing: Compare with original samples and other voices
4. A/B Testing: Test different voices with real users

---

API Integration

Upload Voice Sample

cURL

curl -X POST "https://api.burki.dev/assistants/123/voice-samples/upload" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@voice_sample.wav" \
  -F "name=Professional Voice" \
  -F "description=Clear professional speaking voice"

Create Cloned Voice

cURL

curl -X POST "https://api.burki.dev/assistants/123/cloned-voices/create" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "voice_sample_id": 456,
    "provider": "elevenlabs",
    "name": "Custom Professional Voice",
    "language": "en",
    "enhance_quality": true
  }'

List Cloned Voices

cURL

curl -X GET "https://api.burki.dev/assistants/123/cloned-voices" \
  -H "Authorization: Bearer YOUR_API_KEY"

---

Best Practices

Recording Quality

Professional Recording Setup

Equipment Recommendations:

• Microphone: USB condenser microphone (Audio-Technica AT2020USB+)
• Environment: Quiet room with minimal echo
• Software: Audacity, GarageBand, or professional DAW
• Monitoring: Use headphones to monitor audio quality

Recording Techniques:

1. Consistent Distance: Maintain 6-12 inches from microphone
2. Proper Levels: Keep audio peaks between -12dB and -6dB
3. Room Treatment: Use blankets or acoustic foam to reduce echo
4. Multiple Takes: Record several versions and choose the best

Content Selection

Ideal Voice Sample Content:

• Varied Sentences: Different sentence structures and lengths
• Emotional Range: Include slight variations in tone
• Natural Speech: Conversational, not reading tone
• Complete Thoughts: Full sentences with natural pauses

What to Avoid:

• Background noise or music
• Multiple speakers
• Heavy accents (unless desired)
• Monotone or robotic delivery
• Incomplete sentences or stuttering

Voice Management

1. Naming Convention: Use descriptive, consistent names
2. Version Control: Keep track of voice iterations and improvements
3. Usage Documentation: Document which voices work best for different scenarios
4. Regular Testing: Periodically test voice quality and user satisfaction
5. Cost Monitoring: Track usage and costs across different providers

Security and Privacy

Privacy Considerations: Voice cloning involves processing personal audio data. Ensure you have proper consent and follow privacy regulations when using voice samples.

• Consent: Always obtain explicit consent before using someone’s voice
• Data Protection: Store voice samples securely and follow GDPR/CCPA requirements
• Access Control: Limit who can create and manage cloned voices
• Audit Trail: Keep logs of voice creation and usage
• Retention Policy: Define how long voice samples and models are stored

---

Troubleshooting

Common Issues

Upload Problems

File Upload Fails:

• Check file format is supported (MP3, WAV, FLAC, M4A, OGG)
• Ensure file size is under 50MB
• Verify audio duration is between 10 seconds and 10 minutes
• Check internet connection stability

Audio Quality Issues:

• Use higher sample rate (22kHz+) and bit rate (128kbps+)
• Remove background noise using audio editing software
• Re-record in a quieter environment
• Check microphone positioning and levels

Voice Creation Issues

Cloning Process Fails:

• Verify provider API credentials are valid
• Check account balance with voice cloning provider
• Ensure voice sample meets provider requirements
• Contact provider support for specific error messages

Poor Voice Quality:

• Use higher quality source audio
• Try different provider (ElevenLabs vs Resemble AI)
• Experiment with quality enhancement settings
• Consider recording new samples with better equipment

Performance Issues

Slow Processing:

• Provider processing times vary (ElevenLabs: seconds, Resemble: minutes)
• Check provider status pages for service issues
• Large files take longer to process
• Peak usage times may cause delays

High Costs:

• Monitor usage through analytics dashboard
• Set usage limits and alerts
• Compare provider pricing for your use case
• Optimize voice selection for cost efficiency

---

Provider Comparison

Feature	ElevenLabs	Resemble AI	Coming Soon
Processing Time	Seconds	Minutes	Varies
Quality	Excellent	Excellent	TBD
Languages	29+	English+	TBD
Cost Model	Per character	Per synthesis	TBD
Sample Requirements	30s+	60s+	TBD
Instant Preview	✅	❌	TBD
Emotional Control	Basic	Advanced	TBD
Enterprise Features	Limited	Full	TBD

---

Use Cases

Customer Service

• Consistent Brand Voice: Maintain brand identity across all interactions
• Multilingual Support: Create voices in different languages for global support
• Personality Matching: Match voice characteristics to brand personality

Sales and Marketing

• Campaign Voices: Create specific voices for marketing campaigns
• Regional Variants: Adapt voices for different geographical markets
• Seasonal Adjustments: Modify voice characteristics for holidays or events

Entertainment and Media

• Character Voices: Create unique voices for virtual characters
• Narrator Voices: Professional voices for content narration
• Interactive Experiences: Engaging voices for games and interactive media

Enterprise Applications

• Executive Voices: Clone executive voices for consistent communication
• Training Systems: Consistent voices for e-learning and training
• Brand Ambassadors: Virtual representatives with authentic brand voices

---

Getting Help

📖 Documentation

Complete TTS provider documentation

🎛️ Voice Tuning

Advanced voice configuration guide

💬 Community Support

Get help from the community

🎧 Technical Support

Contact our support team

---

Pro Tip: Start with ElevenLabs for quick prototyping and testing, then consider Resemble AI for production deployments requiring advanced customization and enterprise features.