Eve library

Voice Design Guide

Guide

Revision 1

On this page

This guide covers designing and configuring voices for meditation, spiritual content, and other audio experiences.

Overview#

Voice design involves defining:

  • Voice characteristics (pitch, pace, timbre)
  • Personality traits and speaking style
  • Tradition-specific voice profiles
  • Prosody and emotional expression
  • Model and settings selection

Voice Characteristics#

Core Parameters#

typescript
interface VoiceCharacteristics {
  gender: VoiceGender;      // 'male' | 'female' | 'neutral' | 'androgynous'
  age: VoiceAge;            // 'young' | 'young_adult' | 'middle_aged' | 'mature' | 'elderly'
  pitch: VoicePitch;        // 'very_low' to 'very_high'
  pace: VoicePace;          // 'very_slow' to 'very_fast'
  timbre: VoiceTimbre;      // 'warm' | 'bright' | 'dark' | 'resonant' | etc.
  energy: VoiceEnergy;      // 'calm' | 'relaxed' | 'neutral' | 'engaged' | etc.
  warmth: number;           // 0-1
  clarity: number;          // 0-1
  breathiness: number;      // 0-1
  resonance: number;        // 0-1
  expressiveness: number;   // 0-1
}

Gender Options#

Gender Description Best For
male Masculine voice Wide range of content
female Feminine voice Wide range of content
neutral Neither strongly masculine nor feminine Inclusive content
androgynous Balanced masculine/feminine Spiritual, non-gendered

Age Characteristics#

Age Typical Qualities Content Match
young Bright, energetic, clear Kids content, energizing
young_adult Fresh, dynamic Motivational, fitness
middle_aged Balanced, authoritative General content
mature Warm, experienced Wisdom, meditation
elderly Deep, wise, measured Elder wisdom, traditions

Pitch Settings#

Pitch Description Use Cases
very_low Deep bass Grounding, masculine
low Lower register Authority, calm
medium_low Slightly low Natural, settled
medium Middle range Neutral, versatile
medium_high Slightly high Engaging, friendly
high Higher register Light, feminine
very_high Upper range Children, energetic

Pace Guidelines#

Pace Words/Min Best For
very_slow ~80 Deep meditation, sleep
slow ~100 Meditation, relaxation
measured ~120 Contemplation
medium ~140 General content
moderate ~160 Conversational
fast ~180 Energizing, excitement
very_fast ~200+ Urgency (rare)

Timbre Qualities#

Timbre Character Emotion
warm Friendly, comforting Safety, nurturing
bright Clear, uplifting Joy, energy
dark Deep, mysterious Depth, gravitas
resonant Full, ringing Authority, presence
breathy Airy, intimate Vulnerability, closeness
clear Crisp, distinct Instruction, clarity
husky Textured, raw Emotion, authenticity
rich Full-bodied Warmth, depth
thin Light, delicate Subtlety, gentleness
full Complete, round Completeness

Energy Levels#

Energy Character Content Type
calm Still, peaceful Sleep, deep meditation
relaxed Ease, comfort Relaxation, wind-down
neutral Balanced General content
engaged Attentive, present Active meditation
energetic Lively, dynamic Morning meditation, motivation
passionate Intense, heartfelt Devotional, inspiring
intense Powerful, focused Transformation, breakthrough

Voice Profiles by Content Type#

Sleep Content#

typescript
const sleepVoice: VoiceCharacteristics = {
  gender: 'neutral',
  age: 'mature',
  pitch: 'low',
  pace: 'very_slow',
  timbre: 'warm',
  energy: 'calm',
  warmth: 0.9,
  clarity: 0.7,
  breathiness: 0.6,
  resonance: 0.7,
  expressiveness: 0.3
};

Key traits:

  • Low pitch creates safety and grounding
  • Very slow pace allows for drifting off
  • High warmth for comfort
  • Lower expressiveness for monotonous, sleep-inducing quality

Meditation Content#

typescript
const meditationVoice: VoiceCharacteristics = {
  gender: 'female',
  age: 'mature',
  pitch: 'medium_low',
  pace: 'slow',
  timbre: 'warm',
  energy: 'relaxed',
  warmth: 0.85,
  clarity: 0.8,
  breathiness: 0.4,
  resonance: 0.75,
  expressiveness: 0.5
};

Key traits:

  • Balanced but slightly low pitch
  • Slow but not sleep-inducing pace
  • High warmth and clarity
  • Moderate expressiveness for guidance

Kids Content#

typescript
const kidsVoice: VoiceCharacteristics = {
  gender: 'female',
  age: 'young_adult',
  pitch: 'medium_high',
  pace: 'moderate',
  timbre: 'bright',
  energy: 'engaged',
  warmth: 0.8,
  clarity: 0.9,
  breathiness: 0.2,
  resonance: 0.6,
  expressiveness: 0.8
};

Key traits:

  • Higher pitch for approachability
  • Moderate pace for engagement
  • Bright timbre for energy
  • High expressiveness for storytelling

Energizing Content#

typescript
const energizingVoice: VoiceCharacteristics = {
  gender: 'male',
  age: 'young_adult',
  pitch: 'medium',
  pace: 'moderate',
  timbre: 'bright',
  energy: 'energetic',
  warmth: 0.7,
  clarity: 0.9,
  breathiness: 0.1,
  resonance: 0.8,
  expressiveness: 0.85
};

Key traits:

  • Clear, projecting voice
  • Moderate to fast pace
  • High energy and expressiveness
  • Low breathiness for power

Healing Content#

typescript
const healingVoice: VoiceCharacteristics = {
  gender: 'neutral',
  age: 'mature',
  pitch: 'medium_low',
  pace: 'slow',
  timbre: 'rich',
  energy: 'calm',
  warmth: 0.95,
  clarity: 0.75,
  breathiness: 0.5,
  resonance: 0.8,
  expressiveness: 0.6
};

Key traits:

  • Maximum warmth for safety
  • Rich timbre for depth
  • Moderate breathiness for intimacy
  • Calm but present energy

Tradition-Specific Profiles#

Zen Buddhist Voice#

typescript
const zenVoice: TraditionVoiceProfile = {
  tradition: 'zen_buddhism',
  name: 'Zen Master',
  typicalAccents: ['Japanese', 'neutral'],
  typicalQualities: ['sparse', 'direct', 'spacious'],
  typicalPace: 'slow',
  typicalTimbre: 'clear',
  typicalEnergy: 'calm',
  commonPhrases: [
    'Just sit',
    'What is this?',
    'Before thinking',
    'This moment, complete'
  ],
  sampleTexts: [
    'In zazen, we simply sit. Nothing to achieve. Nowhere to go.',
    'The bell sounds. Can you hear it? Where does it go?'
  ]
};

Tibetan Buddhist Voice#

typescript
const tibetanVoice: TraditionVoiceProfile = {
  tradition: 'tibetan_buddhism',
  name: 'Tibetan Lama',
  typicalAccents: ['Tibetan', 'Indian', 'neutral'],
  typicalQualities: ['warm', 'devotional', 'expansive'],
  typicalPace: 'measured',
  typicalTimbre: 'rich',
  typicalEnergy: 'engaged',
  commonPhrases: [
    'Precious human life',
    'May all beings be free',
    'The nature of mind',
    'Compassion for all sentient beings'
  ],
  sampleTexts: [
    'Visualize the Buddha before you, radiating infinite light and compassion.',
    'May the merit of this practice benefit all beings throughout space and time.'
  ]
};

Hindu/Yoga Voice#

typescript
const yogaVoice: TraditionVoiceProfile = {
  tradition: 'hinduism_yoga',
  name: 'Yoga Guide',
  typicalAccents: ['Indian', 'neutral'],
  typicalQualities: ['devotional', 'flowing', 'embodied'],
  typicalPace: 'slow',
  typicalTimbre: 'warm',
  typicalEnergy: 'engaged',
  commonPhrases: [
    'Breathe in prana',
    'Honor your body',
    'The divine within you',
    'Namaste'
  ],
  sampleTexts: [
    'Breathe in, drawing prana into your heart center. Breathe out, releasing tension.',
    'The light in me honors the light in you. Namaste.'
  ]
};

Secular Mindfulness Voice#

typescript
const secularVoice: TraditionVoiceProfile = {
  tradition: 'secular_mindfulness',
  name: 'Mindfulness Instructor',
  typicalAccents: ['neutral', 'American', 'British'],
  typicalQualities: ['accessible', 'grounded', 'scientific'],
  typicalPace: 'measured',
  typicalTimbre: 'clear',
  typicalEnergy: 'relaxed',
  commonPhrases: [
    'Notice without judgment',
    'When your mind wanders',
    'Bring your attention back',
    'This is natural'
  ],
  sampleTexts: [
    'Simply notice where your attention is right now. There is no right or wrong.',
    'When thoughts arise, acknowledge them and gently return to the breath.'
  ]
};

Prosody and Expression#

Speech Breaks#

typescript
// Break strengths for different contexts
const breakExamples = {
  none: '',                           // No pause
  'x-weak': '<break strength="x-weak"/>', // Minimal pause
  weak: '<break strength="weak"/>',   // Brief pause
  medium: '<break strength="medium"/>', // Normal pause
  strong: '<break strength="strong"/>', // Significant pause
  'x-strong': '<break strength="x-strong"/>' // Long pause
};

// Timed breaks
const timedBreaks = {
  sentence: '<break time="500ms"/>',  // Between sentences
  paragraph: '<break time="1s"/>',     // Between paragraphs
  breath: '<break time="3s"/>',        // For breathing space
  meditation: '<break time="5s"/>'     // Extended silence
};

Prosody Adjustments#

typescript
interface ProsodyParams {
  rate?: number | 'x-slow' | 'slow' | 'medium' | 'fast' | 'x-fast';
  pitch?: number | 'x-low' | 'low' | 'medium' | 'high' | 'x-high';
  volume?: number | 'silent' | 'x-soft' | 'soft' | 'medium' | 'loud' | 'x-loud';
  range?: number; // Pitch range percentage
}

// Example: Slowing down for emphasis
const slowEmphasis: ProsodyParams = {
  rate: 'slow',
  pitch: 'low',
  volume: 'soft'
};

// Example: Energizing passage
const energizing: ProsodyParams = {
  rate: 'medium',
  pitch: 'medium-high',
  volume: 'medium'
};

Emotional Expression#

typescript
// Map content type to emotional range
const emotionalRanges: Record<string, string[]> = {
  sleep: ['peaceful', 'drowsy', 'safe', 'cradled'],
  meditation: ['calm', 'present', 'spacious', 'grounded'],
  healing: ['compassionate', 'nurturing', 'accepting', 'loving'],
  energizing: ['vibrant', 'motivated', 'powerful', 'alive'],
  kids: ['playful', 'curious', 'adventurous', 'wonder-filled']
};

ElevenLabs Voice Settings#

Stability vs. Similarity#

typescript
interface VoiceSettings {
  stability: number;        // 0-1: Higher = more consistent
  similarity_boost: number; // 0-1: Higher = more similar to original
  style?: number;           // 0-1: Style transfer intensity
  use_speaker_boost?: boolean; // Enhance clarity
}

Stability Guidelines:

Content Type Stability Similarity Rationale
Sleep 0.7-0.8 0.7-0.8 Consistency for relaxation
Meditation 0.5-0.7 0.75-0.85 Balanced expression
Kids 0.6-0.8 0.7-0.8 Predictable for children
Storytelling 0.3-0.5 0.8-0.9 Expressive range
Healing 0.5-0.7 0.75-0.85 Warm, consistent
Energizing 0.4-0.6 0.7-0.8 Dynamic expression

Model Selection by Content#

Content Type Primary Model Alternative
Sleep eleven_v3 eleven_multilingual_v2
Meditation eleven_v3 eleven_multilingual_v2
Healing eleven_v3 -
Storytelling eleven_v3 eleven_multilingual_v2
Kids eleven_multilingual_v2 eleven_turbo_v2_5
Energizing eleven_v3 eleven_turbo_v2_5
Real-time eleven_flash_v2_5 eleven_turbo_v2_5

Voice Generation Workflow#

1. Define Requirements#

typescript
const requirements = {
  contentType: 'meditation',
  tradition: 'zen_buddhism',
  targetAudience: 'adults',
  sessionLength: 'long', // For consistency over time
  emotionalGoal: 'calm_presence',
  language: 'en'
};

2. Generate Voice Profile#

typescript
const voiceProfile: GeneratedVoiceProfile = {
  id: 'zen-guide-001',
  characteristics: {
    gender: 'neutral',
    age: 'mature',
    pitch: 'medium_low',
    pace: 'slow',
    timbre: 'clear',
    energy: 'calm',
    warmth: 0.8,
    clarity: 0.9,
    breathiness: 0.3,
    resonance: 0.75,
    expressiveness: 0.4
  },
  accent: {
    region: 'neutral',
    strength: 0.0,
    characteristics: ['clear', 'unaccented']
  },
  qualities: {
    primary: ['spacious', 'direct', 'grounded'],
    secondary: ['wise', 'patient']
  },
  personality: {
    traits: ['mindful', 'present', 'non-judgmental'],
    emotionalRange: ['calm', 'serene', 'peaceful'],
    speakingStyle: ['sparse', 'measured', 'deliberate']
  },
  description: 'A calm, grounded voice embodying Zen simplicity and presence.',
  promptForTTS: 'A mature, gender-neutral voice with clear timbre and calm energy. Speaks slowly and deliberately with natural pauses. Warm but not overly emotional.',
  sampleText: 'Simply sit. Notice the breath. This moment, just as it is.',
  metadata: {
    tradition: 'zen_buddhism',
    generatedAt: Date.now(),
    seed: 12345,
    version: '1.0'
  }
};

3. Configure TTS Settings#

typescript
const ttsConfig = {
  model_id: 'eleven_v3',
  voice_id: 'selected-voice-id',
  voice_settings: {
    stability: 0.6,
    similarity_boost: 0.8,
    style: 0.4,
    use_speaker_boost: true
  },
  output_format: 'mp3_44100_192'
};

4. Apply Content-Specific Tags#

typescript
const taggedContent = `
[calm] [deliberate] Welcome to this moment of stillness.
[pauses]
[warmly] Allow yourself to simply be here.
[thoughtful] Notice your breath, naturally flowing.
[deliberate] In... [pause] and out...
`;

Voice Quality Validation#

Consistency Check#

typescript
interface VoiceQualityMetrics {
  pitchConsistency: number;      // 0-1
  paceConsistency: number;       // 0-1
  emotionalAlignment: number;    // 0-1
  clarityScore: number;          // 0-1
  naturalness: number;           // 0-1
}

function validateVoiceQuality(audio: Buffer): VoiceQualityMetrics {
  // Analyze audio for quality metrics
  return {
    pitchConsistency: 0.92,
    paceConsistency: 0.88,
    emotionalAlignment: 0.85,
    clarityScore: 0.95,
    naturalness: 0.90
  };
}

Content Alignment Check#

typescript
function checkContentAlignment(
  voiceProfile: GeneratedVoiceProfile,
  contentType: string
): boolean {
  const contentEnergy = getExpectedEnergy(contentType);
  const voiceEnergy = voiceProfile.characteristics.energy;

  // Ensure voice energy matches content needs
  const energyMap = {
    sleep: ['calm'],
    meditation: ['calm', 'relaxed'],
    healing: ['calm', 'relaxed', 'engaged'],
    energizing: ['engaged', 'energetic', 'passionate']
  };

  return energyMap[contentType]?.includes(voiceEnergy) ?? false;
}

Best Practices#

1. Match Voice to Content#

typescript
// Good: Calm voice for sleep content
const sleepVoice = {
  energy: 'calm',
  pace: 'very_slow',
  timbre: 'warm'
};

// Avoid: Energetic voice for sleep
const wrongSleepVoice = {
  energy: 'energetic',  // Wrong for sleep
  pace: 'fast',         // Wrong for sleep
  timbre: 'bright'      // Too stimulating
};

2. Consider Session Length#

typescript
// Long sessions: Higher stability for consistency
const longSession = {
  stability: 0.7,
  similarity_boost: 0.8
};

// Short sessions: Can use more expression
const shortSession = {
  stability: 0.5,
  similarity_boost: 0.85
};

3. Test with Sample Text#

typescript
// Always test with representative content
const testTexts = {
  meditation: 'Take a deep breath. Notice the sensation of air entering your body.',
  sleep: 'Allow your body to sink deeper into rest. Let go of the day.',
  energizing: 'Feel the energy rising within you. You are alive and powerful.',
  healing: 'You are safe here. Allow healing light to flow through you.'
};

4. Iterate and Refine#

typescript
async function refineVoice(
  initialProfile: GeneratedVoiceProfile,
  feedback: UserFeedback
): Promise<GeneratedVoiceProfile> {
  const adjustments = interpretFeedback(feedback);

  return {
    ...initialProfile,
    characteristics: {
      ...initialProfile.characteristics,
      warmth: clamp(initialProfile.characteristics.warmth + adjustments.warmth, 0, 1),
      pace: adjustPace(initialProfile.characteristics.pace, adjustments.pace),
      expressiveness: clamp(initialProfile.characteristics.expressiveness + adjustments.expressiveness, 0, 1)
    }
  };
}

5. Maintain Voice Library#

typescript
interface VoiceLibrary {
  voices: Map<string, GeneratedVoiceProfile>;
  categories: Map<string, string[]>;  // category -> voice IDs
  favorites: string[];

  getVoicesByTradition(tradition: SpiritualTradition): GeneratedVoiceProfile[];
  getVoicesByContent(contentType: string): GeneratedVoiceProfile[];
  getRecommendedVoice(requirements: VoiceRequirements): GeneratedVoiceProfile;
}

Common Pitfalls#

1. Too Much Expression in Calming Content#

typescript
// Wrong: High expressiveness for sleep
{ expressiveness: 0.9 }  // Too stimulating

// Better: Low expressiveness for sleep
{ expressiveness: 0.3 }  // Monotonous, sleep-inducing

2. Mismatched Pace and Energy#

typescript
// Wrong: Fast pace with calm energy
{ pace: 'fast', energy: 'calm' }  // Confusing

// Better: Consistent pace and energy
{ pace: 'slow', energy: 'calm' }  // Aligned

3. Ignoring Cultural Context#

typescript
// Consider tradition when selecting accent
const traditionAccents = {
  zen_buddhism: ['neutral', 'Japanese'],
  hinduism_yoga: ['neutral', 'Indian'],
  sufism: ['neutral', 'Middle Eastern'],
  secular_mindfulness: ['neutral']
};

4. Over-Relying on Default Settings#

typescript
// Always customize for content type
const customSettings = getVoiceSettingsForContent(contentType);
// Don't use default settings blindly