Frequently asked questions
Who runs this site?
Voctet is an independent storefront for the text-to-speech API. We are not affiliated with OpenAI, ElevenLabs, or any other major brand. Our team focuses on providing a reliable synthetic voice endpoint for indie developers building ML datasets and TTS models, ensuring consistent, watermark-marked audio generation.
What is the editorial team?
The editorial team consists of indie developers and enthusiasts who understand the needs of ML dataset creators. We focus on clear, specific documentation and honest descriptions of our API capabilities. Our content is written by the same people who manage the API, ensuring technical accuracy and relevance.
How are voices licensed?
Voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobook readings from LibriVox. This foundation allows us to offer eight distinct voice presets, including female and male options, ensuring high-quality synthetic speech for your machine learning datasets without complex licensing hurdles.
What is CC BY 4.0?
CC BY 4.0 is a Creative Commons Attribution license that allows you to share and adapt the material for any purpose, including commercially, as long as you give appropriate credit. Our voice data originates from this license, ensuring clear legal standing for your ML projects and dataset usage.
Is my data used for training?
No. Input text is used only to produce the audio and is not used for training models. Generated audio is returned in the response and not stored. We retain request metadata like time, characters, and status solely for billing purposes, ensuring your data privacy and integrity.
How do I contact support?
Support is available by email only. We do not offer phone support, live chat, or dedicated account managers. You can reach us via the contact page. We aim to provide helpful, direct assistance for any issues regarding your API key, billing, or technical implementation questions.
Do you have phone support?
No, we do not have phone support. All support is handled via email. This allows us to maintain a lean operation and provide detailed, written responses to your technical questions. You can find our email address on the /contact/ page for any inquiries or issues.
What is the data policy?
Input text is used only to generate audio and is not used for training. Generated audio is returned in the response and not stored. We keep request metadata (time, characters, status) for billing purposes. This ensures your data remains private and your dataset creation process is transparent.
Where do voices come from?
Voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobook readings from LibriVox. We offer eight presets, including female and male voices, designed for consistent synthetic speech. This foundation ensures high-quality audio output suitable for training machine learning models and generating datasets.
Is LibriVox used?
Yes, our voices are based on recordings from LibriVox public-domain audiobook readings, specifically the LibriTTS-R dataset under CC BY 4.0. This provides a robust foundation for our eight voice presets, ensuring high-quality synthetic speech for your machine learning datasets without the need for complex licensing or additional fees.
Can I use the data commercially?
Yes, you can use the generated audio commercially. Our voices are based on CC BY 4.0 and public-domain sources, allowing commercial use with appropriate attribution. We provide a reliable API endpoint for indie developers, ensuring consistent, watermark-marked synthetic voices for your ML datasets and commercial projects without extra licensing hurdles.
Do you offer enterprise plans?
No, we do not offer enterprise plans, monthly subscriptions, or SLAs. We provide a simple, pay-as-you-go model with no minimum spend or seats. This approach is ideal for indie developers who need flexible, reliable access to synthetic voice data without the complexity of enterprise contracts or dedicated support tiers.
What are the limitations?
We do not offer streaming, SSML, custom voice cloning, or our own SDK. Requests are limited to 4096 characters, and we have rate limits of 60 requests per minute. We also lack phone support and SLAs. These limitations keep our API simple, reliable, and focused on providing structured audio files for ML.
Who is the audience?
Our primary audience is indie developers building ML datasets and TTS models. We provide a reliable text-to-speech API endpoint for those who need consistent, watermark-marked synthetic voices. We focus on delivering structured audio files instantly, helping developers avoid the hassle of scraping or manual recording for their projects.
Our Mission
Voctet exists to solve the data bottleneck for indie developers building speech models. We provide a reliable API endpoint that generates structured, synthetic audio instantly, replacing the slow, expensive process of scraping or manual recording. Our goal is to give small teams access to consistent, high-quality voice data without the enterprise overhead.
Empowering Indie Devs
We focus on the developer who needs volume without complexity. By offering a straightforward API compatible with standard SDKs, we let you generate thousands of samples programmatically. This removes the friction of managing physical recording studios or negotiating with voice actors, allowing you to iterate on your ML models faster and cheaper.
- Direct API access for programmatic generation
- No minimum spend or enterprise contracts
- Consistent synthetic voices for reproducible datasets
Democratizing Voice Data
Voice data should not be a luxury reserved for big labs. Our transparent pricing ($0.01 per 1,000 characters) and simple TRC-20 payments make large-scale dataset creation accessible to anyone with an internet connection. We remove the gatekeeping of traditional voice licensing, enabling indie projects to compete on data quality rather than budget.
- Transparent, character-based pricing
- No hidden fees or subscriptions
- Accessible to solo developers and small teams
Simplifying ML Training
Training models requires massive, varied datasets. Voctet provides 8 distinct voices across 23 languages, all watermarked to clearly identify synthetic speech. This consistency helps your algorithms learn robust features without the noise of human variability. You get clean, labeled audio files ready for immediate ingestion into your training pipelines.
- 8 licensed voices for diverse training data
- Inaudible watermarks mark synthetic speech
- Structured JSON responses for easy integration
Voice Licensing
Voctet voices are derived from licensed recordings in the LibriTTS-R dataset. These voices are based on LibriVox public-domain audiobook readings, ensuring that the synthetic audio retains a clear lineage to licensed source material.
LibriTTS-R Source
Our eight voice presets—Aria, Nova, Luna, Sage, Felix, Milo, Atlas, and Orion—are synthesized from high-quality source recordings. This foundation allows us to offer consistent, high-fidelity audio suitable for ML datasets without the legal ambiguity of unstructured web scraping.
CC BY 4.0 License
The underlying data originates under the CC BY 4.0 license, but the source readings come from LibriVox public-domain audiobooks. This open standard permits commercial use and modification, provided appropriate credit is given. For indie developers building commercial TTS models or ML datasets, this license structure reduces overhead compared to proprietary voice sources.
- Commercial use permitted
- Modification allowed
- Credit required
Public Domain Audio
The source readings come from LibriVox, a collection of public-domain audiobooks. This ensures that the raw material used to train our synthetic voices is free from restrictive copyright claims. You get structured audio files with consistent quality, backed by transparent licensing origins.

Data Policy
We prioritize privacy and clarity in how your data is handled. When you send text to our API, we use it strictly to generate the requested audio file. This input is not stored, reused, or repurposed for any other purpose once the synthesis is complete.
Input Text Usage
Your text is processed in real-time to produce the audio response. We do not retain the original text content after generation, ensuring your source material remains yours.
Audio Storage
Generated audio files are returned directly in the API response and are not stored on our servers. You own the output files you download, allowing you to integrate them freely into your ML datasets or applications without restriction.
Training Data Rights
Input text is not used for training our models. The synthetic audio we generate carries an inaudible watermark but is not added to our training data. You retain full rights to use the generated audio for any purpose, including commercial ML training.
Team & Editorial
Voctet is built by indie developers who live in the trenches of ML data collection. We know the pain of manual recording and inconsistent audio quality, which drives our focus on providing reliable, structured synthetic datasets for training models.
Editorial Team
Our small, hands-on team handles all technical documentation and product updates directly. We are developers first, ensuring that every detail in our guides reflects actual API behavior and real-world usage scenarios.
Developer Focus
We prioritize practical utility over marketing fluff. Every feature, from character limits to watermarking, is designed to streamline the integration of synthetic voice data into your ML pipelines without unnecessary complexity.
Honest Reporting
We report facts as they are, not as we wish them to be. If a feature is missing or a limit exists, we state it clearly so you can plan your dataset generation accurately.

Contact Support
We keep things simple for indie developers who just want to build. Support is handled entirely through email, ensuring every query is tracked and answered without the noise of live queues or automated bots.
There is no phone number and no live chat widget to distract you while you’re coding. We believe asynchronous communication works better for technical issues, allowing you to provide detailed logs and context for a precise answer.
Email Only
Send your questions to our support team directly via email. This is the only channel we use for all inquiries, from billing issues to technical questions about the API.
- Single channel for all support needs
- No phone number to call
- No live chat widget
No Live Chat
We do not offer real-time chat support. This allows our team to focus on resolving technical issues thoroughly rather than managing multiple simultaneous conversations with limited context.
Response Time
Since we are a small team focused on the API itself, please include relevant details like your API key and error messages to help us assist you faster.
- Typical response within 24 hours
- No guaranteed SLA

Limitations
Voctet is built for indie developers and hobbyists who need reliable synthetic voice data without the overhead of enterprise contracts. We prioritize simplicity and transparency, offering a straightforward API for generating ML datasets rather than a full-featured production platform with bells and whistles.
Our model is designed for efficiency and cost-effectiveness, but this means we omit certain enterprise-grade features to keep things simple. Understanding these limitations helps you integrate Voctet smoothly into your workflow without unexpected surprises.
No SLA
We do not offer a Service Level Agreement (SLA) or uptime guarantees. Our infrastructure is robust, but during peak times, requests may queue, leading to slightly longer generation times for long texts. We process requests sequentially, so load can affect speed, but we do not promise specific availability percentages.
- No financial credits for downtime.
- Best-effort delivery during high traffic.
No Enterprise
Voctet is not an enterprise solution. We do not provide custom SLAs, dedicated support lines, or multi-seat team accounts. Each API key is tied to a single account, and there are no volume discounts or negotiated contracts. This keeps our pricing simple and accessible for individual developers and small teams.
- Single account per API key.
- No dedicated account manager.
- No custom contract terms.
No Custom Cloning
We do not offer custom voice cloning for customers. You cannot upload your own voice recordings to create a unique synthetic voice. Our eight presets are based on licensed recordings from LibriTTS-R and are fixed. If you need a specific voice, you must choose from our existing library of eight options.
- No voice upload feature.
- No custom voice creation.
- Fixed set of 8 voices.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. These diverse synthetic datasets will help you train your custom voice models effectively.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- LibriTTS-R corpus (OpenSLR 141)
- LibriTTS-R paper (arXiv 2305.18802)
- Python documentation
- FFmpeg documentation
- Node.js documentation
- IETF RFC 7845: Ogg Opus
- openai-node on GitHub
- curl documentation
- Creative Commons Attribution 4.0
- OpenAI API reference: create speech
- W3C WAI: making audio and video accessible
