Frequently asked questions
What is the best voice for news?
Q: Which voice is best for news? A: Atlas is designed for neutral, news-style delivery. Other options like Milo offer a friendly, conversational tone, while Orion suits deep, documentary narration. Each voice is based on licensed recordings from LibriVox. You can select any of the eight presets to match your editorial style. All voices support 23 languages, allowing you to maintain consistent audio quality across different regional broadcasts.
How do I handle long articles?
Split your text into parts of up to 4096 characters. Send each part as a separate request to the API. Then, join the resulting audio files. This approach ensures reliable synthesis and avoids hitting character limits. Since requests are processed sequentially, longer texts may take time to generate. Typical speed is roughly real-time, producing about 25 characters per second.
Can I mix voices in one audio file?
The API does not support mixing voices within a single audio file. Each request generates audio using one voice preset. To create a multi-voice broadcast, you must send separate requests for each segment. Then, concatenate the individual audio files using your preferred audio editing tool or script. This method gives you full control over the final broadcast structure.
What if the API is down?
We do not provide an SLA or uptime guarantee. Requests are processed one after another, so under load, your request may wait in a queue. If synthesis fails, the charge is returned in full. Input text is not stored, and generated audio is returned in the response. For critical deadlines, monitor the queue status and plan for potential delays.
Do I need a developer to use it?
A developer is helpful but not strictly required. The API uses a standard JSON body, similar to the OpenAI speech endpoint. You can use the official OpenAI SDKs by changing the base URL. For non-technical users, the /get-api-key/ page provides a ready curl example. You can also use plain HTTP requests to send text and receive audio files directly.
Can I use it in my CMS?
Yes, you can integrate it into your CMS via HTTP requests. The API accepts standard JSON bodies and returns audio files. You can write a simple script to fetch your API key, send the article text, and save the resulting audio. Since there is no official SDK or WordPress plugin, you will need to handle the integration logic within your CMS environment.
How do I split text correctly?
Split your text into chunks of up to 4096 characters. Ensure each chunk is a complete sentence or paragraph to avoid awkward pauses. Send each chunk as a separate request to the /v1/audio/speech endpoint. Then, concatenate the resulting audio files. This method ensures that each request stays within the limit and produces clear, high-quality audio segments.
What is the character limit?
Each request supports up to 4096 characters. Spaces and punctuation count toward this limit. If your article exceeds this length, split it into multiple requests. Each request is processed independently. Failed requests are free, so you can retry if needed. The total cost is $0.01 per 1,000 characters, regardless of the number of requests used.
Preparation: API Key and Balance
Access the /get-api-key/ page on win9online.com to create your account. You can register using Google or an email and password combination. The dashboard displays your active API key, current balance, and character usage. Note that accounts created with email and password do not include the $0.50 trial credit available to Google sign-ins.
Sign up and get your key
Once registered, your account page shows your single active API key starting with 'tts-'. This key authenticates all requests. You can issue a new key at any time, which immediately invalidates the previous one, ensuring secure access for your editorial team.
- Register via Google or email/password
- One active key per account
- New keys invalidate old ones instantly
Top up with USDT if needed
The service accepts USDT on the TRON network (TRC-20) for top-ups. There is a minimum deposit of $5, and funds are credited automatically within minutes of transfer. Payments are processed without invoices, subscriptions, or VAT, keeping billing transparent and simple.
- USDT (TRC-20) only
- Minimum top-up: $5
- No monthly plans or subscriptions
Check your balance and limits
Monitor your balance and character consumption directly on the dashboard. Your rate limit is 60 requests per minute per key, though this drops to 10 per minute before the first top-up. Failed requests do not consume balance or count toward limits.
- 60 requests/minute (standard)
- 10 requests/minute (pre-top-up)
- Failed requests are free
Choosing the Right Voice
Selecting the appropriate voice ensures your news audio matches the editorial tone. The API offers eight distinct presets, each designed for specific broadcasting styles. Choose based on whether you need neutral reporting, energetic promos, or calm instruction.
Atlas for neutral news delivery
Atlas provides a neutral voice ideal for standard reporting. It delivers clear, consistent audio suitable for daily briefings or factual updates where the content takes precedence over personality.
- Neutral tone for factual reporting
- Consistent delivery for daily briefings
- Clear articulation for news segments
Felix for energetic segments
Felix brings an energetic, promo-style delivery. Use this voice for weather updates, sports highlights, or promotional content that requires a brighter, more upbeat tone to engage listeners.
- Energetic and upbeat delivery
- Ideal for promos and weather updates
- Bright tone for engaging segments
Sage for instructional parts
Sage offers a calm, instructional voice. It is well-suited for educational news segments, explainers, or detailed analysis where a steady, reassuring pace helps listeners process complex information.
- Calm and instructional tone
- Steady pace for explainers
- Reassuring delivery for analysis

Sending the First Request
To generate audio, send a POST request to the https://api.win9online.com/v1/audio/speech endpoint. The payload must be valid JSON containing the text input, selected voice model, and desired output format.
Construct the JSON body
Include the input field with your script (max 4096 characters). Specify model as tts-1 or tts-1-hd. Set voice to a preset like atlas. Define response_format (e.g., mp3 or wav) and optional speed between 0.25 and 4.0.
Set the Authorization header
Include the header Authorization: Bearer tts-<your-key> in every request. This authenticates your account and deducts characters from your balance.
Send POST to /v1/audio/speech
Submit the request to the specified endpoint. On success, receive the audio file directly. If synthesis fails, the charge is returned in full.
Handling Long Articles
Long articles often exceed the 4,096 character limit per request. To handle this, split your script into smaller chunks and process each part individually. This ensures every segment is synthesized correctly without truncation.
Once all parts are generated, combine the resulting audio files into a single broadcast-ready file. This workflow allows editorial teams to produce full-length news segments or detailed articles efficiently.
Split text into 4096 char chunks
Break your input text into segments of up to 4,096 characters. Count spaces and punctuation as characters. This limit applies to each individual request sent to the API.
Ensure each chunk is a complete sentence or logical clause where possible. This helps maintain natural pacing when the audio segments are joined later.
- Split at sentence boundaries to avoid abrupt cuts.
- Count all characters, including spaces and punctuation.
- Store each chunk as a separate string variable in your code.
Loop through parts in code
Iterate through your list of text chunks. Send a POST request for each chunk to the /v1/audio/speech endpoint. Wait for the response before proceeding to the next chunk, or manage parallel requests carefully.
Keep track of the generated audio files or their URLs. Each request returns an audio file in the specified format. Store these temporarily on your server or in a cloud bucket.
- Use a loop to process each chunk sequentially.
- Ensure each request includes the correct voice and language parameters.
- Handle any API errors before moving to the next chunk.
Combine audio files after generation
After all chunks are synthesized, merge the audio files in the correct order. Use an audio library or command-line tool like FFmpeg to concatenate the files. This creates a single, continuous audio file.
Verify the combined file for seamless transitions. The API does not handle crossfades, so ensure your text splitting logic minimizes audible gaps between sentences.
- Use FFmpeg to concatenate MP3 or WAV files.
- Maintain the original order of the chunks.
- Test the final file for audio quality and continuity.
Setting the Language
The API can guess the language from your text, defaulting to English for Latin script. However, explicitly setting the language parameter ensures accuracy for non-English content or mixed scripts.
This step is critical for maintaining broadcast quality across the 23 supported languages.
Use ISO code for non-English
For non-English text, always pass the language field with the ISO 639-1 code. This prevents the API from misinterpreting the script, especially when using Latin characters for languages like Turkish.
Supported languages include Arabic, Chinese, Danish, Dutch, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish. Any voice can speak any of these languages.
- Use
enfor English - Use
zhfor Chinese - Use
jafor Japanese - Use
defor German - Use
frfor French
Auto-detect for Latin script
If you omit the language field, the system attempts to detect the language from the text. For Latin-script text that is not English, it defaults to English, which may result in incorrect pronunciation.
Explicitly setting the language code overrides this default behavior, ensuring the correct voice model is applied for your specific editorial content.
Ensure correct pronunciation
Accurate pronunciation relies on the correct language setting. If the language is guessed incorrectly, the voice may struggle with foreign names or terms.
For best results, verify the output audio for proper accent and rhythm, especially when handling proper nouns or technical terms in supported languages.
Error Handling and Retries
News audio workflows often face transient failures. Proper error handling ensures your broadcast pipeline remains reliable without manual intervention. This section covers rate limits, backoff strategies, and timeout management.
Check for 429 rate limit errors
The API enforces a limit of 60 requests per minute per key. New accounts start with 10 requests per minute. Exceeding this returns a 429 status code.
- Rate limit resets every minute.
- Each key has its own limit.
Implement exponential backoff
When a 429 error occurs, do not retry immediately. Wait for a short period, then retry with increasing delays. This prevents overwhelming the server during high demand.
- Start with a 1-second delay.
- Increase delay with each failed attempt.
Handle network timeouts
Long texts may take time to process. Set a reasonable timeout (e.g., 30–60 seconds) for your HTTP client. If the request times out, retry once to ensure the audio is generated.
Optimizing for Speed
To minimize latency, issue multiple synthesis requests in parallel. This approach leverages concurrent processing and reduces the total time required to generate a full news bulletin or long article.
Use parallel requests if possible
Since requests are processed one after another only under load, parallelizing independent requests ensures that idle time is minimized. Group related segments, such as individual news items, and send them simultaneously to maximize throughput.
- Send multiple POST requests concurrently for distinct text segments.
- Maintain separate request IDs to track individual responses.
Monitor queue wait times
During peak usage, requests may wait in a queue before synthesis begins. Monitor the time between submission and the start of generation to identify bottlenecks. If delays persist, consider staggering requests or reducing concurrency slightly.
Adjust speed parameter
Adjust the speed parameter (0.25–4.0) to balance generation speed and audio quality. Higher speeds may reduce processing time slightly but can affect clarity. Test different speeds to find the optimal balance for your editorial workflow.

Integrating with CMS
The NewsAudio API integrates seamlessly into editorial workflows, allowing you to convert scripts into broadcast-ready audio automatically. By connecting your CMS to the API, you can standardize voice output across all articles, ensuring consistent branding without manual intervention for each new piece of content.
Automate script-to-audio workflow
You can build a simple backend service that listens for article updates and sends the text to the API. Using the official OpenAI SDK with the adjusted base URL, your system can handle the JSON payload, including model selection and voice presets like 'atlas'.
- Use webhooks or polling to detect content changes.
- Map CMS fields directly to API input parameters.
- Handle text splitting for long articles automatically.
Trigger API on article publish
Configure your CMS to trigger the audio generation process immediately upon publication or scheduling. This ensures that audio versions are ready for distribution the moment the article goes live, reducing the time-to-air for digital news segments and keeping your content pipeline efficient.
- Set triggers on 'publish' or 'update' events.
- Include optional language ISO codes for non-English texts.
- Manage concurrent requests to stay within rate limits.
Store audio in media library
Once the API returns the audio file, store it in your CMS media library. This makes the audio easily accessible for embedding in articles or exporting for broadcast. You can associate the audio file with the specific article ID for organized retrieval and playback.
- Save files in supported formats like MP3 or WAV.
- Link audio assets to article metadata.
- Ensure proper file naming for easy reference.
Testing and Validation
Before integrating the API into your production news workflow, validate your setup using short, representative phrases. This initial test confirms that your authentication key is valid and that the API responds correctly to your specific voice and language parameters without wasting credits on long-form content.
Test with short phrases first
Start by sending a request with a few sentences. Use the OpenAI SDK or a simple curl command to verify the connection. If the response returns an audio file, your configuration is correct. Proceed to test different voices like Atlas for neutral delivery or Felix for energetic segments.
- Use a 50-character test string to verify connectivity.
- Confirm the correct voice model (tts-1 or tts-1-hd) is selected.
- Check that the audio format (mp3, wav, etc.) matches your CMS requirements.
Verify audio quality and format
Listen to the generated audio to ensure it meets broadcast standards. Check for clarity, proper pacing, and absence of artifacts. The API returns raw audio files; ensure your downstream tools can handle the specific format you requested, such as mp3 or wav.
- Verify the voice tone matches the editorial style.
- Check that the audio duration aligns with expectations.
- Ensure the file format is compatible with your player.
Check character count accuracy
Confirm that the character count sent matches the billing charge. Spaces and punctuation count as characters. Ensure your input text does not exceed the 4096-character limit per request. This prevents unexpected errors and helps you estimate costs for longer articles accurately.
- Count characters including spaces and punctuation.
- Verify the limit of 4096 characters per request.
- Ensure the charged amount matches the input length.
Monitoring Usage
Tracking character consumption ensures accurate billing, as costs are calculated per 1,000 characters including spaces. Each API request returns usage metadata, allowing you to reconcile charges against your generated audio files. Since longer texts are split into chunks, summing these values gives your total consumption.
Track characters consumed
Monitor your dashboard to see exactly how many characters have been processed. Remember that every space and punctuation mark counts toward the 4096-character limit per request. Consistent tracking helps predict costs, especially when publishing high-volume editorial content.
- Check balance after each batch of articles
- Verify character counts match source text length
Monitor rate limit usage
Your key allows 60 requests per minute. Exceeding this triggers a 429 error. Implement exponential backoff in your CMS automation to handle retries gracefully without flooding the API.
- Limit concurrent requests to stay under 60/min
- Use backoff strategies for retry logic
Alert on low balance
Set up notifications when your USDT balance drops below a threshold, such as $10. This prevents unexpected interruptions in your audio generation workflow. Top-ups via TRC-20 are processed automatically within minutes.
- Notify team when balance < $10
- Trigger auto-topup if configured
Common Mistakes to Avoid
Developing a robust news audio pipeline requires attention to API specifics. Even small oversights in request structure or error handling can disrupt automated publishing workflows. By avoiding common pitfalls, editorial teams ensure consistent delivery of broadcast-ready audio without unnecessary delays or billing surprises.
This guide highlights frequent errors encountered when integrating the NewsAudio API into existing CMS environments. Addressing these issues early prevents data loss and maintains the integrity of your audio assets.
Forgetting to set language
The API guesses the language from the script, defaulting to English for Latin text. If you publish content in French, German, or Spanish using Latin characters, the system may misinterpret the voice, resulting in an unnatural accent or mispronunciation.
Explicitly passing the language ISO code (e.g., "fr", "de") ensures the correct voice preset applies, guaranteeing accurate pronunciation for your specific audience.
- Always specify the
languagefield for non-English Latin text. - Default detection works well for English but is unreliable for other languages.
Exceeding character limit
Each request supports a maximum of 4,096 characters, including spaces and punctuation. Sending a long article in a single payload will cause the request to fail, potentially halting your automation script.
Split your text into chunks of 4,096 characters or fewer before sending. Process these chunks sequentially or in parallel, then concatenate the resulting audio files to create the final broadcast-ready file.
- Limit input text to 4,096 characters per request.
- Count spaces and punctuation toward the limit.
Not handling errors properly
Rate limits restrict you to 60 requests per minute per key. If you exceed this, the API returns a 429 error. Ignoring this response can lead to duplicate audio files or wasted credits if not managed correctly.
Implement exponential backoff in your code to retry failed requests. This ensures smooth operation even during high-volume publishing periods or when multiple team members are using the same API key.
- Implement retry logic with exponential backoff for 429 errors.
- Monitor rate limits to avoid service interruptions.

Support and Resources
The documentation provides the definitive reference for all API endpoints, parameters, and response formats. It includes ready-to-use cURL examples and SDK configurations to help you integrate the NewsAudio API into your editorial workflow quickly.
Check the documentation
Visit the /docs/ page for complete technical details. You will find specifications for the speech endpoint, voice presets, and supported audio formats like MP3 and WAV. The guides cover authentication, rate limits, and error codes to ensure smooth implementation.
- API endpoint specifications
- Voice and language options
- Authentication and headers
- Rate limit details
Email support for issues
For account-specific questions or billing inquiries, reach out via email. Unlike live chat, our email support allows for detailed troubleshooting of integration issues. Please include your API key and request details to help us resolve your problem efficiently.
Review FAQ for common problems
The FAQ section addresses frequent questions about character counting, USDT top-ups, and voice selection. It also clarifies that streaming and SSML are not supported, helping you set the right expectations for your broadcast-ready audio pipeline.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. The news publication team collaborates across large divisions to ensure accurate and timely reporting.
23 languages
Any voice speaks any of these languages.
