Frequently asked questions
Do you offer custom voice cloning?
No. AudioPilot uses eight fixed presets based on licensed recordings. We do not offer custom voice cloning for customers, nor do we allow you to upload your own voice or design new ones. You select from the existing set of eight voices for your text-to-speech needs.
Can I upload my own voice?
No. The service relies on eight static voice presets. There is no feature to upload a personal voice sample or create a custom clone. All generated audio uses the standard available voices, ensuring consistent quality without the need for additional voice training or data uploads.
Are the voices based on real actors?
Yes. Every voice is based on licensed recordings from real human speakers. These recordings originate from LibriTTS-R, which includes readings from LibriVox public-domain audiobooks.
What is the source of the voice recordings?
The voices are sourced from LibriTTS-R (CC BY 4.0), which contains readings from LibriVox public-domain audiobooks. The recordings are licensed under Creative Commons Attribution 4.0, ensuring clear usage rights for generated content.
Do I need to credit the voice actors?
A: Our voices are based on LibriTTS-R (CC BY 4.0) and LibriVox public-domain recordings. Since the API returns synthetic speech, you generally do not need to credit individual actors per file. However, verify your specific distribution channel’s requirements. Note that every generated file carries an inaudible watermark marking it as synthetic speech.
Is the audio watermarked?
Yes. Every generated audio file carries an inaudible watermark that marks it as synthetic speech. This watermark is embedded directly into the audio output by the engine. It helps identify the content as AI-generated without affecting the listening experience or the quality of the audio file.
Can I use these voices for branding?
Yes. You can use the generated audio for branding purposes, such as audiobooks, promos, or narration. Since the voices are based on public-domain or CC-licensed recordings, you can incorporate them into your commercial projects. There are no restrictions on using the synthetic output for brand identity.
How is consent handled for the recordings?
The source recordings come from LibriVox, where volunteers have consented to release their readings into the public domain or under CC BY 4.0. Since the data is publicly licensed, consent is already secured for the underlying recordings. You do not need additional consent to generate or distribute the synthetic audio.
Are there any restrictions on voice usage?
There are no specific restrictions on which voice you use for which content. All eight voices can speak any of the 23 supported languages. You can mix voices across different projects. The only limits are technical, such as the 4096-character input limit per request and rate limits.
Do the voices change over time?
No. The eight voice presets are static. They do not evolve or change based on usage or time. Each voice maintains its specific tone and characteristics consistently across all generations. You can rely on the same audio profile for Aria, Nova, Luna, and the other presets every time.
Can I modify the voice tone?
You can adjust the speed between 0.25 and 4.0, which affects the delivery pace. However, you cannot modify the fundamental tone, pitch, or emotion of the voice directly. There are no emotion tags or SSML controls. The voice character remains fixed, with speed being the primary variable.
Is there a limit on how many voices I can use?
No. You can use all eight available voices across your requests. There is no limit on the number of different voices you can call in a single project or over time. You can switch between voices like Aria, Felix, or Orion as needed for your narrative structure.
Do you provide voice samples?
Yes. You can view available voices, languages, and formats by calling the GET /v1/voices endpoint. This endpoint lists all eight presets and their details. You can generate sample audio by sending a text request to the speech endpoint to hear how each voice sounds with your text.
How do I choose the right voice?
Select a voice based on the desired tone: Aria for warm narration, Nova for upbeat content, Luna for storytelling, or Sage for instructional audio. For male voices, use Felix for energy, Milo for conversation, Atlas for news, or Orion for documentaries. Match the voice to your content's mood.
Defining Voice Cloning
Voice cloning is a synthetic voice replication technique that trains a model on specific audio samples from a target speaker. Unlike standard text-to-speech, which uses a fixed voice profile, cloning algorithms learn the unique timbre, pitch, and cadence of a particular individual to generate new speech that sounds identical to them.
This process typically requires hours of high-quality reference audio. The model analyzes these samples to build a custom voice representation, allowing for highly personalized narration where the AI mimics the donor's exact vocal characteristics across any text input.
Synthetic voice replication
Synthetic voice replication creates a digital model of a human voice. By analyzing acoustic features from existing recordings, the system generates new audio that retains the original speaker's unique identity. This allows for the creation of endless dialogue from a single voice profile.
- Digital reconstruction of vocal traits
- Preserves unique timbre and pitch
Training on specific samples
Cloning requires dedicated training data. You must provide specific audio samples, often hours of varied speech, to teach the model the nuances of a particular voice. This training phase is distinct from standard TTS, which uses pre-trained models with fixed voices.
- Requires hours of reference audio
- Analyzes vocal nuances and cadence
Custom voice generation
Once trained, the model enables custom voice generation. You can input any text, and the system produces audio that sounds like the cloned individual. This is ideal for projects needing a specific, recognizable voice without recording new sessions.
- Generates new dialogue from text
- Maintains consistent speaker identity
Our Approach: Preset Voices
AudioPilot relies on eight distinct, pre-defined voice presets rather than training new voices for each project. This approach ensures consistent, high-quality narration for freelance narrators and self-publishing authors without the variability of custom models.
Each voice is selected for its specific tonal qualities, such as the warm clarity of Aria or the calm instruction of Sage, making them ideal for various storytelling workflows.
8 fixed voice presets
We offer four female and four male voices, each designed for specific narrative styles. From the upbeat energy of Nova to the deep documentary tone of Orion, these presets cover the spectrum of audiobook narration needs.
These voices are ready to use immediately via the API. You simply select the voice that best matches your book’s mood, ensuring a professional result with every generation.
- Aria: Warm, clear narrator
- Nova: Bright, upbeat
- Luna: Soft, storytelling
- Sage: Calm, instructional
- Felix: Energetic, promo
- Milo: Friendly, conversational
- Atlas: Neutral, news
- Orion: Deep, documentary
Licensed from LibriVox
All voices are based on licensed recordings from LibriTTS-R (CC BY 4.0), derived from public-domain audiobook readings. This provides a solid legal foundation for commercial use.
The synthetic nature of the output is marked by an inaudible watermark, clearly identifying the audio as synthetic speech while maintaining high fidelity.
No custom training for users
Unlike voice cloning services, we do not train custom voices from your own recordings. You cannot upload a sample to create a unique clone.
This limitation keeps costs predictable and generation times fast, as you work with our established, high-quality presets rather than waiting for model training.
image]:
Consent and Licensing
AudioPilot voices originate from licensed recordings, ensuring clear commercial rights for your audiobook projects. Our library draws from public domain sources, allowing you to focus on narration without worrying about complex royalty structures or ongoing attribution fees for each generated file.
The foundation of our library consists of recordings released under CC BY 4.0 licenses from LibriVox. This means the source material is freely available for commercial use, providing a transparent legal baseline for self-publishing authors and freelance narrators who need reliable, rights-cleared audio assets.
Every audio file generated via our API carries an inaudible digital watermark. This mark identifies the output as synthetic speech, helping platforms and listeners distinguish between human-recorded and AI-generated content. It adds a layer of transparency without affecting the listening experience or audio quality.
CC BY 4.0 recordings
Our voices are built upon recordings licensed under CC BY 4.0, granting you broad commercial usage rights. This license ensures that the source audio can be modified and used in commercial products like audiobooks without additional per-unit royalties. It simplifies your legal workflow, allowing you to generate and distribute content without tracking individual attribution for every file.
- - Broad commercial usage rights included
- - No per-unit royalties required
- - Simplifies legal workflow for publishers
Public domain audiobooks
The voice models are trained on readings from public domain audiobooks. Since these original works have no remaining copyright restrictions, the voices themselves are free from exclusive ownership claims. This makes them ideal for freelance narrators and self-publishing authors who need versatile, royalty-free voices for any genre or project type.
- - Trained on public domain readings
- - No exclusive ownership claims
- - Ideal for any genre
Inaudible watermarks
Each generated audio file contains an inaudible watermark that marks it as synthetic speech. This feature helps platforms verify the origin of the content and maintains transparency with listeners. The watermark does not affect audio quality, latency, or the natural flow of the narration.
- - Identifies synthetic speech
- - Does not affect quality
- - Maintains listener transparency
Technical Differences
TTS synthesizes audio directly from text using fixed models, while cloning requires training a unique model on specific voice samples before generation. Our API skips the training phase, delivering audio in a single request.
Cloning’s two-step workflow adds time and complexity. Our direct approach ensures consistent output without the need for separate model preparation steps.
Text-to-Speech vs Cloning
TTS synthesizes audio directly from text using fixed models, while cloning requires training a unique model on specific voice samples before generation. Our API skips the training phase, delivering audio in a single request.
Cloning’s two-step workflow adds time and complexity. Our direct approach ensures consistent output without the need for separate model preparation steps.
Latency and processing
Cloning often involves a two-step process: training the voice model, then generating audio. Our TTS API generates audio directly in a single request. Typical speed is roughly real time on long texts (about 25 characters per second); short phrases take a few seconds. Requests are processed one after another, so under load a request may wait in a queue.
This linear processing model means you do not manage separate training jobs. You simply send text and receive an audio file, streamlining your production pipeline.
Quality consistency
Because our voices are based on licensed recordings, the output maintains a consistent tone and style across all generations. There is no risk of model drift or inconsistent training quality.
This consistency is crucial for audiobooks, where listeners expect a uniform narrative voice throughout the entire book, from chapter one to the final page.
Why Choose Presets for Audiobooks
Predictable quality
Preset voices like Aria, Luna, and Orion offer consistent narration styles suited for storytelling. Each voice is based on licensed LibriTTS-R recordings, ensuring uniform tone across your entire audiobook. You avoid the variability of custom models, maintaining a professional standard from chapter one to the end.
The eight presets cover warm narrators to deep documentary styles, letting you match the book’s genre without guessing. Since the engine is fixed, the output remains stable, which is critical for long-form content where listener immersion depends on voice consistency.
Lower cost than cloning
Cloning typically requires paying for custom voice training per project, whereas presets are included in the standard per-character rate of $0.01 per 1,000 characters. There are no extra setup fees or per-voice licensing costs. You pay only for the characters you generate, making budgeting straightforward for self-publishing authors.
With a flat rate and no minimum spend, you can produce a full-length audiobook without hidden costs. This model suits freelancers who need to keep overhead low while delivering high-quality narration without negotiating separate licenses for each voice.
Faster generation times
Using presets eliminates the training phase, allowing immediate synthesis. Text is processed at roughly real-time speed, generating about 25 characters per second. For a standard 50,000-character manuscript, this means significantly faster turnaround compared to cloning workflows that require setup and rendering time.
Requests are processed sequentially, but the lack of custom model dependency ensures predictable queuing. You can split long texts into 4,096-character chunks and join the audio files seamlessly, keeping your production pipeline efficient and reducing idle time between chapters.
Legal Considerations for Freelancers
Understanding licensing is critical for freelancers selling audiobooks. AudioPilot provides voices derived from LibriTTS-R, licensed under CC BY 4.0, and public domain recordings. This ensures clear commercial usage rights for the generated audio files you produce for clients or your own self-published titles.
You must verify that your specific distribution platform accepts CC BY 4.0 licensed audio. While the underlying recordings are public domain or openly licensed, the synthetic nature of the output requires careful adherence to platform-specific rules regarding AI-generated content.
Commercial usage rights
The voices are based on licensed recordings, allowing you to use the generated audio commercially. Since the source material is either public domain or CC BY 4.0, you generally retain the right to sell the resulting audiobooks without additional licensing fees per copy.
There are no monthly plans, no seats, and no minimum spend.
- CC BY 4.0 licensed source recordings
- Public domain audiobook readings
- No per-copy licensing fees
Attribution requirements
The CC BY 4.0 license for the source recordings may require attribution depending on how you distribute the audio. While the synthesized output is distinct, you should check if your platform requires disclosing that the voice was AI-generated.
AudioPilot does not provide an official SDK or plugin to auto-handle attribution, so you must manually ensure compliance with the original source licenses when bundling or distributing the final audio product.
- Check CC BY 4.0 attribution rules
- Disclose AI generation if required by platform
Platform terms of service
Different audiobook platforms have varying policies on AI-generated content. Some require explicit disclosure, while others have no restrictions. Always review the specific Terms of Service of your distribution channel before publishing.
Since we do not offer a dedicated WordPress plugin or browser extension to manage these details, you are responsible for ensuring your workflow complies with each platform's current guidelines regarding synthetic voices.
- Verify platform AI content policies
- Manual compliance management required
Common Confusions
Many users assume that preset voices are cloned from specific celebrities or actors, but AudioPilot uses eight fixed presets derived from licensed LibriVox recordings. These are not custom clones of any individual. Each voice is a distinct, consistent character—such as Aria or Felix—designed for storytelling rather than impersonation.
Another common misunderstanding is the expectation of granular control over delivery. Unlike some advanced TTS systems, AudioPilot does not offer emotion tags or SSML. You cannot manually dictate sarcasm or joy for a specific sentence. Similarly, there is no phoneme customization to tweak individual pronunciation, keeping the workflow simple and consistent for audiobook narration.
Preset voices are not clones
AudioPilot offers eight fixed voice presets, not custom clones of specific people. These voices are synthesized from licensed recordings, ensuring a consistent, character-driven tone suitable for narrating chapters. They do not replicate the unique timbre of a single celebrity or actor, but rather provide distinct narrative styles like the warm clarity of Aria or the deep documentary tone of Orion.
No emotion tags available
The API does not support SSML or emotion tags to manually adjust the tone of a sentence. You cannot inject specific emotions like sadness or excitement into a single line of text. The voice maintains a consistent emotional baseline throughout the generated audio, which works well for steady narration but lacks the dynamic range of human performance or higher-level editing tools.
No phoneme customization
There is no phoneme dictionary or custom pronunciation lexicon available in the API. If a word is mispronounced, you cannot adjust the phonetic breakdown to correct it. The system relies on standard language models to interpret the text input. For most standard English, Spanish, or French text, this works seamlessly, but specialized jargon may require manual text adjustments before generation.

Data Usage and Training
Your input text is used strictly to generate the audio file and is not added to our training datasets. This ensures your manuscript remains private and is not used to improve our models for other customers. The generated audio is returned directly in the API response and is not stored on our servers after delivery.
We retain request metadata—such as timestamp, character count, and status—solely for billing and usage verification. This data helps us track your consumption and ensure fair rate-limit enforcement. No audio files or source texts are kept for long-term storage or analysis.
Input text not used for training
Your input text is used strictly to generate the audio file and is not added to our training datasets. This ensures your manuscript remains private and is not used to improve our models for other customers.
Generated audio not stored
The generated audio is returned directly in the API response and is not stored on our servers after delivery. This minimizes data retention risks for your published content.
Metadata retention policy
We retain request metadata—such as timestamp, character count, and status—solely for billing and usage verification. This data helps us track your consumption and ensure fair rate-limit enforcement. No audio files or source texts are kept for long-term storage or analysis.
When to Use Cloning vs TTS
Choosing between cloning and TTS depends on your specific production needs. While cloning offers unique brand identity, preset voices provide consistent, reliable narration ideal for long-form storytelling. Presets eliminate the variability of custom training, ensuring every chapter sounds uniform.
For most freelance narrators and self-publishing authors, the trade-off favors presets. They offer predictable quality and faster generation times without the need for extensive sample collection or custom model training.
Branding needs vs storytelling
Custom cloning excels when you need a distinct, proprietary voice for brand recognition, such as a podcast host or a recurring character. However, for audiobooks, consistency across hundreds of pages is often more valuable than uniqueness.
Preset voices like Aria or Luna are optimized for narrative flow. They maintain a steady tone without the subtle inconsistencies that can arise from limited training data in cloning models.
- Cloning: Unique brand voice, higher maintenance
- Preset TTS: Consistent narrative tone, immediate availability
Cost efficiency
Voice cloning typically involves higher upfront costs for training and often higher per-unit pricing. With our TTS API, you pay a flat rate of $10 per million characters, regardless of the voice used.
For a 100,000-character audiobook, cloning might require a custom model fee plus usage costs. Preset voices incur only the character-based fee, making them significantly more economical for standard projects.
- Cloning: High initial training cost, variable unit price
- Preset TTS: No training fee, flat $10/1M chars
Scalability requirements
If you are producing multiple books or series, preset voices allow you to scale production quickly. You can generate chapters rapidly without waiting for custom models to train.
Our API handles high volumes efficiently, though requests are processed sequentially. Presets are optimized for this workflow, ensuring stable performance as your library grows.
Voice Quality and Realism
AudioPilot generates clear, natural-sounding narration suitable for audiobooks and self-publishing. Each voice preset is derived from licensed recordings, providing a consistent baseline quality without the variability of custom training.
By using fixed models, you avoid the inconsistency often seen in cloning services. The result is predictable, professional-grade audio that meets standard narration expectations for commercial use.
Human-like narration
Our voices use natural intonation and pacing, mimicking human speech patterns. This ensures that your audiobooks sound professional rather than robotic, which is critical for listener retention in the self-publishing market.
- Natural prosody and rhythm
Consistent tone
Each of the eight presets maintains a steady character throughout your text. Unlike cloning, which can drift in style, our presets like Aria or Atlas provide uniform tone from the first to the last chapter, ensuring a cohesive listening experience.
Limitations in emotion
While the narration is clear and engaging, the voices do not support explicit emotion tags or dynamic shifts in sentiment. The tone remains stable, which is often preferred for long-form storytelling where excessive drama can become fatiguing.
- No dynamic emotion control
Future of Voice Technology
Voice technology continues to evolve, but our focus remains on stable, character-based pricing and reliable generation for audiobook production. We prioritize clarity and consistency over fleeting gimmicks, ensuring your narration workflow stays predictable and affordable as the industry shifts toward scalable cloud solutions.
Advancements in TTS
Modern TTS engines deliver clear, human-like narration suitable for audiobooks. Our model processes text efficiently, allowing narrators to produce content rapidly without compromising on clarity.
Market trends
The self-publishing sector increasingly relies on API-driven audio production to lower costs and speed up release cycles. Creators value transparent character-based billing over complex subscription tiers, preferring direct, usage-based payments that scale with their output volume.
What to expect next
We anticipate continued improvements in voice variety and language support within fixed presets, rather than a shift toward expensive custom cloning for every user. Standardization in audio formats and watermarking will likely become more common to verify synthetic content, ensuring trust in distributed audiobook markets.
Our roadmap focuses on maintaining high uptime for our 8 preset voices and optimizing request queuing to keep generation times consistent, even during peak usage periods.
Support and Contact
Support for AudioPilot is handled exclusively through email. We do not offer a live chat widget, phone numbers, or ticketing portals to keep our operations lean and focused on delivering the API infrastructure you rely on for your audiobook production.
Email support only
Send your technical questions, billing inquiries, or API usage issues to our support team. This single channel ensures that every query is logged and addressed by someone with full context of your account and usage history.
No live chat available
We have intentionally omitted real-time chat to reduce overhead and avoid fragmented support threads. You will not find a pop-up widget on this site. If you need immediate assistance, email is the most reliable way to reach us.
Response time expectations
We aim to respond to all emails within one business day. During peak periods, such as major release updates, response times may vary slightly.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. Once you receive the audio files for each chunk, concatenate them in the correct order.
23 languages
Any voice speaks any of these languages.
