Frequently asked questions
What is the best male voice for a corporate IVR?
Atlas is designed as a neutral, news-style voice suitable for corporate IVR systems. It provides clear, consistent synthetic speech without distinct personality traits, ensuring professional communication. Sage is also an option for instructional flows. The choice depends on whether you prefer a neutral broadcast tone or a calm instructional style for your automated calls.
How does the Atlas voice differ from Milo?
Atlas offers a neutral, news-style tone ideal for standard IVR prompts. Milo provides a friendly, conversational tone better suited for interactive customer service interactions. Both are male presets with identical technical specifications, pricing, and language support. Select Atlas for formal announcements and Milo for more engaging, two-way conversational flows.
Can I use Felix for promotional calls?
Yes, Felix is described as energetic and promo-oriented, making it suitable for promotional calls. It supports all 23 languages and works with the standard TTS-1 model. Ensure the text input stays within the 4096-character limit per request. The energetic tone aligns with marketing objectives while maintaining the consistent synthetic quality of the API.
Is Orion too deep for telephone systems?
Orion is described as deep and documentary-style. Telephone systems handle deep voices effectively as long as the audio format (mp3, wav, opus, flac, pcm) is compatible. The API returns standard audio files. There is no specific frequency limitation mentioned for IVR systems. Test Orion with your specific telephony provider to ensure clarity.
Do I need to attribute the voice in my IVR?
Each generated file carries an inaudible watermark marking it as synthetic speech. The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain readings. While the watermark is inaudible, attribution may be required depending on your use of the underlying CC BY 4.0 licensed data sources.
How do I test different voices with my users?
Use the GET /v1/voices endpoint to list available voices and their properties. Generate sample audio files using the POST /v1/audio/speech endpoint with different voice presets. Distribute these audio files to your users for feedback. Adjust the speed parameter (0.25–4.0) to optimize perception for your specific IVR audience.
Can I switch voices mid-conversation?
Yes, you can switch voices by sending a new POST request with a different voice parameter. Each request generates a separate audio file. You must split your text into parts of up to 4096 characters and join the resulting audio files. This allows dynamic voice changes within a single conversation flow.
Are there voice-specific character limits?
No, the character limit is consistent across all voices. Each request supports up to 4096 characters of input text. This limit applies regardless of whether you use a male or female voice. For longer texts, split the content into multiple requests and join the generated audio files programmatically.
Does the speed setting affect voice perception?
Yes, the speed parameter (0.25–4.0) directly affects voice perception and generation time. Faster speeds reduce latency but may impact clarity. Slower speeds improve comprehension but increase processing time. Typical generation is roughly real-time, but requests are processed sequentially, so higher speeds may still queue under load.
What is the difference between warm and energetic voices?
Aria is described as warm and clear, suitable for narration. Felix is energetic and promo-oriented. These describe the tonal quality of the synthetic speech. Technically, both use the same engine and pricing. The difference is purely perceptual. Choose based on whether you want a calming or lively tone for your IVR.
Can I mix male and female voices in one call?
Yes, you can mix voices by sending separate requests for each segment. Generate audio for each voice using the POST endpoint, then join the files. Each request is independent. Ensure the total text length for each segment does not exceed 4096 characters. This allows flexible voice assignment within a single call flow.
How do I handle voice changes in my IVR flow?
Handle voice changes by sending distinct POST requests for each voice segment. Use the GET /v1/voices endpoint to verify available voices. Generate audio for each part, then concatenate the files. Manage the speed and language parameters for each segment to ensure consistent quality across the entire IVR interaction.
Is there a limit to how many times I can use a voice?
No, there is no limit to how many times you can use a specific voice. You can reuse any of the 8 presets indefinitely. Pricing is based on characters processed, not voice usage. Each request is charged at $0.01 per 1,000 characters, regardless of which voice preset is selected.
Do all voices support all 23 languages equally?
Yes, any voice can speak any of the 23 supported languages. The language is detected automatically or set via the optional language parameter. If not set, Latin text defaults to English. All voices have equal capability across Arabic, Chinese, English, German, Japanese, Spanish, and other listed languages without performance differences.
Understanding Voice Roles in IVR
In IVR systems, voice roles dictate user navigation efficiency. Navigation prompts require high clarity for quick action, while informational content benefits from a tone that reduces cognitive load. Distinguishing these roles ensures callers process instructions accurately without confusion.
Listener retention depends on matching voice tone to content complexity. A mismatched tone increases call abandonment rates. Consistent voice usage builds trust, allowing users to anticipate the system's behavior and navigate menus with greater confidence and reduced effort.
Navigation vs Information
Navigation prompts demand precise articulation to guide users through menu options. Informational responses, such as balance inquiries, benefit from a calmer delivery that aids comprehension.
Differentiating these roles prevents listener fatigue. Clear structural cues help users distinguish between actions they must take and information they simply receive.
- Use neutral voices for complex menu structures.
- Apply conversational tones for simple status updates.
Tone and Listener Retention
Consistent tone reduces cognitive strain during long calls. Listeners retain information better when the voice matches the context, such as using a calm voice for detailed explanations.
Inconsistent tone shifts can cause confusion, leading to repeated requests and increased operational costs. Maintaining a steady auditory experience supports higher completion rates.
- Match tone to content density.
- Avoid abrupt style shifts within a single flow.
Brand Consistency
Voice acts as a brand identifier. Using a consistent voice across all IVR touchpoints reinforces brand recognition and professionalism.
Uniform voice selection ensures that users experience a cohesive service, regardless of the specific menu or information type they are accessing.
- Standardize voice selection across regions.
- Align voice characteristics with corporate identity.
Male Voice Presets Overview
The three primary male presets in the Viva Voice Labs API serve distinct functional roles within IVR architectures. Each voice is a preset with a specific tonal quality, ensuring that synthetic speech aligns with the informational density and emotional tone of the caller's journey.
Atlas: Neutral News Style
Atlas provides a neutral, professional tone ideal for delivering factual updates, transaction confirmations, or hold messages. Its consistent pacing reduces cognitive load for callers processing standard account information.
Milo: Friendly Conversational
Milo offers a warm, approachable delivery suitable for customer support flows or greeting sequences. This preset minimizes the robotic feel of automated systems, encouraging clearer communication during troubleshooting steps.
Felix: Energetic Promotional
Felix delivers high-energy audio optimized for promotional announcements or upsell segments. Its dynamic range captures attention effectively, making it appropriate for time-sensitive offers or priority service notifications.
Orion: Deep Documentary Tone
Orion delivers a deep, documentary-grade tone suitable for authoritative IVR announcements. Its vocal profile conveys stability and seriousness, making it ideal for high-stakes information delivery where listener trust is critical. This preset is engineered to reduce cognitive load during complex instruction sets.
Use Cases for Deep Voices
Deploy Orion for system alerts, account verification results, or critical outage notifications. The lower frequency range may be distinct, but no specific data confirms it cuts through background noise more effectively than higher-pitched voices.
Perception of Authority
Listeners often associate deeper tones with expertise and reliability. Using Orion signals that the information provided is official and definitive, reducing the need for users to double-check details or call back for confirmation.
Suitability for Long Prompts
Orion maintains consistent pacing and tone over extended durations, preventing listener fatigue during long hold messages or detailed account summaries. The vocal quality remains stable even at higher speeds, ensuring intelligibility without sounding rushed.
Comparing Voice Characteristics
Selecting the correct male voice for your IVR requires balancing auditory clarity with appropriate energy levels. The choice between Atlas, Milo, and Felix directly influences how callers perceive the system's reliability and responsiveness.
Warmth and Clarity
Atlas provides a neutral, news-style delivery ideal for factual prompts, ensuring high comprehension rates in noisy environments. Milo offers a friendly, conversational tone that reduces caller anxiety during account verification steps.
- Atlas: Neutral, news-style delivery
- Milo: Friendly, conversational tone
Energy and Pace
Felix delivers an energetic, promotional style suitable for short, action-oriented prompts like menu options or promotional announcements. Adjust the speed parameter (0.25–4.0) to match the caller's expected reading pace without sacrificing clarity.
- Felix: Energetic, promo style
- Speed range: 0.25–4.0
Professionalism vs Approachability
Choose Atlas for maximum professionalism in legal or financial IVRs, where neutral tone reduces ambiguity. Opt for Milo when approachability is prioritized, such as in customer service queues, to foster a more engaging interaction.
- Atlas: Maximum professionalism
- Milo: Prioritized approachability
Technical Considerations for Voice Selection
To select the optimal voice for your IVR, you must evaluate technical constraints alongside tonal qualities. Audio format dictates bandwidth usage and playback compatibility, while speed adjustments directly influence character consumption and call duration.
Audio Format Impact
Choose between mp3 for efficient bandwidth or wav for uncompressed quality. Selecting opus or flac depends on your infrastructure’s decoding capabilities and storage requirements for archived calls.
Speed Adjustment Effects
Adjusting speed from 0.25 to 4.0 changes generation time and character usage. Faster speeds reduce latency but may impact comprehension; slower speeds increase clarity but extend call duration.
Background Noise Compatibility
Ensure selected voices remain intelligible; however, the API does not provide specific noise cancellation or background noise optimization features. Test voice clarity in noisy environments to prevent listener fatigue and reduce repeat calls.
Licensing Implications
Viva Voice Labs utilizes the LibriTTS-R dataset, licensed under CC BY 4.0, and sourced from LibriVox public domain audiobook readings. This foundation ensures that your IVR audio files are generated from legally cleared, licensable synthetic voices suitable for commercial call center operations.
The use of licensed recordings provides a clear chain of title for generated speech. This reduces legal ambiguity when deploying high-volume IVR systems, as the underlying voice models are derived from public domain sources with established open licenses.
LibriTTS-R License Terms
The LibriTTS-R dataset operates under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. This permits commercial use of the generated audio, provided appropriate credit is given to the original dataset creators if required by your internal compliance standards.
- CC BY 4.0 license allows commercial use.
- Derived from LibriVox public domain readings.
- Clear attribution requirements for the source dataset.
Public Domain Origins
Voice data originates from LibriVox, a platform hosting public domain audiobook readings. These recordings are free of restrictive copyright claims on the underlying spoken content, allowing us to train and serve consistent synthetic voices without ongoing royalty obligations per call.
- Sourced from LibriVox public domain archives.
- No per-call royalty fees.
- Stable licensing structure for high-volume usage.
Commercial Usage Rights
Generated audio files are cleared for commercial deployment in IVR, phone trees, and automated notifications. The synthetic nature of the output, combined with the open license of the source material, supports broad commercial distribution without additional licensing layers for standard call center applications.
- Cleared for commercial IVR deployment.
- No per-call royalty fees.
- Supports broad commercial distribution.
Testing and Validation
Validating synthetic voices ensures they meet IVR standards before full deployment. A/B testing allows developers to compare listener retention and comprehension across different male presets.
A/B Testing Voices
Run parallel tests with Atlas and Milo to measure completion rates. Use the API’s speed parameter to simulate different call center paces.
- Test 4096-character limits with long prompts
- Compare mp3 vs wav clarity
- Monitor drop-off rates
User Feedback Loops
Collect explicit ratings after voice interactions. Use simple yes/no questions to gauge comprehension without increasing call time significantly.
- Post-call surveys
- Error rate tracking
Performance Metrics
Track generation speed; typical speed is roughly real time (about 25 characters per second), but requests are processed sequentially and may queue under load. Ensure the 60 requests per minute limit does not cause queue delays during peak hours.
- Latency measurements
- Queue wait times
- Success rate per voice

Implementation Best Practices
Standardize voice assignments across all IVR flows to prevent listener confusion. Each voice serves a specific functional role, such as Atlas for news-style updates or Milo for conversational prompts. Consistent usage reinforces brand identity and reduces cognitive load for callers navigating complex menus.
Consistent Voice Usage
Assign a single male voice to each distinct menu node. Avoid switching voices within the same logical flow, as this disrupts the user’s mental model. Use Atlas for neutral informational segments and Milo for interactive dialogues. This consistency ensures predictable audio cues, aiding callers in anticipating the type of information being presented.
- Map each IVR menu to a specific voice preset.
- Maintain the same voice for all prompts within a single flow.
- Document voice assignments in your API configuration for easy reference.
Mixing Male and Female Voices
Strategically combine male and female voices to differentiate between system announcements and user interactions. For instance, use Atlas for automated system alerts and switch to a female voice like Aria for human-like customer support prompts. This contrast helps callers distinguish between machine-generated instructions and interactive elements, improving overall usability without increasing complexity.
Ensure smooth transitions between voices by maintaining consistent speed and tone settings. Use the speed parameter to align the pacing of different voices, preventing jarring audio shifts. This approach supports a natural dialogue flow, enhancing the professional appearance of your IVR system.
- Use male voices for system announcements and female voices for interactive prompts.
- Keep speed settings consistent across mixed voices to avoid abrupt transitions.
- Limit voice switches to distinct logical sections of the IVR tree.
Contextual Voice Matching
Match the voice’s emotional tone to the context of the message. Use Felix for energetic promotional offers within the IVR and Orion for authoritative, documentary-style information. Aligning voice characteristics with content type enhances message retention and ensures the tone matches the caller’s expectations, whether they are receiving urgent updates or detailed instructions.
Common Pitfalls
Selecting the correct male voice preset requires aligning vocal characteristics with specific IVR functions. Common errors include assigning energetic tones to complex instructions or ignoring how listener fatigue impacts call resolution rates.
Overusing Energetic Voices
Applying Felix’s energetic tone to all menu prompts can cause listener fatigue. High energy works for promotions but may feel aggressive during status checks, increasing abandonment rates in high-volume call centers.
- Reserve high-energy voices for promotional or upsell flows
- Avoid energetic presets for complex navigation instructions
- Monitor drop-off rates after energetic prompt segments
Mismatched Tone for Context
Using a deep, documentary voice like Orion for quick transactional updates creates a disconnect. Listeners expect neutral clarity for account balances, not a slow, authoritative narration style, leading to confusion and repeated calls.
- Match voice personality to information density
- Use neutral tones for data-heavy prompts
- Ensure tone aligns with brand voice guidelines
Ignoring Listener Fatigue
Long prompts with high-speed settings exacerbate listener fatigue. Even with optimal voice selection, excessive speed or lack of pauses forces users to re-listen, increasing average handle time. Test at 1.0x speed before deployment.
- Limit prompt length to under 15 seconds
- Avoid maximum speed settings for first-time callers
- Include natural pauses to aid comprehension
Integration with Call Center Software
Viva Voice Labs enables direct integration with standard PBX systems via simple HTTP triggers or middleware bridges. Developers can configure their telephony infrastructure to send specific text payloads to the API endpoint, receiving synthesized audio files ready for playback during call routing sequences.
This architecture supports both static and dynamic voice assignments, allowing call center operators to maintain consistent IVR experiences while adapting to real-time call contexts. The system handles the translation between telephony events and API requests efficiently.
PBX Integration Methods
Connect your PBX to the API using standard HTTP POST requests. Middleware tools like Asterisk or FreePBX can route text inputs directly to the speech endpoint, ensuring audio generation suitable for typical IVR hold messages and menu prompts; note that generation is not real-time streaming and may queue under load.
- Use standard HTTP POST requests for PBX routing
- Supports Asterisk and FreePBX middleware bridges
- Low-latency generation for standard menu prompts
API Triggering for Specific Flows
Trigger voice synthesis based on specific call flows, such as error messages or account status updates. By mapping distinct text strings to specific voice presets, you ensure that the tone matches the context of the customer interaction precisely.
- Map text strings to specific voice presets
- Trigger synthesis for error messages or status updates
- Ensure tone matches customer interaction context
Dynamic Voice Selection
Implement dynamic voice selection by passing the voice parameter in your API request. This allows your system to switch between male voices like Atlas for news updates or Milo for conversational assistance based on real-time call data or user history.
Note that voice selection does not affect the character-based pricing or rate limits, ensuring consistent operational costs regardless of the voice chosen for a specific call segment.
- Pass the
voiceparameter for dynamic selection - Switch between Atlas or Milo based on call data
- Pricing remains consistent across all voice choices
Cost Implications of Voice Choice
Voice selection does not alter your operational costs. Whether you deploy Atlas for news updates or Felix for promotions, the rate remains fixed at $0.01 per 1,000 characters. Your budget depends on text volume, not voice prestige.
Character count includes spaces and punctuation. Long IVR prompts increase costs linearly. Splitting text into 4,000-character chunks ensures consistent processing without extra fees.
Uniform Pricing Across Voices
All eight presets, including deep and energetic options, share identical character-based pricing. There are no premium tiers for specific vocal qualities.
You can switch voices mid-campaign without adjusting your cost structure. This allows flexible A/B testing without financial penalty.
- No voice-specific surcharges
- Consistent rate: $0.01 per 1,000 characters
Character Count Variations
Every character, including spaces, counts toward the 1 million character billing unit. Precise text management controls expenditure.
Failed requests incur no charge, protecting your budget from API errors during high-volume IVR calls.
Budgeting for Voice Changes
Switching from a calm voice to an energetic one requires no new contract or minimum spend. Balance does not expire.
Top-ups via USDT (TRC-20) on the TRON network are processed automatically, ensuring continuous operation regardless of voice selection changes.
Future-Proofing Your IVR
Viva Voice Labs provides a stable API with fixed rate limits of 60 requests per minute, ensuring predictable scaling for your IVR infrastructure. The character-based pricing model, at $0.01 per 1,000 characters, remains uniform regardless of voice selection or audio format, simplifying long-term cost forecasting for high-volume call centers.
Scalability of Voice Selection
Since all eight voice presets operate on the same underlying engine and pricing tier, you can scale voice usage without adjusting budget models. The consistent $10 per 1 million characters rate applies equally to deep documentary tones or neutral news styles, allowing flexible load distribution.
Updating Voice Catalogs
The voice catalog is static, consisting of eight presets derived from the LibriTTS-R dataset. As no new voices are added dynamically, your IVR logic can rely on a permanent list of available voices, eliminating the need to handle catalog shifts or deprecated voice IDs in your API integrations.
Maintaining Consistency
With a fixed set of voices, you can lock in specific audio personalities for your IVR flows without fear of sudden changes. This stability supports long-term brand consistency, as the same voice ID will always produce the same synthetic output, ensuring predictable listener experiences across years of operation.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. This voice is optimal for general call center announcements where clarity is paramount.
23 languages
Any voice speaks any of these languages.
