PDF Text to Speech API for Audiobook Production

By ScribeRead editorial team · Updated 2026-11-18

ScribeRead provides a character-based Text-to-Speech API designed for self-publishers and no-code developers. This page details how to convert large PDF documents into high-quality audio files efficiently, handling the volume and complexity of long-form content without manual formatting.

You will find technical specifications for PDF parsing, supported languages, and voice profiles. The guide covers integration with standard OpenAI SDKs, pricing structures, and data privacy terms, ensuring a reliable workflow for audiobook production.

PDF Text to Speech API for Audiobook Production

Frequently asked questions

Can I use the API to read large PDF files directly?

No. The API accepts plain text strings up to 4096 characters, not PDF files. You must extract the text from your PDF using your preferred tool or library before sending it to the endpoint. The service converts the provided text string into audio, requiring you to handle the document parsing step separately.

How does the API handle long texts exceeding 4096 characters?

You must split the text into parts of up to 4096 characters each. Send each part as a separate request and join the resulting audio files. The API does not automatically split long texts or merge the audio segments; this workflow is managed by your application or script.

Is the audio generated suitable for commercial audiobooks?

Yes. The generated audio files are returned in the response and are not stored by the service. The audio carries an inaudible watermark marking it as synthetic speech. Input text is used only for production and not for training, making the output suitable for commercial distribution in audiobooks.

Which voices are best for fiction versus non-fiction?

Luna offers a soft storytelling tone, while Aria provides a warm, clear narrator style. Sage delivers a calm, instructional voice, and Atlas provides a neutral, news-style delivery. Felix is energetic, while Orion is deep and documentary. All eight voices support 23 languages and can be adapted to various genres based on your specific content needs.

Do I need to install any software to use the API?

No. The API is accessed via standard HTTP requests. You can use the official OpenAI SDKs by changing the base URL, or send plain HTTP requests. No mobile SDK, WordPress plugin, or browser extension is offered. Any environment capable of making HTTP POST requests can use the service.

How are characters counted in the billing?

Billing is based on the total number of characters in the input text. Spaces and punctuation count as characters. The rate is $0.01 per 1,000 characters. Failed requests are free, and the charge is returned in full if synthesis fails. Each request processes up to 4096 characters.

What happens if a request fails during processing?

If synthesis fails, the charge is returned in full. You are not billed for characters processed in a failed request. You can retry the request or split the text differently. The API ensures that you only pay for successfully generated audio files.

Can I change the speed of the narration?

Yes. You can adjust the speed parameter between 0.25 and 4.0. This allows you to slow down or speed up the narration to suit your content. The speed adjustment applies to the generated audio file directly within the request parameters.

Are the voices based on human recordings or AI generated?

The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. Each generated file carries an inaudible watermark marking it as synthetic speech. The voices are presets derived from these human recordings, not custom AI models trained on new data.

Is there a limit to how many books I can produce?

No. There is no limit to the number of books you can produce. You pay per character used. The only constraint is the rate limit of 60 requests per minute per key. You can generate as much content as your budget and API key allow.

Do you support languages other than English?

Yes. The API supports 23 languages including Arabic, Chinese, French, German, Japanese, Spanish, and more. Any voice can speak any of these languages. If the language is not set, it is guessed from the script, but you should set it explicitly for non-English Latin-script text.

How long does it take to generate audio for a short chapter?

Typical speed is roughly real time, about 25 characters of text per second of generation. Short phrases take a few seconds. Requests are processed one after another, so under load, a request may wait in a queue. Longer texts generate at approximately real-time speed.

Can I use the API with Zapier or Make.com?

Yes. Since the API uses standard HTTP requests with a JSON body, it can be integrated with automation tools like Zapier or Make.com. You can configure these tools to send POST requests to the endpoint with your API key and text data to generate audio.

What is the cost for processing a typical 50,000 character book?

The cost is $0.50. The rate is $0.01 per 1,000 characters. For 50,000 characters, you multiply 50 by 0.01. This price applies to successful requests. Failed requests are free. The free trial includes $0.50 credit.

First request

curl

curl https://api.playjp77.com/v1/audio/speech \
  -H "Authorization: Bearer $TTS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"tts-1","voice":"aria","input":"Hello from the API.","response_format":"mp3"}' \
  --output hello.mp3

Automate Audiobook Creation from PDF Sources

Sound wave illustration for: Automate Audiobook Creation from PDF Sources

ScribeRead provides a RESTful API endpoint compatible with standard OpenAI SDKs, allowing no-code platforms to convert PDF text directly into audio. By changing the base URL to https://api.playjp77.com/v1, developers can integrate speech synthesis into automated publishing pipelines without custom infrastructure.

The system handles the volume of long-form content by accepting up to 4096 characters per request. Clients can split large documents programmatically and join the resulting audio files, ensuring efficient processing of full-length audiobooks without manual formatting overhead.

Direct API integration for no-code platforms

Direct API integration for no-code platforms

The API uses a standard POST request structure, enabling immediate integration with tools like Zapier or Make. Authentication is handled via a Bearer token, ensuring secure access to speech synthesis endpoints for automated workflows.

  • Compatible with official OpenAI SDKs
  • Standard JSON request body
  • Simple Bearer token authentication

Handling large character counts efficiently

Each request supports a maximum input of 4096 characters. For longer texts, split the content into chunks and combine the output files. This approach ensures consistent processing even for extensive manuscripts.

  • 4096 character limit per request
  • Character counting includes spaces
  • Joinable audio output

Consistent voice across long documents

Select one of eight voice presets to maintain a uniform narrative tone throughout your audiobook. The engine applies the same voice profile to each chunk, ensuring seamless listening experiences across the entire document.

  • Single voice per batch
  • Uniform tone for fiction and non-fiction
  • No voice drift between chunks

Supported Languages and Voice Profiles

Sound wave illustration for: Supported Languages and Voice Profiles

ScribeRead supports 23 languages, including English, Spanish, French, German, Japanese, and Arabic. Any voice can speak any of these languages, ensuring consistent audio quality across diverse audiobook projects. This flexibility allows self-publishers to produce multilingual editions from a single source text without managing separate voice assets.

23 languages with native-sounding voices

We support 23 languages: Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish. If the language field is omitted, the system 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 accuracy.

8 distinct voice presets for varied genres

Select from 8 licensed presets based on LibriTTS-R and LibriVox recordings. Female voices include Aria (warm narrator), Nova (bright), Luna (soft storytelling), and Sage (calm instructional). Male options are Felix (energetic), Milo (friendly), Atlas (neutral news), and Orion (deep documentary). These presets cover fiction, non-fiction, and educational content.

  • 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

Automatic language detection with fallback

The API automatically detects language from the input script if the language field is not set. This reduces configuration overhead for single-language projects. However, for mixed-language texts or non-English Latin scripts, manually specifying the language code is recommended to prevent misidentification and ensure correct pronunciation.

Technical Specifications for PDF Parsing

Sound wave illustration for: Technical Specifications for PDF Parsing

The API accepts text inputs up to 4,000 characters per request. For lengthy PDF documents, clients must split the content into chunks and concatenate the resulting audio files. This method ensures consistent processing and avoids token limits while maintaining narrative continuity across chapters.

4096 character input limit per request

Each API call processes a maximum of 4,096 characters, including spaces and punctuation. Larger documents require client-side splitting into sequential requests. The service returns individual audio files that must be joined to form the complete audiobook.

  • Limit: 4,096 characters per request
  • Supports standard Latin and non-Latin scripts
  • Concatenation required for full document reconstruction

Multiple audio formats: MP3, WAV, FLAC

Select output formats via the response_format parameter. Supported options include MP3, WAV, OPUS, FLAC, and PCM. FLAC and WAV provide lossless quality for archival purposes, while MP3 optimizes file size for distribution.

  • MP3 for web distribution
  • WAV and FLAC for lossless archival
  • PCM for raw data processing

Speed control for narration pacing

Adjust the speed parameter between 0.25 and 4.0 to match desired pacing. Standard narration operates at a rate of approximately 25 characters per second. Slower speeds enhance clarity for complex texts, while faster speeds reduce total generation time for large batches.

  • Range: 0.25x to 4.0x
  • Standard speed approx. 25 chars/sec
  • Adjustable for listener preference

Data Privacy and Usage Rights

Sound wave illustration for: Data Privacy and Usage Rights

ScribeRead ensures that input text is used solely for the purpose of generating the requested audio file and is not utilized to train our underlying models. This isolation guarantees that your manuscript content remains distinct from the training data pool, preserving the uniqueness of your self-published works.

Input text not used for training models

Your source text, whether derived from a PDF or direct input, is processed exclusively for synthesis. It is not retained for model improvement, ensuring that your narrative data remains private and separate from our general dataset.

No persistent storage of generated audio

The synthesized audio files are returned directly in the API response and are not stored on our servers after delivery. This minimizes data retention risks, allowing you to manage storage and distribution of your audiobook files independently.

Transparent licensing for commercial use

Each file contains an inaudible watermark identifying it as synthetic speech, which is standard for AI-generated content but does not restrict your commercial rights.

Pricing Structure for Self-Publishers

Sound wave illustration for: Pricing Structure for Self-Publishers

ScribeRead charges $0.01 per 1,000 characters, billed per million-character blocks. Failed requests incur no charge, ensuring you only pay for successful audio generation. This transparent, usage-based model eliminates fixed costs for irregular publishing schedules.

Pay per million characters usage

Each request processes up to 4,096 characters. Pricing is strictly character-based, including spaces and punctuation. There are no hidden fees for format selection or speed adjustments.

No monthly subscription fees

Accounts operate on a pay-as-you-go basis with no minimum spend or recurring billing cycles. Balances do not expire, allowing you to top up only when needed via USDT (TRC-20).

Unlike traditional SaaS platforms, we do not require monthly subscriptions or seat licenses. Your balance remains available indefinitely until used.

Free trial credit for new accounts

New accounts created with Google sign-in receive $0.50 in trial credit, valid for 7 days. This provides approximately 50,000 characters of test audio to verify voice quality and integration.

Accounts created with email and password start without trial credit. Top-ups start at $5, credited automatically within minutes of transfer.

Pricing Structure for Self-Publishers

API Reliability and Rate Limits

Sound wave illustration for: API Reliability and Rate Limits

ScribeRead provides a stable foundation for audiobook production by enforcing a standard rate limit of 60 requests per minute per API key. This capacity supports consistent batch processing for self-publishers without requiring complex infrastructure scaling. Requests are processed one after another, so under load a request may wait in a queue.

60 requests per minute standard limit

The API enforces a strict limit of 60 requests per minute per active key. This limit applies uniformly to all requests, regardless of text length or voice selected. New accounts start with 10 requests per minute until the first top-up is processed.

  • Standard limit: 60 requests/minute/key
  • Initial limit: 10 requests/minute for new accounts
  • Limits reset every 60 seconds

Queue processing for high-volume batches

Requests exceeding the per-minute capacity are placed in a processing queue. Since requests are processed one after another, high-volume batches may experience slight delays during peak usage. This sequential processing ensures that each audio file is generated with consistent quality before moving to the next.

Real-time status codes for errors

The API returns standard HTTP status codes to indicate request success or failure. Failed synthesis requests are free, with the charge returned in full. You can monitor request status via the response headers and body, allowing for immediate retry logic in your no-code workflows.

  • HTTP 200: Success
  • HTTP 429: Rate limit exceeded
  • HTTP 500: Internal server error

Voice Selection for Narrative Styles

Sound wave illustration for: Voice Selection for Narrative Styles

ScribeRead’s eight voice presets are engineered to match specific narrative tones, allowing self-publishers to select the optimal voice for their content genre. Each voice is derived from licensed recordings, ensuring consistent quality across long-form audiobooks.

Warm narrators for fiction genres

Aria provides a warm, clear narration style ideal for fantasy and romance novels. Luna offers a soft, storytelling approach that enhances emotional depth in character-driven narratives.

Clear instructional voices for non-fiction

Sage delivers a calm, instructional tone suitable for educational materials and how-to guides. Nova adds a bright, upbeat energy that keeps listeners engaged in business or self-help content.

Neutral tones for news and reports

Atlas provides a neutral, professional voice perfect for news summaries and factual reports. Orion offers a deep, documentary style that adds gravitas to historical or biographical texts. Felix brings energetic clarity for promotional content.

  • Aria: Warm, clear narrator
  • Luna: Soft, storytelling
  • Sage: Calm, instructional
  • Nova: Bright, upbeat
  • Atlas: Neutral, news
  • Orion: Deep, documentary
  • Felix: Energetic, promo
  • Milo: Friendly, conversational

Integration with No-Code Workflows

Sound wave illustration for: Integration with No-Code Workflows

ScribeRead leverages standard OpenAI SDK compatibility, allowing no-code platforms to integrate TTS by simply adjusting the base URL. This allows connection to tools like Zapier or Make via standard HTTP requests.

Standard OpenAI SDK compatibility

Use the official OpenAI SDKs with the base URL https://api.playjp77.com/v1. This approach ensures your existing no-code logic remains intact while accessing our TTS capabilities.

  • Compatible with standard OpenAI SDKs
  • Simple base URL configuration
  • No custom library required

Simple cURL examples for testing

Verify your integration immediately using standard cURL commands. The request structure mirrors the OpenAI speech endpoint, allowing for quick validation of voices, formats, and speed settings.

  • Direct JSON body structure
  • Standard authentication headers
  • Quick validation of endpoints

Easy webhook or polling integration

For asynchronous workflows, integrate via polling or webhooks. The API returns clear status codes, ensuring your automation script can reliably track request progress and handle failures without manual intervention.

  • Clear status codes
  • Asynchronous processing support
  • Reliable error handling

Quality Assurance and Audio Output

Sound wave illustration for: Quality Assurance and Audio Output

ScribeRead generates audio with an inaudible watermark marking it as synthetic speech. This ensures clear provenance for audiobook production without affecting listener experience. Input text is used only to produce the audio and is not used for training.

Inaudible watermarking for synthetic speech

Every generated file carries an inaudible watermark that marks it as synthetic speech. This provides transparent provenance for commercial audiobooks without impacting audio quality or listener experience.

High-definition audio options available

Select 'tts-1-hd' for higher fidelity or 'tts-1' for standard efficiency. Both models use the same engine and price ($10 per 1 million characters). Output formats include mp3, wav, opus, flac, and pcm to suit your distribution pipeline.

Consistent output across batch jobs

Split large PDFs into parts of up to 4096 characters and join the audio files. Requests are processed sequentially, ensuring consistent voice tone across long documents. Failed requests are free, protecting your character balance during bulk generation.

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. His style keeps long-form non-fiction engaging without sacrificing clarity.

Avatar of the voice aria
ariafemale
Avatar of the voice nova
novafemale
Avatar of the voice luna
lunafemale
Avatar of the voice sage
sagefemale
Avatar of the voice felix
felixmale
Avatar of the voice milo
milomale
Avatar of the voice atlas
atlasmale
Avatar of the voice orion
orionmale

23 languages

Any voice speaks any of these languages.

Sources and further reading

  1. Wikipedia
  2. LibriTTS-R paper (arXiv 2305.18802)
  3. IETF RFC 7845: Ogg Opus
  4. LibriTTS-R corpus (OpenSLR 141)
  5. Python documentation
  6. ISO 639 language codes
  7. OpenAI API reference: create speech
  8. MDN: Web Speech API
  9. openai-node on GitHub
  10. openai-python on GitHub
  11. Creative Commons Attribution 4.0
  12. LibriVox public-domain audiobooks

More on this site