Frequently asked questions
Which voice is best for dyslexia?
We do not designate a single voice as best for dyslexia. All eight presets—Aria, Nova, Luna, Sage, Felix, Milo, Atlas, and Orion—are clear and distinct. You may select any voice based on user preference. The API returns standard audio formats.
How do I split long texts for reading?
The API accepts up to 4096 characters per request. For longer texts, you must split the input into chunks of this size. After receiving the audio files for each chunk, you must join them sequentially in your application. This ensures the entire text is synthesized without exceeding the character limit per individual API call.
Can I adjust the reading speed?
Yes, you can adjust the speed between 0.25 and 4.0. Set the speed parameter in your request body. This allows users to control playback rate directly. The API processes requests sequentially, so speed adjustments do not affect processing time significantly. This feature supports diverse reading preferences common in accessibility applications.
Does the API detect language automatically?
The API guesses the language from the script if not specified. Latin script defaults to English. For other Latin-script languages, you must set the language parameter explicitly. If you omit it for non-English text, the output may be incorrect. Supported languages include Arabic, Chinese, and German, among others listed in our documentation.
What is the best audio format for web?
Q: What is the best audio format for web? A: We support mp3, wav, opus, flac, and pcm. For web applications, mp3 or opus are typically efficient. Mp3 offers broad compatibility. Opus provides better compression. Choose based on your client requirements. All formats are generated instantly upon request.
How do I integrate with React?
Use the official OpenAI SDK or plain HTTP. Change the base URL to https://api.6zzz.info/v1. Send a POST request to /v1/audio/speech with your text and voice. Use your Bearer token for authentication. React can handle the response as a blob or array buffer. No specific React SDK is provided.
Is the API WCAG compliant?
The API itself does not guarantee WCAG compliance. It generates audio files based on your input. Compliance depends on how you implement the audio player and controls in your user interface. We provide the audio data. You must ensure your application meets accessibility standards by providing proper ARIA labels and controls.
How much does it cost per chapter?
Pricing is $10 per million characters. Spaces and punctuation count. A 50,000-character chapter costs $0.50. Failed requests are free. There are no monthly fees. You pay only for the characters processed. Top-ups are via USDT on TRON. Balance does not expire. Costs scale linearly with text length.
What happens if the API is slow?
Requests are processed sequentially. Under load, requests may wait in a queue. Typical speed is roughly real-time. Short phrases take a few seconds. If the API is slow, your request is likely queued. There is no SLA or uptime guarantee. You may experience delays during high traffic periods.
Can I store the generated audio?
The API returns the audio in the response but does not store it. You are responsible for storing the generated files if needed. Input text is used only for synthesis and not for training. You can store the audio files on your own servers or cloud storage for future use by your users.
Do I need to handle errors manually?
Yes, you must handle errors in your integration. If synthesis fails, the charge is returned in full. You should check the HTTP status code and response body. Implement retry logic if necessary. Failed requests do not incur costs. Proper error handling ensures a smooth user experience when audio generation encounters issues.
How do I test voice quality?
Use the /v1/voices endpoint to list available voices. Send a test text to the /v1/audio/speech endpoint. Listen to the output audio file. You can adjust speed and format. Test with different voices like Aria, Nova, or Felix. Compare outputs to determine which voice suits your accessibility needs best.
Is student data used for training?
No, input text is used only to produce the audio. It is not used for training models. Generated audio is returned in the response and not stored by us. Request metadata is kept for billing purposes only. Your data remains private and is not utilized for improving our synthesis engine.
How do I scale for many users?
Each key has a limit of 60 requests per minute. You can issue a new key if needed, but the old one stops immediately. Scale by managing keys or distributing load. Requests are queued under load. Monitor your usage via the dashboard. There is no enterprise tier or SLA for scaling guarantees.
Choosing the Right Voice for Reading
Selecting the appropriate voice preset directly impacts reading comprehension for users with dyslexia. Clarity and tonal consistency reduce cognitive load by providing predictable audio patterns, allowing users to focus on decoding text rather than interpreting speech nuances.
Our API offers eight distinct presets, with Aria and Sage serving as primary tools for educational content. Aria provides a warm, clear narration, while Sage delivers a calm, instructional tone.
Clarity and tone importance
Consistent vocal characteristics help users establish reading rhythm. A stable tone minimizes distraction, enabling better retention of complex information. Developers should test voice presets with target users to ensure the selected tone aligns with their specific accessibility needs.
- Reduce cognitive fatigue during long reading sessions
- Maintain consistent pacing for better comprehension
- Support diverse learning styles through tone variation
Aria for narrative content
Aria is designed as a warm, clear narrator. The voice maintains a steady cadence, making it suitable for extended text inputs.
- Warm, clear narrator style
- Ideal for narrative and story content
- Supports English and 22 other languages
Sage for instructional text
Sage offers a calm, instructional tone optimized for educational material. Its neutral delivery reduces auditory distraction, making it effective for textbooks, instructions, and technical reading. Users can adjust speed between 0.25x and 4.0x to match their processing speed.
- Calm, instructional delivery
- Optimized for educational and technical text
- Supports speed adjustment from 0.25x to 4.0x
Managing Input Character Limits
Each API request accepts a maximum of 4,096 characters, including spaces and punctuation. This constraint requires developers to parse input text before synthesis. Long articles or chapters must be divided into smaller chunks to fit within this limit.
4096 character constraint
The character limit applies per request, not per file. Text exceeding this length must be segmented. Each segment generates a separate audio file. This ensures predictable processing times and avoids timeouts.
Splitting long texts
Split text at natural boundaries, such as paragraph breaks or sentence ends. Preserve the sequence of segments to maintain narrative flow. Use the input field for each chunk individually. Ensure no single request exceeds 4,096 characters.
- Break text at logical sentence or paragraph boundaries.
- Ensure each request stays within the 4,096-character limit.
- Maintain the original order of segments for sequential playback.
Joining audio segments
Receive separate audio files for each text segment. Concatenate these files in the correct order on the client side. Use standard audio libraries to merge MP3 or WAV streams. This creates a continuous listening experience for the user.
- Concatenate audio files in the original text sequence.
- Use client-side libraries for seamless audio merging.
- Ensure no gaps or overlaps between joined segments.
Speed Adjustment for Comprehension
Adjusting playback speed is critical for dyslexic readers, who often benefit from slower, clearer audio to process text without cognitive overload. The API allows precise control over synthesis speed, enabling developers to tailor the listening experience to individual user needs. This flexibility supports varied reading paces and improves comprehension for users with different processing speeds.
The API supports speed values ranging from 0.25x to 4.0x, allowing users to slow down narration for better clarity or speed up for review. This range accommodates both detailed study and rapid skimming, ensuring the audio matches the user’s current focus and comfort level.
Slower speeds for clarity
For users who struggle with rapid speech, setting the speed between 0.25x and 0.75x provides clearer enunciation and reduces auditory fatigue.
By reducing the rate, developers ensure that each word is distinctly audible, supporting better retention and understanding of the material.
- Ideal for detailed study or complex texts.
- Reduces auditory fatigue for sensitive listeners.
- Improves phoneme clarity and distinction.
Range from 0.25x to 4.0x
The API accepts speed values from 0.25 to 4.0, offering extensive customization. Users can start at a comfortable pace and gradually increase speed as their reading confidence grows. This adaptability makes the tool suitable for both novice and experienced readers seeking efficient information consumption.
Extreme speeds like 4.0x are useful for quick reviews of familiar content, while moderate speeds around 1.0x–1.5x suit standard reading. The linear adjustment ensures predictable changes in audio duration, helping users estimate time requirements accurately.
- Supports speeds from 0.25x to 4.0x.
- Allows gradual pace adjustment.
- Suitable for both review and detailed reading.
User-controlled playback
Integrate speed controls directly into your application’s UI, allowing users to adjust playback in real time. This user-driven customization empowers individuals to find their optimal listening speed, enhancing engagement and reducing frustration. Simple sliders or buttons can map to the API’s speed parameter.
Ensure that speed settings persist across sessions if needed, providing a consistent experience. By giving users control, you accommodate diverse learning styles and cognitive preferences, making the reading assistance more inclusive and effective for a wider range of dyslexic readers.
- Real-time UI controls for speed adjustment.
- Empowers user customization.
- Supports diverse learning styles.
Language Detection Accuracy
Language handling in AudioLex Access relies on script detection by default, which assumes Latin scripts are English. For non-English Latin text, you must explicitly set the language code to ensure accurate synthesis.
Script-based guessing
The API infers language from the input script. Latin characters default to English. Cyrillic or Greek scripts trigger their respective languages. This automatic detection simplifies basic implementations for single-language content.
- Latin script defaults to English
- Cyrillic text triggers Russian detection
- Greek text triggers Greek detection
Explicit language codes
For mixed scripts or ambiguous text, pass the ISO language code in the request body. This overrides automatic detection. Supported languages include Arabic, Chinese, Danish, and 20 others.
- Pass 'language' field in JSON body
- Overrides default script-based inference
- Ensures correct pronunciation for specific voices
Handling mixed languages
Each request supports one language code. For mixed-language texts, split the input by language segments. Send separate requests for each segment and join the resulting audio files.
- One language code per request
- Split text by language boundaries
- Join audio segments client-side
Audio Format Selection
The API supports MP3, WAV, OPUS, FLAC, and PCM outputs. Selecting the correct format ensures optimal playback quality and resource usage within your reading application.
MP3 for mobile efficiency
MP3 provides a balanced compression ratio, making it ideal for mobile devices where bandwidth and storage are limited. This format allows smooth playback of synthesized speech without significant data consumption.
WAV for high fidelity
WAV delivers uncompressed audio, preserving the full quality of the synthesized voice. Use this format when audio fidelity is prioritized over file size, such as in high-quality accessibility tools.
OPUS for bandwidth saving
OPUS offers superior compression at low bitrates, significantly reducing data transfer costs. This format is suitable for users with limited internet connectivity or when processing large volumes of text.
Integration with React or Angular
Integrating the AudioLex Access API into React or Angular applications requires handling the asynchronous speech synthesis response efficiently. Developers must manage the API base URL and authentication headers within their service layer to ensure secure token transmission for each request.
Fetch API usage
Use the native Fetch API to POST text to the endpoint. The request body must include the model, voice, and input text. The response provides an audio file that can be converted into a Blob object for immediate playback.
Audio element handling
Create an HTML5 audio element dynamically or reuse a persistent instance. Assign the Blob URL to the src attribute to initiate playback. This approach ensures compatibility with standard web audio contexts and allows for standard playback controls.
State management for playback
Manage playback state (playing, paused, stopped) within your framework's state management system. Update UI components to reflect the current status, ensuring that user interactions like clicking a 'stop' button correctly halt the audio stream without memory leaks.
Accessibility Compliance
Our API delivers clean, synthetic audio that integrates seamlessly with standard accessibility patterns, ensuring compliance with WCAG 2.1 guidelines for non-text content. By providing reliable text-to-speech conversion, we enable developers to create reading experiences that support users with dyslexia and visual impairments without requiring complex infrastructure.
WCAG guidelines support
The generated audio files are standard formats (MP3, WAV, OPUS) that browsers recognize natively. The consistent audio output allows for predictable behavior across assistive technologies.
- Standard audio formats ensure broad compatibility.
- Consistent output supports predictable assistive technology behavior.
- No proprietary codecs required for playback.
Keyboard navigation for audio
Since we do not provide a proprietary SDK, you retain full control over the DOM structure. This allows precise implementation of keyboard navigation for play, pause, and skip controls. Developers can bind standard ARIA attributes to the HTML5 audio element, ensuring focus management aligns with user expectations for reading applications.
- Full DOM control enables custom ARIA attributes.
- Standard HTML5 audio element supports native keyboard events.
- Flexible structure aids focus management implementation.
Screen reader compatibility
Because the audio is returned as a direct file, it can be easily annotated with ARIA labels. This ensures that visually impaired users receive accurate auditory feedback without interference from background noise or complex UI elements.
- Clear enunciation aids human and machine listening.
- Direct file return allows easy ARIA annotation.
- Minimal UI interference improves reading flow.

Cost Management for Startups
AudioLex Access charges $0.01 per 1,000 characters, ensuring predictable expenses for dyslexia-focused apps. Failed requests incur no cost, protecting your budget during development. The API processes text sequentially, so long documents may queue under load, but billing remains strictly tied to successful synthesis.
Predictable per-character cost
Cost scales linearly with usage. Each character, including spaces and punctuation, counts toward the total. This model eliminates surprise fees, allowing precise forecasting for student reader applications based on estimated text volume.
- $0.01 per 1,000 characters
- Failed requests are free
- Spatial punctuation counts as characters
No subscription overhead
There are no monthly plans, seats, or minimum spends. You only pay for what you generate. Accounts start with a $0.50 trial for 7 days. Top-ups via USTRC20 are processed within minutes.
- No monthly plans
- No seats or minimum spend
- Balance does not expire
Monitoring usage metrics
The dashboard displays balance, trial status, and characters used. Request metadata is retained for billing but input audio is not stored. This allows you to track consumption accurately without storing large audio files.
- Real-time balance display
- Character count tracking
- Billing metadata retention
Error Handling in Reading Apps
Robust error handling ensures uninterrupted reading experiences for users with dyslexia. When network connectivity drops, your application should cache the last successfully synthesized segment and resume playback once the connection is restored, preventing cognitive disengagement.
Network failure recovery
Implement exponential backoff for failed requests. If a synthesis fails due to transient network issues, retry up to three times with increasing delays. This approach minimizes user frustration while maintaining audio continuity.
API rate limit warnings
Monitor the 60 requests per minute limit closely. If you approach the threshold, display a non-intrusive notification indicating potential delays. This allows users to pause reading or adjust speed settings without abrupt interruption.
- Check
Retry-Afterheader in 429 responses - Display estimated wait time if queue is long
User feedback on failures
Provide clear, accessible error messages when synthesis fails. Use ARIA live regions to announce failures to screen readers. Offer a simple retry button that re-submits the specific text segment without requiring full page reloads.
- Announce failure via
aria-liveregion - Offer one-click retry for failed segments
Testing Voice Performance
Validating synthesis performance ensures reading accessibility meets user expectations. Measuring latency helps developers optimize the user experience, particularly for real-time reading assistance where delays can disrupt comprehension. Accurate performance data allows teams to adjust speed and format settings effectively.
Latency measurement
Monitor the time between sending the request and receiving the audio file. Short phrases typically take a few seconds, while longer texts process at roughly real-time speed (25 characters per second). Requests are processed sequentially, so high concurrency may introduce queue delays.
- Track start-to-finish duration for various text lengths.
- Monitor queue wait times during peak usage.
- Compare performance across different response formats.
Audio quality assessment
Verify that the synthetic watermark is present but unobtrusive. Test voice selection for specific content types, such as Aria for narratives or Sage for instructional text.
- Assess voice clarity for dyslexic readers.
- Check audio fidelity in MP3 and WAV formats.
- Verify watermark presence without affecting playback.
User preference testing
Conduct tests with target users to determine optimal speed settings between 0.25x and 4.0x. Gather feedback on voice tone and language detection accuracy. Use this data to refine default settings for accessibility features in your application.
- Collect user feedback on voice selection.
- Test speed adjustments for comprehension.
- Validate language detection accuracy.
Data Privacy for Student Data
For applications serving students, understanding how input data is handled is critical. Our API is designed with privacy as a core constraint, ensuring that the text you send for synthesis is used exclusively for the immediate production of the audio file.
No training on input text
Unlike many consumer-grade services, we do not use your input text to train our models or improve voice quality. The text sent in the request body remains isolated to that specific synthesis job.
No audio storage
The generated audio file is returned directly in the HTTP response and is not persisted on our servers. Once the client receives the audio data, our infrastructure does not retain a copy, reducing the risk of data exposure in case of a breach.
Metadata retention policy
We retain only the necessary request metadata—specifically the timestamp, character count, and request status—for billing and operational purposes. This limited dataset is sufficient to verify usage and process charges without storing the actual content or audio files.
Scaling for Multiple Users
AudioLex Access supports multiple concurrent users through a standardized rate limit of 60 requests per minute per API key. This limit applies regardless of the number of active users sharing the key, requiring careful monitoring to prevent synthesis delays during peak reading sessions.
Rate limit considerations
Each API key is capped at 60 requests per minute. New accounts start with 10 requests per minute until the first top-up. Exceeding this limit may cause requests to queue, increasing latency for end-users. Monitor usage metrics via the dashboard to adjust key allocation based on concurrent user load.
- 60 requests per minute per key
- 10 requests per minute for new accounts
- Automatic queuing during peak load
Concurrency management
Requests are processed sequentially within the rate limit. For applications with high concurrency, consider distributing requests across multiple API keys or implementing client-side queuing to smooth traffic spikes. This ensures consistent playback experience without abrupt pauses.
Load balancing strategies
Since the API processes requests one after another, spread user requests over time or use multiple keys to maintain throughput. Avoid simultaneous bursts that exceed the per-minute limit. Implement exponential backoff in your client to handle temporary rate limit errors gracefully.
- Distribute requests across multiple keys
- Implement client-side queuing
- Use exponential backoff for retries
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. WAV provides uncompressed, lossless audio, preserving every detail of the synthesized voice.
23 languages
Any voice speaks any of these languages.
