Frequently asked questions
Which voice sounds most human?
Human perception is subjective. Our voices are based on licensed recordings from LibriTTS-R and LibriVox. Each preset, like Aria or Felix, offers a distinct tone. You can test all eight voices via the /v1/voices endpoint to find the natural fit for your specific narration style.
Can I use Aria for both French and English?
Yes. Any of our eight voice presets can speak any of the 23 supported languages. Aria works for both French and English. If you send Latin-script text without specifying a language, the system defaults to English. For French, you should explicitly set the language parameter to ensure accurate pronunciation and intonation.
Is Orion too deep for young audiences?
Orion is described as deep and documentary-style. Whether it suits young audiences depends on your content. You can adjust the speed between 0.25 and 4.0 to change the delivery pace. We do not provide emotion tags or custom voice design, so you must test Orion directly with your target demographic to decide.
Do voices sound different in MP3 vs WAV?
The audio content remains identical regardless of format. You can choose from mp3, wav, opus, flac, or pcm via the response_format field. MP3 offers compression for smaller file sizes, while WAV provides uncompressed quality. The voice characteristics do not change between formats; only the file size and encoding differ.
Can I adjust the emotion of a voice?
No. Our API does not support emotion tags or custom voice design. The voices have fixed characteristics: Aria is warm, Luna is soft, and Felix is energetic. You can control the speed (0.25–4.0) and language, but you cannot dynamically alter the emotional tone of the generated speech during synthesis.
Which voice is best for audiobooks?
Luna is designed for soft, storytelling narration, making it highly suitable for audiobooks. Aria, with its warm narrator style, is also an excellent choice. Since voices have fixed tones rather than subjective rankings, we recommend testing both Luna and Aria with your specific text to determine which best fits your narrative style and pacing needs.
Do all voices support all 23 languages?
Yes. Every voice preset supports all 23 listed languages, including English, French, German, and Japanese. You can assign any voice to any language. If you do not set the language parameter, the system guesses it from the script, defaulting to English for Latin text. Explicitly setting the language ensures accuracy.
Is there a voice specifically for news?
Atlas is described as neutral and news-oriented. It provides a clear, professional tone suitable for news narration. While other voices can read news, Atlas is curated for this style. You can adjust the speed to match typical news pacing, but the fundamental tone remains fixed as neutral and professional.
How does speed affect voice quality?
Speed ranges from 0.25 to 4.0. Changing speed alters the playback rate without changing the pitch or voice identity. Higher speeds make the narration faster, which may affect clarity depending on the content. The underlying voice model remains the same; only the duration and rate of the audio file change.
Can I clone a voice for my podcast?
No. We do not offer custom voice cloning or allow you to upload your own voice. You must choose from our eight preset voices. These are based on licensed recordings from LibriTTS-R and LibriVox. You cannot create a unique voice model specific to your podcast using our current API.
Are the voices copyrighted?
The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. Generated files carry an inaudible watermark marking them as synthetic speech. We do not provide specific legal advice on usage rights, but the source materials have specific licensing terms you should review for commercial use.
Do voices sound different on mobile devices?
The audio file is generated on our servers and returned to you. The voice characteristics are fixed in the file. Differences in perception on mobile devices depend on your device’s speakers, volume settings, and audio codecs. The API output remains consistent regardless of where you play the file.
Can I use multiple voices in one episode?
Yes. You can generate separate audio files for each voice and join them. Send multiple requests to the /v1/audio/speech endpoint, specifying different voices for each segment. You can then concatenate the resulting audio files to create a single episode with varied narration styles or character voices.
Which voice is best for interviews?
Milo, described as friendly and conversational, is well-suited for interview-style narration. Atlas offers a neutral news tone. You can adjust the speed (0.25–4.0) to match a natural pace. Since we do not support emotion tags, choose Milo for a warm vibe or Atlas for a neutral tone. Both voices handle the required languages effectively, ensuring your podcast sounds professional without needing expensive studio time.
Female Voices Overview
Our female voices offer distinct tonal qualities tailored for podcast narration, ensuring your content resonates with listeners. Each voice is built from licensed LibriTTS-R recordings, providing a professional foundation for high-quality audio generation.
Aria: The Warm Narrator
Aria delivers a warm, clear narration style ideal for long-form storytelling or audiobooks. Her consistent tone reduces listener fatigue during extended episodes.
Nova: The Upbeat Host
Nova brings a bright, upbeat energy perfect for lively discussions or intro segments. Her dynamic delivery keeps audiences engaged from the first second.
Luna: The Storyteller
Luna offers a soft, storytelling voice that excels in narrative-driven content. Her gentle pacing creates an intimate atmosphere for personal stories.
Male Voices Overview
Male voices bring distinct energy to podcast narration, from high-octane promos to steady news delivery. Each preset is tuned for specific tonal qualities, ensuring your audio matches the genre's expectations without requiring studio engineering.
Felix: The Energetic Promo
Felix delivers a bright, upbeat tone perfect for promotional segments or dynamic intros. His energetic delivery keeps listeners engaged during short, punchy scripts where attention spans are shortest.
- Ideal for ad reads and show openers
- High energy, clear articulation
Milo: The Friendly Conversationalist
Milo sounds like a friend chatting over coffee, making him ideal for interview-style podcasts or casual storytelling. His conversational flow reduces listener fatigue during longer episodes.
- Best for dialogue-heavy content
- Natural, relaxed pacing
Atlas: The Neutral News Anchor
Atlas provides a balanced, professional tone suitable for news updates or factual narration. His neutrality ensures the content, not the voice, remains the focus for serious topics.
- Neutral, professional delivery
- Consistent volume and tone
Specialized Voices
While many voices handle general narration, two presets are optimized for specific audio textures that define professional podcast production. Sage and Orion provide distinct acoustic signatures suitable for educational and documentary formats respectively.
Both voices utilize the same underlying engine, ensuring consistent quality while offering tonal variety that keeps long-form content engaging without requiring multiple license tiers.
Sage: Calm Instructor
Sage delivers a calm, instructional tone ideal for educational podcasts and guided meditations. Its steady pacing helps listeners absorb complex information without feeling rushed.
Use Sage for tutorial segments or explainer episodes where clarity and patience are key. The voice remains neutral enough to avoid distracting from the core message.
- Ideal for step-by-step tutorials
- Reduces listener fatigue in long lessons
- Clear enunciation for technical terms
Orion: Deep Documentary
Orion offers a deep, resonant voice perfect for documentary-style storytelling. Its rich timbre adds gravitas to historical narratives or investigative journalism segments.
Pair Orion with ambient soundscapes for immersive experiences. The voice’s depth cuts through background music effectively, maintaining focus on the narrative arc.
- Best for narrative-driven episodes
- Adds emotional weight to stories
- Excellent for historical contexts
Voice Selection by Genre
Selecting the right voice for your podcast genre directly impacts listener retention and engagement. By matching vocal characteristics to content type, you ensure your narration feels intentional and professional without requiring expensive studio time.
Our curated voices are designed to handle the specific tonal demands of different podcast styles, from suspenseful storytelling to clear instruction.
True Crime Podcasts
For true crime narratives, Atlas provides a neutral, news-like delivery that maintains objective tension without becoming overly dramatic. Alternatively, Orion’s deep, documentary-style tone adds gravity and seriousness to case details, keeping listeners engaged through the mystery.
Use Luna for softer, storytelling-driven segments or Atlas for hard-hitting factual reporting. The choice depends on whether you want to emphasize the emotional arc or the factual timeline of the crime.
- Atlas for neutral, factual reporting
- Orion for deep, serious documentary tone
Educational Content
Sage is the ideal choice for educational podcasts, offering a calm, instructional tone that helps listeners absorb complex information without distraction. Its clear enunciation ensures that key concepts are delivered with precision.
Pair Sage with Nova for lighter, upbeat educational segments or quick tips. The contrast between Sage’s calm instruction and Nova’s bright energy can help maintain listener interest during longer lessons.
- Sage for calm, instructional clarity
- Nova for bright, upbeat tips
Comedy Shows
Nova’s bright and upbeat personality works well for comedy, adding energy to punchlines and light banter. Felix’s energetic promo style can also enhance comedic timing, especially for short skits or ad-reads within the episode.
Use Felix to inject high energy into comedic segments, ensuring the tone remains lively and engaging. Avoid slower voices like Aria or Luna unless you are aiming for a specific deadpan or ironic effect.
- Nova for bright, upbeat energy
- Felix for energetic, lively delivery
Voice Selection by Language
Language selection is critical for podcast authenticity. While the API guesses language from script, explicit ISO codes ensure precision for French or English narration. This prevents awkward defaults and guarantees the synthetic voice matches your target audience’s linguistic expectations.
French Narration
Set the language field to fr for consistent French output. Without this, Latin-script text defaults to English, which sounds unnatural for French narratives. Ensure your input text uses correct French orthography to maximize clarity.
- Use
language: "fr"for all French episodes. - Accents and special characters in text affect pronunciation.
- Default Latin script assumes English without explicit coding.
English Conversations
English narration benefits from explicit language: "en" setting to avoid ambiguity. Voices like Aria or Milo deliver clear, natural speech patterns. For conversational podcasts, adjust speed to 1.0–1.5 for a dynamic, engaging flow that mimics human pacing.
- Explicitly set
language: "en"for clarity. - Use speed 1.0–1.5 for conversational tone.
- Aria and Milo excel in English storytelling.
Multilingual Episodes
Switch voices or languages within a single episode by sending separate requests. Each request handles up to 4096 characters. Concatenate the resulting audio files to create seamless multilingual content. This allows mixing narration in different supported languages in one show.
- Split long texts into <4096 char chunks.
- Concatenate audio files for seamless playback.
- Mix languages by changing
languageper request.
Technical Voice Features
Technical precision ensures your podcast sounds professional. Our API lets you fine-tune audio delivery for specific narrative styles. Adjust speed and choose the right file format to match your production needs.
Pronunciation accuracy is handled automatically. The system detects language from the text script, defaulting to English for Latin characters. You can force a specific language code for non-English content to ensure correct accent and rhythm.
Speed Control Range
Adjust speech rate from 0.25x to 4.0x. This allows you to slow down complex explanations or speed up casual banter. Fine-tune the pace to match your podcast’s energy without re-recording.
Slower speeds aid clarity for dense educational content. Faster speeds keep comedy or news segments moving briskly. Test different rates to find the optimal listening experience for your audience.
- Range: 0.25 to 4.0
Format Compatibility
Export audio as MP3, WAV, OPUS, FLAC, or PCM. Choose MP3 for efficient storage or WAV for lossless editing. OPUS offers high quality at low bitrates.
FLAC provides lossless compression for archival quality. PCM is raw data for custom processing pipelines. All formats are generated instantly upon request, ensuring compatibility with major podcast hosting platforms.
Pronunciation Accuracy
The model guesses language from the script. Latin text defaults to English. For French narration or other Latin-script languages, set the language parameter explicitly to avoid English accent artifacts.
Non-Latin scripts are handled automatically for supported languages like Japanese or Chinese. If your text uses Latin characters but is in French or German, specify the ISO code to ensure correct phonetics.
Pricing Impact of Voices
Voice selection does not affect your billing rate. Whether you choose the warm narration of Aria or the deep documentary tone of Orion, the cost remains identical. This uniformity simplifies budgeting for podcasters who experiment with different personalities across episodes or series without worrying about variable rates for specific character types.
The pricing structure is strictly based on input length. Every character, including spaces and punctuation, counts toward the total. This transparency means you can accurately forecast costs for long-form narrations or short intro segments with precision, ensuring no hidden fees appear regardless of the voice model used.
Uniform Pricing Model
All eight voices share the same base rate of $10 per million characters. This consistency allows creators to mix voices freely within a single podcast series without adjusting their financial model. Whether producing a comedy show with Milo or a news segment with Atlas, the character cost is constant.
No Extra Cost for HD
The tts-1-hd model offers higher fidelity but charges the exact same rate as tts-1. You can upgrade your audio quality for documentary-style narration or premium intros without incurring additional charges. This makes high-definition audio accessible for all episode types, from casual conversations to detailed historical accounts.
Value per Character
At $0.01 per 1,000 characters, the API provides predictable costs for content planning. A typical 5,000-character script segment costs just $0.05. This low marginal cost encourages frequent A/B testing of voices to find the optimal tone for your specific audience without significant financial risk.
Failed requests are free, ensuring you only pay for successfully generated audio. This protects your budget during peak usage times when queueing might occur, as you are not charged for attempts that do not complete delivery of the audio file.
- $10 per 1 million characters
- $0.01 per 1,000 characters
- Failed requests are free
- No extra cost for HD audio

Audio Quality Comparison
The audio output quality depends on your chosen format. MP3 files are smaller, while WAV files preserve raw fidelity without compression artifacts.
MP3 vs WAV Clarity
MP3 offers efficient compression suitable for most podcast distributions. WAV provides lossless quality, ideal for archival or further editing in post-production studios.
- MP3: ~128kbps, efficient storage
- WAV: Lossless, higher fidelity
Noise Floor
The API generates clean audio with a consistent noise floor. The API generates clean audio with a consistent noise floor.
Dynamic Range
Audio maintains natural dynamic range across all formats. The synthesis engine preserves speech dynamics, ensuring emotional cues are clear without artificial compression artifacts common in cheap TTS.
Licensing Details
Our voices are built on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobooks from LibriVox. This foundation ensures your podcast narration carries a professional, natural tone without the hassle of clearing individual rights for each voice.
You get clear commercial rights to use the generated audio in your episodes, ads, and social media clips. The inaudible watermark marks the files as synthetic, but does not restrict your distribution or monetization.
LibriTTS-R License
The core voice data comes from the LibriTTS-R dataset, released under the Creative Commons Attribution 4.0 International license. This allows broad commercial use, provided proper attribution is maintained where required by your platform.
- CC BY 4.0 license for base recordings
- Clear, standardized commercial usage rights
- No hidden attribution complexities for standard podcast use
Public Domain Sources
Additional voice models are trained on readings from LibriVox, which consists of public-domain audiobooks. Since the underlying text and recordings are free of copyright restrictions, you can use the resulting synthetic voices without fearing copyright strikes.
Commercial Rights
You own the rights to the final audio files generated through our API. Whether you are producing a solo show or a sponsored segment, the audio is yours to distribute globally without additional royalty payments.
- Full commercial use for podcasts and videos
- No per-download or per-stream royalties
- Unlimited distribution across all platforms
Listener Experience
The voice you choose directly impacts how long listeners stay engaged. A mismatch between tone and content can cause fatigue, making even the best narration feel tiresome over time.
By aligning voice characteristics with your genre, you ensure a consistent auditory brand that keeps your audience invested from intro to outro.
Voice Fatigue Factors
Long-form podcasts benefit from softer, slower voices like Luna or Sage to prevent listener ear fatigue. High-energy voices like Nova or Felix work well for shorter segments but may become exhausting if used for hour-long episodes.
Pacing matters too. Using the speed parameter between 0.25 and 4.0 allows you to fine-tune the delivery. Slowing down complex explanations helps clarity, while speeding up casual banter keeps the energy up.
- Use Luna for long storytelling to reduce strain
- Limit Felix’s high energy to short promo segments
- Adjust speed to match natural speech patterns
Engagement Metrics
Consistent voice quality leads to higher completion rates. When the narration sounds professional and clear, listeners are less likely to drop off.
Clear audio ensures consistent quality, helping listeners connect with the story.
- Higher retention with consistent voice selection
- Clear audio formats improve emotional connection
- Professional sound reduces listener distraction
Audience Retention
Matching the voice to the genre—like Atlas for news or Orion for documentaries—builds trust and keeps listeners coming back.
Multilingual episodes using the same voice for different languages can create a unique, cohesive listening experience that stands out.
- Genre-specific voices build listener trust
- Multilingual consistency enhances brand identity
- Professional narration encourages repeat listens
Integration Tips
Integrating our TTS API into your podcast workflow ensures consistent audio quality across episodes. Use the same voice for intros to build brand recognition, while varying voices for different segments to maintain listener interest. Our API supports speed adjustments from 0.25x to 4.0x, allowing precise control over pacing.
Automating your audio generation saves hours of studio time. By sending text via POST requests, you can generate narration instantly. This approach works seamlessly with existing content management systems, letting you focus on storytelling rather than recording logistics.
Consistent Branding
Maintain a recognizable audio signature by selecting one primary voice for your show’s intro and outro. Since all voices support all 23 languages, you can keep branding consistent even when covering multilingual topics.
Use the speed parameter to match your standard pacing. Whether you choose Aria for a warm narrator feel or Felix for energetic promos, keeping the speed between 0.8x and 1.2x ensures a natural listening experience.
- Stick to one voice for recurring segments.
- Use consistent speed settings for brand identity.
- Leverage language tags for multilingual consistency.
Voice Switching Strategies
Switch voices between segments to keep listeners engaged. For example, use Luna for storytelling sections and Atlas for news updates. This variation prevents voice fatigue during long episodes.
You can switch voices within a single script by splitting the text into chunks. Each chunk is processed separately, allowing you to combine different vocal styles in the final audio file.
- Alternate voices for different topics.
- Split long scripts for varied tones.
- Use distinct voices for interviews vs. narration.
Automated Pipelines
Integrate the API with your content management system using simple HTTP requests. Send your script text to https://api.playgb168.com/v1/audio/speech and receive the audio file directly. This allows for instant generation of episode audio from written drafts.
Set up automated scripts to handle character limits. Since each request supports up to 4096 characters, your pipeline should split longer scripts into manageable chunks. Join the resulting audio files to create a complete episode.
- Use HTTP POST for instant generation.
- Split text at 4096 character limits.
- Automate audio assembly from chunks.
Future Voice Updates
The API currently offers eight core voices. While the catalog remains stable, new presets may be added periodically to support emerging podcast styles or languages. Updates are deployed to the live endpoint without requiring code changes from your integration.
New Voice Releases
New voices are released directly to the /v1/voices endpoint. You can check the endpoint regularly for additions to the eight current presets. There is no fixed schedule for new releases.
Voice Customization
The API does not offer custom voice cloning or uploading your own voice recordings. You cannot create unique voice profiles for specific brands. Each voice remains a fixed preset based on licensed LibriTTS-R recordings.
- No voice cloning
Feedback Loop
We monitor usage patterns to identify demand for specific tones. You can send feature requests via email. Popular requests may influence future voice selections or speed adjustments.
- Email support only
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. Each voice is built from licensed LibriTTS-R recordings, providing a professional foundation for high-quality audio generation.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- LibriTTS-R corpus (OpenSLR 141)
- MDN: Web Speech API
- IETF RFC 7845: Ogg Opus
- curl documentation
- ISO 639 language codes
- Creative Commons Attribution 4.0
- W3C WAI: making audio and video accessible
- openai-python on GitHub
- Python documentation
- OpenAI API reference: create speech
- LibriTTS-R paper (arXiv 2305.18802)
