Frequently asked questions
Which voice is best for French?
All eight voices support French. Aria offers a warm, clear narrator style suitable for general exhibits. Nova provides a bright, upbeat tone for interactive displays. Luna delivers soft storytelling for historical narratives. Sage is ideal for instructional content. Each voice maintains consistent volume and precise pronunciation, ensuring clarity for museum visitors regardless of the specific voice selected.
Can all voices speak all languages?
Yes, every voice can speak all 23 supported languages, including French, German, Spanish, and Japanese. You specify the language via the language parameter. If omitted, the system guesses the language from the script, defaulting to English for Latin text. This allows flexible voice selection independent of the target language for your audioguide content.
What is the difference between tts-1 and tts-1-hd?
Both models use the same engine and incur the same price of $10 per 1 million characters. The tts-1-hd designation indicates high-definition output, but both accept the same request fields. Choose tts-1 for standard processing or tts-1-hd for high-definition audio. Both support the same eight voices and 23 languages with identical speed and character limits.
Are the voices realistic?
The voices are based on licensed recordings from LibriTTS-R and LibriVox public-domain audiobook readings. They provide clear, consistent narration suitable for museum exhibits. Each generated file carries an inaudible watermark marking it as synthetic speech. The audio is realistic for informational content, though it does not include emotion tags or custom voice cloning features.
Do the voices sound like real people?
The voices are synthesized from licensed recordings of real people, specifically from LibriVox public-domain audiobook readings. They do not mimic specific celebrities or politicians. The output is synthetic speech with an inaudible watermark. While realistic for narration, they are distinct presets (e.g., Aria, Nova) rather than direct recordings of specific individuals.
Can I clone a voice?
No, custom voice cloning is not offered. You cannot upload your own voice or design custom voices. You must choose from the eight available presets: Aria, Nova, Luna, Sage, Felix, Milo, Atlas, or Orion. These voices are based on licensed recordings and are available for all supported languages without additional cloning fees.
What is the source of the voices?
The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. This source provides a foundation for the eight presets. The audio generated is synthetic, marked with an inaudible watermark, and uses text input to produce speech without storing the input text for training purposes.
Is there a license for commercial use?
The source recordings are from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. You can use the generated audio for commercial audioguides and museum exhibits. The API charges $10 per 1 million characters. No additional licensing fees are mentioned for the output audio beyond the standard usage costs.
How do I choose between Aria and Luna?
Aria is a warm, clear narrator voice, ideal for general museum exhibits and straightforward informational content. Luna is a soft, storytelling voice, better suited for historical narratives or immersive experiences. Both support all 23 languages. Choose Aria for clarity and Luna for a softer, more narrative tone in your audioguide applications.
What is the best voice for children?
Q: What voice is best for children? A: Nova offers a bright, upbeat tone, while Luna provides a soft, storytelling style suitable for children’s narratives. Both voices support all 23 languages, allowing you to match the audio to your specific content style, whether educational or entertaining, without language constraints.
Do voices have accents?
Voices are designed for clear narration and do not have strong regional accents. They support 23 languages with precise pronunciation. You can specify the language code to ensure correct pronunciation. The voices are based on standard recordings, providing consistent speech patterns suitable for international museum exhibits and tourism content without unintended accent variations.
Can I mix voices in one audio?
No, each request generates audio for a single voice. You cannot mix voices within one audio file. To switch voices, you must make separate API requests for each segment and join the audio files. This allows you to create varied audioguide experiences by combining different voice presets in your application logic.
What is the watermarked audio?
Every generated audio file carries an inaudible watermark that marks it as synthetic speech. This watermark identifies the audio as AI-generated without affecting playback quality. It is embedded automatically by the engine. This feature helps identify synthetic content in your museum audioguides or tourism applications without requiring additional metadata or visible indicators.
How do I test voice quality?
Send a short text via the POST /v1/audio/speech endpoint with your preferred voice and language. Use the free trial credit of $0.50 (50,000 characters) to test multiple voices. Check the returned audio file for clarity, speed, and tone. Adjust the speed parameter between 0.25 and 4.0 to optimize for your exhibit's pacing.
Voice Categories
Selecting the right voice is critical for museum audioguides, where clarity and tone must match the exhibit's context. Our engine provides eight voice presets. This guide helps product teams align voice selection with visitor expectations, ensuring a consistent and professional audio experience across all exhibits.
Narrator Voices
Aria, Luna, and Sage serve as reliable narrators. Aria offers a warm, clear tone ideal for general introductions. Luna provides a soft storytelling quality for historical narratives. Sage delivers a calm, instructional voice suitable for detailed exhibit descriptions.
- Aria: Warm, clear narrator
Conversational Voices
Milo and Nova create a more engaging, approachable atmosphere. Milo is friendly and conversational, perfect for interactive tours. Nova is bright and upbeat, suitable for modern art exhibits or family-friendly content. These voices help maintain visitor interest without feeling overly formal.
- Milo: Friendly, conversational
Documentary Voices
Felix, Atlas, and Orion cater to specific documentary styles. Atlas provides a neutral, news-like tone for factual accuracy. Orion offers a deep, authoritative voice for serious historical or scientific content. Felix adds energy for promotional or short-form audio clips.
- Atlas: Neutral, news
Female Voices
For museum audioguides, female voices offer distinct tonal qualities that shape visitor engagement. Aria, Nova, and Luna provide three reliable presets optimized for clear narration, energetic exploration, and soft storytelling. Each voice provides clear narration with consistent volume.
Aria: Warm Narrator
Aria delivers a warm, clear narrative style ideal for standard exhibit descriptions. Her tone remains steady and professional, making her suitable for general information where clarity outweighs emotional intensity.
- Best for: General exhibit overviews
- Trait: Warm, clear narrator
Nova: Bright Upbeat
Nova provides a bright, upbeat energy suitable for interactive displays or children’s areas. Her higher tempo and lively tone keep visitors engaged without sounding artificial or overly promotional.
- Best for: Interactive displays
- Trait: Bright, upbeat
Luna: Soft Storytelling
Luna offers a soft, storytelling voice perfect for historical narratives or emotional exhibits. Her gentle pacing allows complex stories to unfold without overwhelming the listener in quiet gallery spaces.
- Best for: Historical narratives
- Trait: Soft, storytelling
Male Voices
Male voices offer distinct tonal ranges for museum and tourism audio content. Each preset targets specific listener expectations, from energetic promotional spots to deep documentary narration.
Selecting the right male voice depends on the exhibit type. Use these presets to maintain consistent auditory identity across your audioguide collection.
Felix: Energetic Promo
Felix delivers an energetic, promotional tone ideal for short attraction highlights or ticket upsells. His voice is energetic and clear.
Use Felix for dynamic transitions or time-sensitive announcements where engagement is critical.
- High energy delivery
- Optimized for short segments
- Promotional focus
Milo: Friendly Conversational
Milo provides a friendly, conversational style suitable for interactive exhibits or self-guided tours. He sounds like a knowledgeable local guide rather than a narrator.
Ideal for casual historical tours or nature trails where a personal touch enhances the visitor experience.
- Approachable tone
- Conversational pacing
- Visitor engagement
Atlas: Neutral News
Atlas offers a neutral, news-style delivery perfect for factual historical data or technical exhibit descriptions. His voice remains unbiased and clear.
Use Atlas for informational plaques, safety instructions, or structured chronological narratives where clarity outweighs emotional appeal.
- Neutral delivery
- Clear articulation
- Fact-focused tone
Orion: Deep Documentary
Orion provides a deep, documentary-style voice for immersive historical contexts or solemn memorials. His lower register adds gravity to significant events.
Best suited for long-form historical tours or museum installations requiring a deep, documentary tone.
- Deep resonance
- Documentary style
- Immersive quality
Language Matching
Language handling is automatic but requires precision for optimal audioguide quality. The API guesses the language from the script, defaulting to English for Latin text. For non-English Latin-script content, explicitly set the language parameter to ensure accurate pronunciation.
Incorrect language detection can result in unnatural accents or mispronounced proper nouns, which disrupts the visitor experience. Explicit configuration eliminates ambiguity, especially in mixed-language exhibits or regions with multiple dominant languages.
French Pronunciation
Set language: 'fr' to ensure correct handling of silent letters, nasal vowels, and liaisons. This is critical for historical tours in Francophone regions where precise enunciation enhances immersion.
Without explicit language setting, French text might be read with an English cadence, reducing authenticity for local visitors.
- Set
language: 'fr'for French text. - Ensures correct liaison and nasal vowel handling.
- Improves authenticity for local visitors.
English Clarity
English is the default for Latin scripts. The engine uses tts-1 or tts-1-hd models optimized for clear, neutral pronunciation suitable for general museum content.
Use voice: 'atlas' for neutral, news-style clarity or voice: 'aria' for a warm, narrator tone. Both deliver high clarity for instructional content.
- Default language for Latin scripts.
- Use
atlasfor neutral clarity. - Use
ariafor warm narration.
Multilingual Mix
For exhibits spanning multiple languages, generate separate audio files for each language segment rather than relying on auto-detection. This ensures consistent pronunciation quality across the entire tour.
You can mix voices and languages within a single tour by concatenating audio files. For example, use nova for English segments and luna for French storytelling sections.
- Generate separate files per language.
- Concatenate audio for mixed tours.
- Mix voices per language segment.
Tone Analysis
Selecting the right vocal tone ensures your audioguide aligns with the exhibit’s atmosphere. Whether guiding visitors through a silent gallery or energizing a children’s zone, the voice sets the cognitive load and emotional context. Consistency in tone prevents listener fatigue during long audio tours.
The API provides distinct voice profiles to match specific content needs. Each voice is engineered for clarity and natural pacing, ensuring that complex historical details or simple directional cues are understood without distraction. Proper tone selection enhances retention and user engagement.
Informative Tone
For educational content, clarity is paramount. Voices like Atlas (Neutral News) or Sage (Calm, Instructional) deliver clear narration for instructional content without imposing unnecessary emotion. This approach minimizes cognitive load, allowing visitors to focus on historical facts or exhibit details.
Use these voices for detailed plaques, timeline explanations, or audio labels where precision matters more than drama. The neutral delivery ensures that complex terminology is articulated distinctly, supporting better comprehension for diverse audiences.
- Atlas: Neutral, news-style delivery for factual accuracy
- Sage: Calm, instructional tone for detailed explanations
- Best for: Historical timelines, artifact descriptions, educational plaques
Engaging Tone
To maintain interest in longer tours, use warmer, more dynamic voices. Aria (Warm Narrator) and Luna (Soft Storytelling) create an immersive experience suitable for art exhibits or cultural stories. Their natural inflection keeps listeners engaged without feeling artificial.
Nova (Bright Upbeat) works well for younger audiences or interactive exhibits where energy is desired. These voices add personality to the audio experience, making the tour feel like a guided conversation rather than a static recording.
- Aria: Warm, clear narrator for artistic context
- Luna: Soft, storytelling voice for cultural narratives
- Nova: Bright, upbeat tone for interactive exhibits
- Best for: Art galleries, cultural stories, family tours
Professional Tone
For corporate or formal museum settings, a balanced tone conveys authority and reliability. Milo (Friendly Conversational) bridges the gap between strict professionalism and approachability, making it ideal for guided tours in modern institutions.
Orion (Deep Documentary) offers a gravitas suitable for significant historical events or memorial exhibits. His deep, documentary-style voice adds weight to the content, signaling importance and respect for the subject matter.
- Milo: Friendly, conversational for modern institutions
- Orion: Deep, documentary tone for significant events
- Best for: Corporate tours, memorials, formal exhibitions
Use Case Matching
Matching voice characteristics to exhibit content ensures auditory coherence across diverse museum sections. Our API provides eight distinct presets, allowing product teams to align audio identity with curatorial intent.
Art Exhibits
Aria’s warm, clear narration suits static visual displays, while Luna’s soft storytelling enhances interactive installations. Use Nova for vibrant, contemporary pieces to maintain an upbeat tone that matches modern art aesthetics.
- Aria for classical galleries
- Luna for interactive media
- Nova for contemporary art
Historical Tours
Atlas provides a neutral, news-like delivery ideal for factual historical timelines. Orion’s deep, documentary style works well for immersive heritage sites, adding gravity to significant events without sounding overly dramatic.
- Atlas for factual timelines
- Orion for heritage sites
- Milo for casual history
Nature Trails
Sage’s calm, instructional voice is optimal for guided walking paths. Felix’s energetic promo style suits high-traffic visitor centers or short, engaging highlights. Ensure consistent speed settings (1.0x) to match natural walking paces.
- Sage for guided paths
- Felix for visitor centers
- Consistent 1.0x speed
Technical Constraints
Selecting the right voice for your audioguide requires balancing narrative intent with technical constraints. Each voice preset carries specific acoustic characteristics that influence how well content translates across different exhibit types, from quiet galleries to busy outdoor trails.
Understanding these limitations ensures consistent playback quality across all devices and languages without unexpected audio artifacts or timing issues.
Character Limits
Each API request accepts a maximum of 4096 characters, including spaces and punctuation. For longer tours, split the script into chunks that fit this limit and join the resulting audio files sequentially.
This constraint applies per request, so plan your segment boundaries carefully to avoid cutting mid-sentence.
- Max 4096 characters per request
- Spaces and punctuation count toward the limit
- Split long scripts into sequential chunks
Speed Variations
Adjustable speed ranges from 0.25x to 4.0x. Standard narration typically runs at 1.0x, generating roughly 25 characters per second. Slower speeds improve clarity for complex historical details, while faster rates suit quick transitions.
Under load, requests may queue, so test peak usage times to ensure consistent delivery during high-traffic museum hours.
- Range: 0.25x to 4.0x
- Typical rate: ~25 chars/sec at 1.0x
- Queueing may occur under high load
Format Compatibility
Output formats include mp3, wav, opus, flac, and pcm. Choose mp3 for efficient storage and web playback, or wav for high-fidelity local devices. Opus offers smaller file sizes.
All formats are returned directly in the API response, allowing immediate integration into your audio player or CMS without conversion steps.
- Supported: mp3, wav, opus, flac, pcm
- MP3 for web efficiency
- WAV for high-fidelity local playback

Testing Strategy
Testing ensures your audioguide audio meets technical standards before deployment. Use the API to generate short samples, verify pronunciation accuracy, and confirm volume consistency across different voice presets.
Sample Generation
Generate test clips using the POST /v1/audio/speech endpoint. Input up to 4096 characters per request to test long-form narration. Check that audio formats (mp3, wav, opus) render correctly in your target devices.
- Test edge cases like special characters and numbers.
A/B Testing
Compare two voices side-by-side for the same script. Use aria versus nova for art exhibits, or atlas versus orion for historical tours. Measure listener engagement metrics to determine which voice yields better completion rates.
User Feedback
Collect feedback from real visitors using the museum app. Ask specific questions about clarity and tone preference. Iterate on voice selection based on this data to optimize the final audio guide experience.
Consistency
Consistent audio delivery is critical for audioguide usability. Inconsistent pacing or volume disrupts the visitor experience, especially in noisy museum environments where clear narration is essential.
This section outlines how to maintain uniformity across your audio content through precise voice and speed configuration.
Voice Selection
Stick to a single voice per exhibit or tour segment. Switching voices mid-tour can confuse listeners and break immersion. For historical tours, Atlas provides a neutral baseline, while Luna offers a softer approach for storytelling sections.
Ensure the selected voice matches the content’s tone. Using Felix for a somber memorial might feel mismatched, whereas Orion suits deep historical documentaries.
- Use Atlas for neutral, informational content.
- Choose Luna for narrative-driven exhibits.
- Avoid switching voices within a single audio track.
Speed Settings
Set a consistent speed across all tracks for a specific tour. The default speed works for most narration, but slower speeds (0.75–0.9) improve clarity for complex historical terms. Keep this setting uniform to prevent jarring transitions between tracks.
Test the speed with your longest tracks to ensure they remain within the 4096-character limit per request, avoiding multi-part concatenation which can introduce slight pauses.
- Maintain a single speed setting per tour.
- Use 0.75–0.9 for complex terminology.
- Mixing speeds is technically possible but may cause jarring transitions.
Audio Quality
Generate all audio in a consistent format, such as MP3, to ensure uniform playback across different devices. MP3 offers a good balance of quality and file size for web-based audioguides.
Monitor the character count closely. Each request handles up to 4096 characters. Longer texts require splitting, so ensure the join points are seamless by using a consistent speed and voice.
Note that generated audio includes an inaudible watermark, marking it as synthetic speech. This does not affect listening quality but confirms the audio’s origin.
- Use MP3 for consistent web playback.
- Split texts at logical points under 4096 chars.
- Audio includes an inaudible synthetic watermark.
Licensing
Our voices are derived from licensed recordings sourced from LibriTTS-R, a dataset built from LibriVox public-domain audiobook readings. This foundation ensures that the synthetic voices we provide are legally grounded in well-established, open-source audio materials, offering product teams a reliable and clear path for commercial integration without complex rights negotiations.
LibriTTS-R License
The LibriTTS-R dataset is distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. This means the source recordings are free to use, modify, and distribute, provided appropriate credit is given. Our API leverages this license to deliver consistent, high-quality synthetic speech for museum and tourism applications.
- Open-source dataset for TTS training
- Commercial use permitted under CC BY 4.0
- No additional licensing fees for voice usage
CC BY 4.0
Under CC BY 4.0, you can integrate our audio outputs into your audioguide systems without worrying about restrictive copyright terms. The license allows for both non-commercial and commercial use, making it ideal for museum exhibits, tourism apps, and educational content where audio narration is a core component of the user experience.
- Attribution required for source dataset
- No warranty provided by original authors
- Compatible with most open-source projects
Public Domain Source
The original audiobook readings come from LibriVox, a volunteer-driven project where texts are in the public domain. This ensures that the underlying content—both text and audio—does not carry additional copyright restrictions. Combined with the CC BY 4.0 license, this creates a transparent and legally sound foundation for your audio products.
- Texts are public domain
- Volunteer-recorded audio
- No celebrity or performer rights issues
Audio Watermark
Every generated audio file includes an inaudible watermark that identifies the content as synthetic speech. This technical marker ensures that users and platforms can distinguish AI-generated narration from human-recorded audio without affecting the listening experience.
Inaudible Mark
The watermark is embedded directly into the audio waveform during synthesis. It remains transparent to museum visitors and tourists, preserving the immersive quality of the audioguide while maintaining a digital signature for provenance.
- Embedded during generation
- Inaudible to human ears
- Preserves audio fidelity
Synthetic Detection
Third-party tools can detect this specific watermark to verify that the audio was produced by our TTS engine. This allows museums and content distributors to audit their media library and confirm the source of their narration assets.
Transparency
Using synthetic voices with clear detection markers supports transparency in media consumption. It helps audiences understand that the narration is generated by AI, which is increasingly important for educational and historical content where authenticity matters.
Cost Implications
Pricing is uniform across all eight voices: $0.01 per 1,000 characters, regardless of voice selection or audio format. Failed requests incur no charge, ensuring you only pay for successful synthesis. The free trial provides $0.50 credit, sufficient for approximately 50,000 characters, valid for seven days upon Google sign-in.
Voice Pricing
All voices share the same rate structure. You do not pay extra for higher fidelity or specific character types. The cost scales linearly with text volume, making budgeting predictable for large museum collections.
- Uniform rate for all 8 voices
- No premium for 'HD' model
- Failed requests are free
Length Impact
Text is charged by character count, including spaces and punctuation. Requests are limited to 4,096 characters per call. Splitting longer tours into multiple requests does not change the per-character rate but may add slight overhead for connection setup.
- Spaces count as characters
- 4096 character limit per request
- No volume discounts
Budget Planning
Top-ups via USDT (TRC-20) are credited automatically. There are no monthly subscriptions or seats, so you only spend what you use. Balance does not expire, allowing you to top up infrequently for long-term projects.
- No monthly plans
- Balance never expires
- USDT TRC-20 payments only
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. Selecting the right voice is critical for museum audioguides, where clarity and tone must match the exhibit's context.
23 languages
Any voice speaks any of these languages.
