Skip to main content

​
Overview

Voice settings control how your agent sounds during conversations. The right voice selection can significantly impact caller engagement, trust, and overall conversation success. Configure voice identity, language preferences, and speech characteristics to match your brand and audience.
Voice Settings Include:
  • Voice selection from extensive library
  • Language and dialect configuration
  • Voice characteristics (gender, style, accent)
  • Speech synthesis parameters
  • Multi-language support

​
Voice Selection

​
Accessing the Voice Selector

  1. Navigate to your Single Prompt Agent
  2. Expand the Configuration panel
  3. Click on Voice Settings section
  4. Click the voice preview button or Change to open voice selector

​
Voice Library

The voice library contains 100+ professional AI voices across multiple providers: Voice Providers:
  • Deepgram - Natural conversational voices
  • ElevenLabs - High-quality expressive voices
  • PlayHT - Diverse accent and dialect options
  • Azure - Microsoft’s neural voices
  • Custom - Your cloned voices (if any)
Each provider has unique characteristics. Test voices from multiple providers to find the perfect match for your use case.

​
Browsing Voices

The voice selector organizes voices into tabs: Four Main Tabs:
  1. All Voices - Complete library of available voices
  2. Favorite Voices - Voices you’ve marked as favorites (star icon)
  3. Currently Used - Voices recently used in your agents
  4. My Voices - Custom voices you’ve created via voice cloning

​
Voice Card Information

Each voice displays:
  • Profile Image - Visual representation of the voice
  • Name - Voice identifier (e.g., “Sarah - Professional”, “Ahmad - Friendly”)
  • Language - Primary language (English, Arabic, etc.)
  • Dialect/Accent - Regional variant (US English, Egyptian Arabic, etc.)
  • Gender - Male, Female, or Neutral
  • Style - Conversational or Narrator
  • Play Button - Preview the voice with sample audio
  • Favorite Icon - Star to add/remove from favorites

​
Previewing Voices

To preview a voice:
  1. Find a voice card in the selector
  2. Click the Play button (▶)
  3. Listen to the sample audio
  4. Click again to stop playback
Preview samples are short demos. Always test voices in actual conversation scenarios before finalizing your selection.
Best practices for previewing:
  • Listen to multiple voices in the same category
  • Preview with headphones for better audio quality
  • Compare male vs female voices for your use case
  • Test different accents for international audiences

​
Selecting a Voice

To select a voice:
  1. Browse or filter to find your preferred voice
  2. Click the voice card
  3. Voice is automatically selected
  4. Voice preview appears in Voice Settings
  5. Click Save to apply changes
The selected voice will be used for all conversations with this agent.

​
Filtering Voices

Use filters to narrow down the voice library and find the perfect match:

​
Language Filter

Select the primary conversation language: Available Languages:
  • English - All English voices (US, UK, Australian, etc.)
  • Arabic - All Arabic voices (various dialects)
  • All - Show voices from all languages
The language filter should match your agent’s conversation language. Using an English voice for Arabic conversations (or vice versa) will result in poor pronunciation.

​
Gender Filter

Filter by voice gender:
  • Male - Masculine voices
  • Female - Feminine voices
  • All - Show all genders
Gender Selection Considerations:
  • Match your brand personality and target audience
  • Consider cultural preferences in your market
  • Test both to see which generates better engagement
  • No right answer - depends on use case and preference

​
Dialect Filter (Language-Specific)

For Arabic, select specific regional dialects: Arabic Dialects Available:
  • Egyptian (EGY) - Broad Middle East appeal, widely understood
  • Saudi Arabian (KSA) - Gulf region, formal contexts
  • Emirati (UAE) - Gulf region, business contexts
  • Jordanian (JOR) - Levantine region
  • Lebanese (LEB) - Levantine region, cosmopolitan
  • Syrian (SYR) - Levantine region
  • Palestinian (PLS) - Levantine region
  • Iraqi (IRQ) - Iraqi dialect
  • Bahraini (BAH) - Gulf region
  • All Dialects - Show all Arabic voices
Dialect Matching Tips:
  • Match dialect to your target market for best results
  • Egyptian Arabic has widest recognition across MENA
  • Gulf dialects (KSA, UAE) for GCC markets
  • Levantine dialects for Syria, Jordan, Lebanon, Palestine
English Dialects: English voices typically indicate accent in the voice name:
  • US English
  • UK English
  • Australian English
  • Canadian English
  • Indian English

​
Style Filter

Filter by voice style/type: Conversational
  • Natural, friendly tone
  • Informal delivery
  • Best for: Customer service, sales, support calls
  • Sounds like a helpful person having a conversation
Narrator
  • Clear, articulate delivery
  • Professional announcer style
  • Best for: Announcements, instructions, information delivery
  • Sounds like a professional speaker or broadcaster
Search voices by name: 
Type: "Sarah", "professional", "friendly", "Ahmad"
The search filters in real-time as you type.

​
Clearing Filters

Click Clear Filters to reset all filters and show the complete library. Filter Results Count: The interface displays “Filter Result X Voice” showing how many voices match your current filters.

​
Voice Characteristics

​
Understanding Voice Properties

Each voice has distinct characteristics that affect how callers perceive your agent: Pitch
  • High pitch: Youthful, energetic
  • Medium pitch: Neutral, professional
  • Low pitch: Authoritative, calming
Pace
  • Fast: Energetic, efficient
  • Medium: Conversational, balanced
  • Slow: Deliberate, easy to understand
Tone
  • Warm: Friendly, approachable
  • Neutral: Professional, business-like
  • Authoritative: Confident, commanding
Emotion
  • Expressive: Shows emotion in delivery
  • Neutral: Steady, consistent tone
  • Subdued: Calm, understated
Match voice characteristics to your use case:
  • Customer Support: Warm, medium pace, friendly
  • Sales: Confident, slightly faster, enthusiastic
  • Healthcare: Calm, slower, reassuring
  • Legal/Finance: Professional, measured, authoritative

​
Language Configuration

​
Primary Language Setting

The language setting determines:
  • Speech recognition model
  • Response generation language
  • Voice compatibility
  • System message language
To configure language:
  1. Scroll to Voice Settings configuration
  2. Language is auto-set when you select a voice
  3. English voice → English language
  4. Arabic voice → Arabic language
Changing language after setup requires changing the voice as well. Voices are language-specific.

​
Language Dialect Switcher

Enable automatic dialect adaptation within the same language. What it does: Detects the caller’s specific dialect and adapts the agent’s speech patterns and understanding to match. Example Use Cases: Arabic:
  • Agent configured with Egyptian Arabic
  • Caller speaks Saudi dialect
  • Agent adapts recognition to better understand Saudi pronunciation
  • Maintains Egyptian voice but improves comprehension
English:
  • Agent configured with US English
  • Caller speaks with strong Scottish accent
  • Agent adapts to better understand regional variations
  • Improves accuracy without changing agent’s voice
To enable:
  1. Open Call Settings section
  2. Enable Language/Dialect Switcher toggle
  3. Save changes
Benefits:
  • Better understanding of regional accents
  • Improved speech recognition accuracy
  • More natural conversations with diverse callers
Trade-offs:
  • Slight increase in processing time (~50-100ms)
  • Works best within same language family

​
Voice Management

​
Favoriting Voices

Build a curated collection of preferred voices: To add to favorites:
  1. Find a voice you like
  2. Click the Star icon on the voice card
  3. Voice appears in “Favorite Voices” tab
To remove from favorites:
  1. Click the filled Star icon
  2. Voice is removed from favorites
Benefits of favoriting:
  • Quick access to your top choices
  • Easy comparison between preferred voices
  • No need to search repeatedly
  • Consistent voice selection across agents
Create a favorites collection of 3-5 voices for different use cases: one for customer service, one for sales, one for announcements, etc.

​
Recently Used Voices

The Currently Used tab shows voices you’ve recently used across all agents in your project. Features:
  • Automatically updated when you select a voice
  • Sorted by most recent usage
  • Includes voices from all agents
  • Quick access for brand consistency
Use cases:
  • Quickly reuse the same voice across multiple agents
  • Maintain brand consistency
  • Find the voice from your successful agent

​
Copying Voice Information

Copy Voice ID:
  1. Hover over voice card
  2. Click actions menu (three dots)
  3. Select Copy ID
  4. Voice ID copied to clipboard
Copy Voice Name:
  1. Hover over voice card
  2. Click actions menu
  3. Select Copy Name
  4. Voice name copied to clipboard
Voice IDs are useful for API integrations and programmatic agent configuration.

​
Voice Selection Best Practices

​
Matching Voice to Brand

Luxury/Premium Brands
  • Sophisticated, refined voices
  • Professional narrator style
  • Slower, deliberate pace
  • Clear pronunciation
Tech/Startup Brands
  • Modern, approachable voices
  • Conversational style
  • Efficient, natural pace
  • Friendly but professional
Healthcare/Medical
  • Warm, reassuring voices
  • Clear, easy to understand
  • Slower pace for comprehension
  • Professional but caring
Retail/Hospitality
  • Friendly, energetic voices
  • Conversational style
  • Upbeat tone
  • Welcoming personality
Financial/Legal
  • Authoritative, trustworthy voices
  • Professional delivery
  • Clear, measured pace
  • Formal tone

​
Matching Voice to Use Case

Customer Support
  • Gender: Test both, no clear preference
  • Style: Conversational
  • Tone: Friendly, patient, helpful
  • Pace: Medium
  • Example: “Sarah - Friendly” or “James - Support”
Sales Calls
  • Gender: Test both for conversion rates
  • Style: Conversational
  • Tone: Confident, enthusiastic
  • Pace: Slightly faster, energetic
  • Example: “Alex - Energetic” or “Maya - Sales”
Technical Support
  • Gender: Clear pronunciation over gender
  • Style: Conversational or Narrator
  • Tone: Professional, reassuring
  • Pace: Medium, clear enunciation
  • Example: “David - Professional” or “Amira - Clear”
Appointment Reminders
  • Gender: Neutral preference
  • Style: Conversational
  • Tone: Friendly but professional
  • Pace: Medium
  • Example: “Emily - Reminder” or “Omar - Scheduler”
Surveys/Feedback
  • Gender: Warm, trustworthy voices
  • Style: Conversational
  • Tone: Neutral, encouraging
  • Pace: Slower, allows time to think
  • Example: “Rachel - Neutral” or “Khalid - Survey”
Emergency/Urgent
  • Gender: Authoritative voices
  • Style: Narrator
  • Tone: Clear, commanding
  • Pace: Deliberate, not rushed
  • Example: “Marcus - Authoritative” or “Fatima - Alert”

​
Cultural Considerations

Arabic-Speaking Markets:
  • Egyptian Arabic: Widely understood across Middle East, good for broad appeal
  • Gulf Dialects (KSA, UAE): Essential for GCC markets, business contexts
  • Levantine: Syria, Jordan, Lebanon, Palestine - cosmopolitan appeal
  • Gender preferences vary by culture and context - test both
English-Speaking Markets:
  • US English: North American audiences
  • UK English: European, Middle East business contexts
  • Australian: Australian, New Zealand markets
  • Neutral/International: Global SaaS, tech companies
Cultural norms vary significantly. What works in one market may not work in another. Always test with your target audience.

​
A/B Testing Voices

Test multiple voices to find what works best: Testing Framework:
  1. Select 2-3 candidate voices from your favorites
  2. Create test agents with identical prompts, different voices
  3. Run 20-50 test calls per voice
  4. Measure metrics:
    • Call completion rate
    • Average call duration
    • Customer satisfaction (if surveyed)
    • Task completion success rate
    • Caller feedback
  5. Analyze results and select winner
Small changes in voice can have significant impact on engagement. Don’t skip A/B testing for high-volume agents.

​
Voice Customization (Advanced)

​
Custom Voice Cloning

Create unique branded voices using voice cloning:
  1. Record or provide sample audio (3-30 minutes)
  2. Upload via Voice Cloning feature
  3. Process takes 10-30 minutes
  4. Custom voice appears in “My Voices” tab
  5. Select for your agent like any other voice
Benefits:
  • Unique brand identity
  • Consistent voice across all channels
  • Use executive or founder’s voice
  • Recreate retired employee voices
Best for:
  • Brand differentiation
  • Celebrity endorsements
  • Multilingual cloning (one voice, many languages)
  • Consistency with existing IVR
→ Learn More: Voice Cloning

​
Troubleshooting Voice Issues

​
Voice Sounds Robotic or Unnatural

Possible Causes:
  • Low-quality voice selected
  • Voice-text mismatch (Arabic voice speaking English)
  • Technical issues with provider
Solutions:
  1. Try a different voice from same provider
  2. Test voices from different providers
  3. Verify language matches voice language
  4. Check for provider outages in Live Call Logs

​
Voice is Too Fast or Too Slow

Solution: Voice speed is inherent to each voice selection. To adjust:
  1. Try a different voice with your preferred pace
  2. Adjust your prompt to use shorter/longer sentences
  3. For custom voices, re-clone with different sample audio pace
Voice speed/stability parameters are currently tied to voice selection. Future updates may include speed adjustment controls.

​
Voice Doesn’t Match Accent/Dialect Needed

Solution:
  1. Use Dialect Filter to find exact match
  2. If no exact match, choose closest available
  3. Consider custom voice cloning for specific accents
  4. Enable Language Dialect Switcher for better caller understanding

​
Voice Sounds Different in Production vs Preview

Possible Causes:
  • Preview audio is compressed
  • Actual synthesis uses different parameters
  • Network/audio quality differences
Solution:
  1. Always test via Test Agent feature (browser call or phone)
  2. Don’t rely solely on preview audio
  3. Test in production-like environment

​
Can’t Find Previously Used Voice

Solutions:
  1. Check Currently Used tab
  2. Use search with voice name
  3. Check favorites if you favorited it
  4. Filter by language/gender to narrow down

​
Voice Settings Configuration Panel

​
Accessing Voice Settings

In your Single Prompt Agent:
  1. Expand Configuration panel (right side)
  2. Scroll to Voice Settings section
  3. Click to expand

​
Voice Settings Display

The Voice Settings panel shows: Selected Voice Card:
  • Profile image (if available)
  • Voice name
  • Language/dialect flag icon
  • Change button to open voice selector
To change voice:
  1. Click Change button
  2. Voice selector modal opens
  3. Browse, filter, search for new voice
  4. Click voice card to select
  5. Modal closes automatically
  6. New voice appears in settings
  7. Click Save to apply
Voice changes take effect immediately for new calls. Active in-progress calls continue with the previous voice.

​
LLM Temperature & Model Settings

While not directly voice settings, LLM configuration affects speech patterns:

​
Temperature

Temperature controls response creativity and variability: Range: 0.0 - 1.0 Default: 0.2 Low Temperature (0.0 - 0.3):
  • More consistent, predictable responses
  • Follows script/guidelines closely
  • Less creative variation
  • Best for: Customer support, information delivery, compliance
Medium Temperature (0.4 - 0.7):
  • Balanced creativity and consistency
  • Natural variation in phrasing
  • Maintains guidelines while sounding natural
  • Best for: Sales, conversational agents
High Temperature (0.8 - 1.0):
  • More creative, varied responses
  • May deviate from script
  • More human-like unpredictability
  • Best for: Creative conversations, less structured interactions
GPT-5 family models require temperature = 1.0 (fixed). This is automatically enforced.
To adjust temperature:
  1. Open LLM Configuration section
  2. Adjust Temperature slider
  3. Save changes

​
LLM Model Selection

The LLM model affects response quality, speed, and voice-text alignment: Available Models:
  • GPT-5 Series - Most advanced, temperature fixed at 1.0
  • GPT-4.1 Series - Balanced quality and speed
  • Gemini Series - Google’s models
  • DeepMyst Optimized - Telephony-optimized models
DeepMyst optimized models are specifically tuned for voice conversations and may provide better results than standard models.
→ Learn More: LLM Configuration

​
Voice + Prompt Synergy

Your voice selection and prompt should work together:

​
Matching Voice to Prompt Personality

If your prompt says “friendly and casual”:
  • ✓ Select conversational, warm voice
  • ✗ Don’t select formal narrator voice
If your prompt says “professional and authoritative”:
  • ✓ Select professional narrator or business voice
  • ✗ Don’t select casual, youthful voice
If your prompt says “empathetic and caring”:
  • ✓ Select warm, slower-paced voice
  • ✗ Don’t select fast, transactional voice

​
Voice Affects Perceived Personality

Even with identical prompts, different voices create different perceptions: Professional Female Voice:
  • Perceived as helpful, organized
  • Good for: Healthcare, education, customer service
Professional Male Voice:
  • Perceived as authoritative, knowledgeable
  • Good for: Finance, legal, technical support
Casual Female Voice:
  • Perceived as friendly, approachable
  • Good for: Retail, hospitality, lifestyle
Casual Male Voice:
  • Perceived as confident, casual
  • Good for: Sales, tech support, informal services
These are generalizations based on common perceptions, not universal truths. Always test with your specific audience.

​
Multi-Language Scenarios

​
Single Language Agents

Most agents operate in a single language:
  1. Select language: English or Arabic
  2. Choose voice in that language
  3. Write prompt in that language
  4. System messages appear in that language
Simple and straightforward.

​
Bilingual Callers

For callers who might speak either language: Option 1: Separate Agents
  • Create one English agent
  • Create one Arabic agent
  • Route calls based on caller’s language preference
Option 2: Language Dialect Switcher
  • Configure agent in primary language
  • Enable Language Dialect Switcher
  • Agent adapts to caller’s dialect within same language family
Single Prompt Agents cannot switch between entirely different languages (English ↔ Arabic) mid-call. For true multilingual support, create separate agents or use Flow Agent with language detection.

​
Voice Selection Checklist

Before finalizing your voice selection, verify:
  • Language matches - Voice language = agent language
  • Dialect is appropriate - Matches target market if applicable
  • Gender fits brand - Tested with target audience
  • Style matches use case - Conversational vs Narrator
  • Tone aligns with prompt - Voice personality matches written personality
  • Tested in real calls - Used Test Agent feature, not just preview
  • Clear audio quality - No distortion, artifacts, or robotic sound
  • Pace is appropriate - Not too fast or slow for use case
  • Team approval - Stakeholders reviewed and approved
  • A/B tested - If high-volume, compared 2-3 options
Use this checklist before launching your agent to production. Voice is often the first impression callers have of your brand.

​
Next Steps

Write Effective Prompts

Craft compelling agent instructions and personality

Call Behavior Settings

Configure response timing and interaction controls

Intelligence Features

Enable gender detection, smart call end, and more

Voice Cloning

Create custom branded voices
Need more control? Explore Flow Agent Voice Configuration for node-level voice customization.