Frequently asked questions
Which voice is best for official notices?
Atlas is designed for neutral, news-style delivery, making it ideal for official notices. Sage offers a calm, instructional tone suitable for clear instructions. Both voices support all 23 languages. You can adjust the speed between 0.25 and 4.0 to ensure clarity for public information systems.
Can I use multiple voices in one project?
Yes, you can use multiple voices within a single project. Since the API processes text up to 4096 characters per request, you can split longer texts into segments and assign different voices to each segment. This allows you to combine voices like Atlas for news and Luna for storytelling in one application.
Is Aria good for long-form content?
Aria provides a warm, clear narrator voice well-suited for long-form content. The API handles texts up to 4096 characters per request; you must split longer inputs and join the audio files. Generated audio is returned directly in the response and not stored, ensuring privacy. You can adjust speed to manage generation time for accessibility tools.
How does Nova sound in German?
Nova sounds bright and upbeat in German. Any voice can speak any of the 23 supported languages, including German. You must set the language parameter to 'de' for non-English Latin-script text to ensure correct pronunciation. The voice maintains its energetic character regardless of the language used in the input text.
Is Felix too energetic for hospitals?
Felix is energetic and promo-focused, which may be too lively for some hospital settings. Sage offers a calm, instructional tone that might be more appropriate for medical instructions. You can adjust the speed between 0.25 and 4.0 to slow down Felix if needed, or switch to a more neutral voice like Atlas.
Does Milo work well in Spanish?
Milo is friendly and conversational, making it effective for Spanish audio. Spanish is one of the 23 supported languages. Any voice can speak any supported language, so Milo works well for conversational content. Ensure you set the language parameter to 'es' for accurate pronunciation and natural flow in your accessibility applications.
What is the difference between Luna and Sage?
Luna has a soft, storytelling tone, while Sage is calm and instructional. Luna is ideal for narrative content, whereas Sage suits clear instructions or notices. Both voices support all 23 languages. You can choose based on whether your public information system requires engaging storytelling or straightforward, calm delivery for user guidance.
Can Orion be used for documentaries?
Orion is described as deep and documentary-style, making it suitable for that use case. You can generate audio via the API and integrate it into your projects. The voice supports all 23 languages. Since audio is returned in the response and not stored, you retain full control over how you use the generated files.
Are voices gender-specific?
The voices are categorized as female (Aria, Nova, Luna, Sage) or male (Felix, Milo, Atlas, Orion). However, any voice can speak any of the 23 supported languages. The gender perception is based on the preset style, but you can use any voice for any language or content type regardless of its label.
Do voices change with speed?
Voices do not change their fundamental character with speed adjustments, but their perception may vary. You can set the speed between 0.25 and 4.0. Slower speeds make audio clearer for visually impaired users, while faster speeds reduce generation time. The voice remains consistent; only the playback rate changes within these limits.
Is there a voice for children?
There is no specific voice designed exclusively for children. The eight presets are adult voices with distinct styles like warm, bright, or deep. You can adjust the speed to make any voice sound slightly different, but the timbre remains that of an adult. No custom voice cloning or child-specific models are offered.
Can I use these voices in apps?
Yes, you can use these voices in apps via the REST API. Send a POST request to the speech endpoint with your text and voice selection. The API returns the audio file directly. You can integrate this into any application that supports HTTP requests, using standard SDKs or plain HTTP.
Do voices support dialects?
Voices support 23 specific languages but do not explicitly support distinct dialects within those languages. You can set the language parameter to ISO codes like 'de' or 'es'. The voice will use the standard pronunciation for that language. Dialect variations are not guaranteed, so test specific regional terms for your target audience.
Is there a voice for medical terms?
No voice is specifically optimized for medical terminology. All voices use standard pronunciation for the 23 supported languages. You may need to test specific medical terms to ensure clarity. Since there is no custom pronunciation lexicon, complex terms might require manual adjustment of the input text.
Criteria for selection
Selecting the right voice for public sector and NGO websites requires balancing clarity, authority, and approachability. Your audience includes visually impaired users and non-native speakers, so the audio must be instantly understandable without requiring cognitive effort to decode. A poor choice can obscure critical information, while the right voice ensures accessibility and trust.
The selection process should prioritize voices that minimize ambiguity. For government announcements, neutrality prevents bias. For community updates, warmth invites engagement. By mapping specific voices to content types, you ensure that the tone matches the message, whether it is a urgent alert or a friendly invitation.
Clarity and intelligibility
Public information must be consumed quickly and accurately. Voices like Atlas deliver a neutral, news-ready tone that reduces listener fatigue during long broadcasts. Aria provides a clear, warm narration style that helps non-native speakers follow complex instructions. Both options ensure that every word is distinct, reducing the need for users to replay audio.
- Use Atlas for factual, time-sensitive announcements.
- Choose Aria for detailed, step-by-step instructions.
- Avoid overly expressive voices for dense data.
Neutral tone for authority
When representing a government body or established NGO, the voice should convey reliability without imposing emotion. Atlas is designed for this purpose, offering a consistent, professional delivery that sounds like a standard news anchor. This neutrality helps users focus on the content rather than the speaker's personality, which is crucial for legal or regulatory updates.
Sage complements this with a calm, instructional tone. It is ideal for explaining new procedures or digital services. Its steady pace prevents the listener from feeling rushed, which is common when users are navigating unfamiliar government portals or accessing critical health information.
- Atlas provides a standard, unbiased news tone.
- Sage offers a calm, instructional pace.
- Avoid dramatic inflections in regulatory texts.
Warmth for community engagement
For community-focused content, such as local events or volunteer drives, a warmer voice builds connection. Aria’s soft, narrator style feels inviting rather than mechanical. This approach encourages users to stay engaged with longer texts, such as newsletters or community stories, without feeling like they are listening to a rigid machine.
Luna adds a storytelling quality that is perfect for historical context or charity narratives. Her soft delivery makes the content feel personal, which is essential for NGOs trying to humanize their mission. Pairing warmth with clear articulation ensures that emotional content does not sacrifice intelligibility for visually impaired users.
- Aria creates an inviting, narrative feel.
- Luna adds storytelling depth for charities.
- Use warmth to boost engagement in newsletters.
Top voices for news and updates
For public sector websites, clarity and trust are paramount when delivering news and updates. The right voice ensures information is accessible to visually impaired users and non-native speakers without introducing subjective bias or emotional distortion.
Atlas: Neutral and news-ready
Atlas provides a neutral, clear delivery ideal for breaking news, emergency alerts, and official city council announcements. Its consistent tone conveys authority without distraction, making it suitable for high-volume text generation where intelligibility is critical.
- Best for: Emergency alerts, daily news briefs, official records
- Voice type: Male, neutral
Sage: Calm and instructional
Sage offers a calm, instructional tone perfect for explaining complex procedures or government guidelines. Its steady pacing helps users process detailed information, such as tax forms or voting instructions, reducing cognitive load for non-native speakers.
- Best for: How-to guides, policy explanations, instructional content
- Voice type: Female, calm
Aria: Warm narrator style
Aria brings a warm, clear narrator style to longer-form news stories or community updates. While still professional, its slight warmth can make public information feel more engaging without sacrificing the clarity required for accessibility compliance.
- Best for: Community stories, feature articles, monthly newsletters
- Voice type: Female, warm
Voices for educational content
Educational content demands clarity and engagement. The right voice helps learners retain information without fatigue. Our API supports eight voices that can be used for teaching materials, ensuring your audio feels natural and accessible.
Luna: Soft storytelling
Luna offers a soft, storytelling tone ideal for narrative-based lessons or history modules. Her calm delivery reduces cognitive load for visually impaired users and supports non-native speakers who benefit from gentle pacing.
Milo: Conversational and friendly
Milo sounds like a helpful colleague. This voice works well for FAQs, step-by-step tutorials, or interactive quizzes. Its conversational style makes technical instructions feel less rigid and more approachable for diverse audiences.
Nova: Bright and upbeat
Nova brings energy to short updates, announcements, or vocabulary drills. Her bright tone keeps listeners engaged during repetitive educational segments. Use her for content that benefits from a lively, motivating atmosphere.
Voices for promotional material
Promotional content for public sector websites requires a balance of energy and trust. The right voice ensures announcements capture attention without sounding overly commercial or artificial. Selecting the appropriate tone helps convey urgency for time-sensitive updates or depth for detailed project explanations.
Our API offers distinct presets designed to match these specific communicative needs, allowing developers to tailor the auditory experience to the content's purpose.
Felix: Energetic and promo-focused
Felix brings an energetic quality ideal for promotional materials, event announcements, or community initiatives. His tone is upbeat and engaging, making him suitable for short bursts of information that need to grab attention quickly.
Use Felix when promoting local festivals, volunteer drives, or new service launches where a lively delivery encourages immediate user action.
- Ideal for time-sensitive public announcements
- Engages users in community event promotions
- Suitable for short, high-energy audio clips
Orion: Deep and documentary-style
Orion provides a deep, authoritative voice well-suited for documentary-style narrations or detailed project overviews. His tone conveys seriousness and depth, making him appropriate for explaining complex public works or historical context.
Deploy Orion for in-depth explanations of infrastructure projects, annual reports, or educational documentaries where a deep, narrative tone is preferred.
- Best for detailed project overviews
- Conveys authority and seriousness
- Ideal for historical or educational narratives
Matching tone to message
Selecting between Felix and Orion depends on the emotional weight of the message. Use Felix for high-energy updates and Orion for substantive, detailed content.
Ensure consistency by mapping specific content types to these voices in your API configuration, maintaining a coherent auditory brand across your public information systems.
image}:
Language-specific considerations
Selecting a voice for public sector content requires understanding how each preset handles different languages. Since any voice can speak any of the supported 23 languages, consistency depends on choosing the right preset rather than the language itself. This approach simplifies maintenance for NGOs operating across borders.
Voice consistency ensures that your audience recognizes your brand regardless of the language. By assigning specific voices to specific content types, you maintain a uniform auditory experience, which is crucial for accessibility and trust in public information systems.
Voice consistency across languages
Every voice can speak all 23 supported languages, from Arabic to Turkish. This flexibility allows you to maintain a consistent tone across multilingual websites without managing multiple voice profiles. For instance, Atlas remains neutral and authoritative in both English and German, ensuring clear news delivery.
Use the language parameter to specify the ISO code. While the API guesses language from the script, explicitly setting it prevents errors, especially with Latin-script text that isn't English. This guarantees the intended voice pronunciation and style are applied correctly.
- - Support for 23 languages including Spanish, French, and Hindi
- - Any voice can speak any supported language
- - Explicit
languageparameter prevents guessing errors
Accent and dialect neutrality
The voices are designed to be neutral and clear, avoiding strong regional dialects. This makes them ideal for public information where clarity is paramount. Sage’s calm instructional tone works well for diverse audiences, while Milo’s conversational style remains friendly without being overly colloquial.
When serving non-native speakers, neutral accents reduce cognitive load. The API’s synthetic nature ensures consistent pronunciation, unlike human recordings that might vary by region. This reliability is key for accessibility tools used by visually impaired users or those with hearing impairments.
- - Neutral accent design aids comprehension
- - Consistent pronunciation across regions
- - Ideal for non-native speaker audiences
Testing non-English scripts
Always test your content in the target language before deployment. Use the GET /v1/voices endpoint to verify available formats and languages. Generate short samples to check for clarity, especially with complex scripts like Chinese or Arabic.
Monitor character counts carefully, as spaces and punctuation count toward the 4096-character limit. Ensure your text preprocessing handles special characters correctly to avoid synthesis errors. Regular testing helps maintain high-quality audio outputs across all supported languages.
- - Test samples in target languages
- - Verify
GET /v1/voicesendpoint data - - Handle special characters in text preprocessing
Technical setup for voice selection
Integrating text-to-speech requires mapping specific voices to content categories to maintain consistent user experience. This approach ensures that news updates sound authoritative while educational materials remain engaging. Developers can assign voice presets based on content type, such as using Atlas for public notices and Luna for storytelling segments.
The API provides a GET /v1/voices endpoint to retrieve available voices, their IDs, and supported languages. This allows dynamic configuration without hardcoding values. Developers should fetch this list at startup to populate their voice selection menus, ensuring alignment with current API capabilities.
Mapping voices to content types
Assign voice presets based on content function. Use Atlas for neutral news, Sage for instructions, and Aria for narration. This standardization helps visually impaired users recognize content type by voice alone, improving navigation efficiency in public information systems.
- News: Atlas (neutral)
- Education: Sage (calm)
- Storytelling: Luna (soft)
Using the voices endpoint
Call GET /v1/voices to list available presets. The response includes voice IDs like atlas or sage and supported languages. Cache this data locally to reduce API calls and ensure your application knows which voices are active.
- Endpoint: GET /v1/voices
- Returns: voice ID, name, language support
Dynamic voice switching
Implement logic to switch voices per request. Send the voice parameter in the JSON body alongside input and model. This allows different sections of a page to use different voices without reloading the entire page.
- Set
voicein request body - Support multiple voices per session
Accessibility compliance
Integrating AI-generated audio into public sector websites requires adherence to accessibility standards to ensure content is usable by everyone, including visually impaired users and non-native speakers. The synthetic nature of the voices, while human-like, must be implemented with clear controls and consistent labeling to meet these expectations effectively.
This section outlines how to structure audio delivery to comply with WCAG guidelines, allowing users to easily control playback and understand the nature of the content being presented.
WCAG guidelines for audio
Ensure all audio content has visible controls for play, pause, and volume. Provide text alternatives or transcripts for any audio that conveys essential information, satisfying WCAG 1.2.1 and 1.2.2 criteria. Avoid autoplay without user interaction to prevent disruption for screen reader users.
The inaudible watermark identifying synthetic speech does not affect accessibility but helps users understand the content type. Ensure audio elements are properly labeled with ARIA attributes so assistive technologies can identify them as generated speech rather than live recordings.
- Provide explicit play/pause controls for all audio blocks.
- Include text transcripts for audio-only content.
- Label audio elements with ARIA attributes like
role="region".
User preference settings
Allow users to choose their preferred voice or disable audio entirely. Since the API supports multiple voices, let users select from options like Aria for news or Milo for conversational content via a simple dropdown. Store these preferences in cookies or local storage for a consistent experience.
Enable users to adjust the speed of the audio between 0.25x and 4.0x to match their listening speed. This flexibility is crucial for non-native speakers or users with cognitive disabilities who may need slower or faster playback rates.
- Offer a voice selection dropdown in settings.
- Include a speed adjustment slider (0.25–4.0x).
- Allow users to mute all site audio globally.
Standardized voice usage
Assign specific voices to consistent content types across the site. Use Atlas for news updates to convey authority, and Luna for storytelling sections to provide warmth. This consistency helps users develop familiarity and trust with the audio content over time.
Clearly indicate when audio is AI-generated. While the voice is realistic, transparency about the synthetic nature helps manage user expectations. Use labels like "AI Narrated" to ensure clarity without detracting from the user experience.
- Use Atlas for official announcements.
- Use Luna for community stories.
- Label audio as 'AI Narrated' for transparency.

Cost analysis by voice
All eight preset voices—Aria, Nova, Luna, Sage, Felix, Milo, Atlas, and Orion—share identical pricing structures, ensuring predictable costs regardless of which voice you select for your public information system. The cost is strictly $0.01 per 1,000 characters, with spaces and punctuation counting toward the total. This uniformity simplifies budgeting for NGOs and indie developers who mix voices across different content types without worrying about variable rate cards.
Uniform pricing across voices
Whether you generate short announcements using Atlas or longer educational narratives with Luna, the charge remains consistent. You pay only for the character count in the input text, making it easy to forecast expenses for accessibility tools or public websites. Failed requests are free, so you only pay for successful audio synthesis.
- $0.01 per 1,000 characters
- All 8 voices cost the same
- Spaces and punctuation count as characters
- Failed requests are free
Model choice impact on cost
The API offers two models: "tts-1" and "tts-1-hd". Both engines deliver the same voice presets and incur identical pricing per character. The choice between them depends on audio quality preferences rather than budget constraints, as there is no cost difference for selecting the high-definition model over the standard one.
Budgeting for variety
Since you can switch voices dynamically without price changes, you can assign specific tones to different content categories—such as using Felix for promo materials and Sage for instructional guides—without impacting your total spend. Start with the $0.50 Google sign-up trial to test voice diversity, then top up via USDT (TRC-20) for ongoing usage.
Case studies in public sector
Public sector organizations benefit from consistent audio branding. By selecting specific voices for distinct content types, institutions can improve accessibility and user engagement across their digital platforms.
City council announcements
Use Atlas for neutral, authoritative updates on zoning changes or meeting minutes. Its clear, news-ready tone ensures critical information is understood by all residents, including those with visual impairments.
- Atlas for official notices
- Sage for instructional guides
NGO fundraising campaigns
Deploy Aria’s warm narration for emotional stories or Felix’s energy for urgent calls to action. This variety keeps listeners engaged while maintaining a professional standard.
- Aria for storytelling
- Felix for promos
Museum audio guides
Luna’s soft storytelling style suits historical narratives, while Orion’s deep tone works well for documentary-style exhibits. Both offer consistent quality across multiple languages.
- Luna for history
- Orion for documentaries
Common voice combinations
Pairing distinct voices with specific content types creates a structured audio experience for public sector websites. This approach helps users navigate between different information categories without cognitive fatigue, ensuring that news updates feel distinct from educational content.
By standardizing these combinations, indie developers can maintain consistency across their accessibility tools, making digital public services more intuitive for visually impaired users and non-native speakers who rely on clear audio cues.
Alternating between narrator and news
Use Aria for general narration and Atlas for breaking news updates. This contrast helps listeners distinguish between context and urgent information.
Atlas’s neutral tone conveys authority for city council announcements, while Aria’s warmth keeps community engagement high during routine updates.
- Aria for narrative context
- Atlas for urgent updates
Using Milo for Q&A sections
Milo’s conversational style suits FAQ sections, making information feel approachable rather than bureaucratic.
Pair Milo’s friendly tone with simple text to ensure clarity for non-native speakers accessing public services.
Pairing Luna with educational text
Luna’s soft, storytelling voice works well for museum audio guides or historical context.
Her calming presence reduces listener fatigue during longer educational segments compared to faster-paced voices.
Quality assurance tips
Quality assurance ensures your TTS outputs meet public sector standards for clarity and accessibility. By systematically reviewing audio samples and testing with real users, you can verify that synthetic speech remains intelligible across different devices and contexts. This process helps maintain consistent user experience for visually impaired and non-native speakers relying on your content.
Regular checks prevent subtle degradation in audio quality and ensure your chosen voices align with the intended tone of your public information systems.
Listening to sample outputs
Before deployment, generate short samples for each voice and language combination. Verify that the audio sounds natural and matches the expected tone. Check that the speed settings (0.25–4.0) do not distort intelligibility, especially for complex announcements.
Review the generated MP3 or WAV files to ensure they load correctly in your web player. Confirm that the audio length approximates real-time generation, allowing you to estimate processing times for longer texts.
- Test samples on mobile and desktop browsers.
- Verify audio formats (mp3, wav, opus) are supported by your target devices.
- Check volume levels for consistency across voices.
Checking for artifacts
Listen closely for unnatural pauses, clipping, or robotic intonation that may disrupt the user experience. Since requests are processed sequentially, ensure that long texts split into parts of up to 4096 characters join smoothly without awkward gaps or volume shifts.
Use the tts-1-hd model if you need higher fidelity for critical announcements. Remember that all files carry an inaudible watermark, so you cannot distinguish synthetic audio from natural recordings by ear alone.
- Check for clipping or distortion at high speeds.
- Verify smooth transitions between split text parts.
- Ensure no unexpected silence between sentences.
User testing with audiences
Engage target users, such as visually impaired individuals or non-native speakers, to test the audio in real-world scenarios. Gather feedback on clarity, pace, and emotional tone. This helps identify issues that automated checks might miss, such as difficulty understanding specific accents or dialects.
Use this feedback to refine voice selection and speed settings. For example, Atlas may be preferred for news, while Luna works better for storytelling. Regular user testing ensures your TTS solution remains accessible and effective.
- Test with screen readers on iOS and Android.
- Get feedback from non-native speakers for clarity.
- Survey users on audio comfort and understanding.
Maintenance and updates
Maintaining consistent audio quality in public sector applications requires periodic review of your voice configurations. Since the API generates speech dynamically, you should periodically test your selected voices against your content types to ensure they still match your accessibility goals. Regular checks help identify if a voice has become too monotonous for long-form announcements or if a warmer tone is needed for community updates.
You can also adjust your voice selections based on user feedback or seasonal changes in your content strategy. For instance, switching from a neutral news voice to a warmer narrator voice during holiday campaigns can improve engagement. Ensure your dynamic voice switching logic is robust and that users can easily revert to their preferred settings.
Monitoring voice performance
Track usage patterns and user feedback to ensure voices remain effective. Regularly test edge cases, such as mixed-language texts or special characters, to maintain clarity.
Use automated scripts to verify that audio generation remains within expected character-per-second speeds, ensuring timely delivery for real-time updates.
- - Review user feedback quarterly.
- - Test new content types with selected voices.
- - Monitor API response times for delays.
Updating voice selections
Re-evaluate voice choices as your content evolves. If you introduce more instructional content, consider switching to Sage for clarity. For promotional materials, Felix or Nova may be more engaging.
Ensure your mapping logic is updated to reflect any new content categories or audience segments you are targeting.
- - Align voices with content type.
- - Update mappings for new content.
- - Test with new audience segments.
Handling voice changes
If voice presets are updated or new voices are added, test them thoroughly before deploying to production. Communicate any significant changes to your users if their experience is affected.
Keep a record of voice versions and changes to facilitate easy rollback if a new selection proves less effective.
Remember that voice consistency across languages is key; ensure that the same voice is used for the same content type in all supported languages.
- - Test new voices before deployment.
- - Communicate changes to users.
- - Maintain voice consistency across languages.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. Allow users to choose their preferred voice or disable audio entirely.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- openai-python on GitHub
- W3C WAI: making audio and video accessible
- ISO 639 language codes
- curl documentation
- LibriTTS-R corpus (OpenSLR 141)
- Creative Commons Attribution 4.0
- LibriVox public-domain audiobooks
- LibriTTS-R paper (arXiv 2305.18802)
- OpenAI API reference: create speech
- TRON developer hub: TRC-20
- openai-node on GitHub
