Frequently asked questions
How do I authenticate?
Include the header Authorization: Bearer tts-... with your unique API key in every request. You obtain this key on the /get-api-key/ page. Each account holds one active key; issuing a new key immediately invalidates the previous one, ensuring secure access to the TTS endpoint.
What is the base URL?
The API base URL is https://api.bet44.info/v1. The primary endpoint for generating speech is POST https://api.bet44.info/v1/audio/speech. You can also retrieve available voices and models via GET /v1/voices and GET /v1/models respectively using this base path.
Which models are available?
Two models are available: tts-1 and tts-1-hd. Both utilize the same engine and pricing structure. You may select either model in your request body based on your preference, as there is no functional difference in cost or output quality between them.
How do I change the voice?
Specify the desired voice in the voice field of your JSON request body. Choose from eight presets: aria, nova, luna, sage, felix, milo, atlas, or orion. Each voice can speak any of the 23 supported languages, allowing flexible selection based on your content needs.
What formats are supported?
You can request audio in mp3, wav, opus, flac, or pcm formats. Set the response_format field in your request to your preferred type. The API returns the generated audio file directly in the response, ready for immediate download or playback by your application.
Can I use the Python SDK?
Yes, the official OpenAI Python SDK works seamlessly. Configure the base_url to https://api.bet44.info/v1 and provide your API key. This allows you to use standard SDK methods for speech generation without writing custom HTTP clients, leveraging existing code structures for easy integration.
What is the rate limit?
The limit is 60 requests per minute per API key. Before your first top-up, the limit is 10 requests per minute. Requests are processed sequentially, so high volume may cause queuing. Monitor your usage on the dashboard to stay within these constraints effectively.
How do I handle errors?
Failed requests are free, and the charge is returned in full. If synthesis fails, the API returns an error status with details. Check the response status code and message to diagnose issues like invalid parameters or network problems, then retry as needed without cost.
Does it support async requests?
The API does not support streaming or WebSocket connections for real-time partial audio. Requests are processed sequentially. You must wait for the full audio file to be generated and returned in the response before proceeding. Use standard HTTP async libraries if you need non-blocking calls.
How long does generation take?
Generation is roughly real-time, averaging 25 characters of text per second of audio. Short phrases take a few seconds. Under load, requests may queue. Plan your integration accordingly, as processing time depends on text length and current server queue status.
Can I stream the audio?
No, streaming is not supported. The API returns the complete audio file in the response body after generation finishes. You cannot receive partial chunks of audio. Download the full file before playback. This ensures deterministic output but requires waiting for the entire synthesis process.
Is there a limit on request size?
Each request supports up to 4096 characters of input text. For longer texts, split them into parts of this size and join the resulting audio files. Spaces and punctuation count as characters. This ensures consistent processing times and predictable billing for your API calls.
How do I reset my API key?
Issue a new key from the /get-api-key/ page. The old key stops working immediately when the new one is active. This provides a simple reset mechanism without downtime. You can generate a new key at any time to invalidate the previous one securely.
What is the response format?
The response contains the generated audio file in the requested format (mp3, wav, opus, flac, or pcm). The audio is returned directly in the HTTP response body. No JSON wrapper is used for the audio data itself, allowing direct saving to file or memory buffers.
First request
curl
curl https://api.bet44.info/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.mp3Python (openai)
from openai import OpenAI
client = OpenAI(base_url="https://api.bet44.info/v1", api_key="tts-...")
with client.audio.speech.with_streaming_response.create(
model="tts-1", voice="nova", input="Hello from Python.", response_format="mp3"
) as response:
response.stream_to_file("hello.mp3")Node.js (openai)
import OpenAI from "openai";
import fs from "node:fs";
const client = new OpenAI({ baseURL: "https://api.bet44.info/v1", apiKey: process.env.TTS_API_KEY });
const res = await client.audio.speech.create({ model: "tts-1", voice: "milo", input: "Hello from Node.js." });
fs.writeFileSync("hello.mp3", Buffer.from(await res.arrayBuffer()));| voice | |
|---|---|
aria | female |
nova | female |
luna | female |
sage | female |
felix | male |
milo | male |
atlas | male |
orion | male |
| language | |
|---|---|
ar | العربية |
zh | 中文 |
da | Dansk |
nl | Nederlands |
en | English |
fi | Suomi |
fr | Français |
de | Deutsch |
el | Ελληνικά |
he | עברית |
hi | हिन्दी |
it | Italiano |
ja | 日本語 |
ko | 한국어 |
ms | Bahasa Melayu |
no | Norsk |
pl | Polski |
pt | Português |
ru | Русский |
es | Español |
sw | Kiswahili |
sv | Svenska |
tr | Türkçe |
Authentication
Authentication relies on a single active Bearer token per account, issued via the /get-api-key/ dashboard. The token format is strictly tts-<unique-string>. Pass this value in the Authorization header as Bearer tts-<unique-string> for every request to https://api.bet44.info/v1. Only one key remains valid; generating a new one immediately invalidates the previous key.
Bearer Token Format
The authentication header must follow the standard Bearer scheme. Ensure no extra whitespace or quotes surround the token string. The prefix tts- distinguishes our keys from generic API tokens.
API Key Generation
Keys are created upon sign-up via Google or email. The dashboard displays your active key alongside your current balance and usage statistics. You may generate a new key at any time if needed.
Key Rotation Policy
Rotation is instantaneous. Upon issuing a new key, the old one stops working immediately. This eliminates overlap risks but requires immediate client-side updates to avoid 401 errors.
Speech Generation Endpoint
The POST https://api.bet44.info/v1/audio/speech endpoint accepts a JSON payload and returns synthesized audio. Requests are processed sequentially; under load, a request may queue until resources are available. Typical generation speed is roughly real-time, though short phrases may take a few seconds to complete.
Send your text via the input field, selecting a model and voice from the available presets. The API validates all parameters before synthesis begins, ensuring deterministic behavior. Failed requests incur no charge, and the audio file is returned directly in the response body for immediate use.
POST Method Specification
Use the HTTP POST method to send your request to the speech endpoint. The request body must be a valid JSON object containing the text to be synthesized and the desired output parameters. The API processes one request at a time, so concurrent requests will be queued.
JSON Body Structure
The request body is a standard JSON object. It includes the text input, the model identifier, the voice preset, the desired audio format, and an optional speed multiplier. All fields are strictly typed to ensure consistent processing across different client implementations.
- "model": "tts-1" or "tts-1-hd"
- "input": String (max 4096 characters)
- "voice": One of the 8 preset voice IDs
- "response_format": mp3, wav, opus, flac, or pcm
- "speed": Float between 0.25 and 4.0
- "language": Optional ISO code (e.g., "en", "es")
Required Fields
Every request must include the model, input, and voice fields. The input text cannot exceed 4096 characters per request. If the text is longer, split it into parts and join the resulting audio files. The voice field must match one of the eight available presets.
Model Selection
The API offers two model identifiers, tts-1 and tts-1-hd, which utilize the same underlying synthesis engine. Both models support the full range of eight voice presets and accept identical input parameters, including text up to 4096 characters and speed adjustments between 0.25 and 4.0.
tts-1 Engine
The tts-1 model provides standard synthesis with a generation speed of approximately 25 characters per second. It is suitable for most applications where real-time performance on long texts is required without specific audio fidelity constraints.
tts-1-hd Engine
Selecting tts-1-hd triggers the high-definition engine variant. While the processing pipeline remains identical, this model may offer enhanced audio quality depending on the specific voice preset used, without altering the request structure.
Price Parity
There is no cost difference between the two models. Both are billed at $10 per 1 million characters, regardless of whether tts-1 or tts-1-hd is specified in the request body.
- Same engine infrastructure
- Identical character counting
- No per-model surcharge
Voice Parameters
Select from eight deterministic voice presets designed for specific audio contexts. Each voice is generated from licensed recordings, ensuring consistent tonal quality across all languages. The system allows precise matching of output character to use case, from narrated content to conversational interfaces.
All voices support the full set of 23 languages. Language detection occurs automatically for Latin-script text, but explicit language codes are required for non-English Latin inputs to ensure accurate pronunciation. Each voice maintains its distinct tonal characteristics regardless of the language spoken.
8 Preset Voices
The API provides four female and four male voice options. Female options include Aria for clear narration, Nova for upbeat content, Luna for storytelling, and Sage for instructional material. Male options cover Felix for energetic promos, Milo for conversational tones, Atlas for neutral news delivery, and Orion for deep documentary styles.
- 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
Voice ID Strings
Reference each voice using its lowercase identifier string in the voice parameter. The API maps these strings to the specific audio engine profiles. Use GET /v1/voices to retrieve the complete list of available voice IDs and their associated metadata.
Gender and Tone Variations
Voices are categorized by gender and primary tonal attribute. Select based on the desired listener experience rather than strict gender. All voices can speak any supported language without changing their core timbre, allowing flexible casting across different projects.
Audio Format Options
The API returns audio in five supported formats: mp3, wav, opus, flac, and pcm. The response_format parameter determines the output container and codec. MP3 is the default and offers balanced compression, while WAV provides uncompressed lossless quality. Opus delivers high efficiency for low-bandwidth transmission, and FLAC ensures bit-perfect archiving without data loss.
MP3 Compression
MP3 is the default output format. It provides a standard balance between file size and audio fidelity, suitable for most web and mobile applications. The generated file is playable in standard media players.
WAV Lossless
WAV output delivers uncompressed audio. This format is ideal for post-production workflows where further editing is required, as it preserves the full audio data without compression artifacts. The resulting files are larger but retain maximum quality for downstream processing.
Opus and FLAC Support
Opus provides efficient compression suitable for low-bandwidth transmission, while FLAC offers lossless compression for archival purposes. Both formats ensure high audio integrity. Choose Opus for efficient storage or FLAC if you require a compressed but bit-perfect copy of the generated speech.

OpenAI SDK Integration
The API mirrors the OpenAI speech endpoint structure, allowing developers to reuse existing integration code with minimal adjustments. By overriding the base URL to https://api.bet44.info/v1, applications can authenticate using standard Bearer tokens and send requests via familiar SDK methods.
Base URL Override
Point your SDK client configuration to https://api.bet44.info/v1. This single change directs all subsequent requests to our infrastructure while maintaining the standard request shape expected by OpenAI-compatible clients.
Standard SDK Methods
Use the standard audio speech creation method. Pass the model (tts-1 or tts-1-hd), input text, voice ID, and desired response format. The SDK handles the JSON serialization and returns the audio binary data.
Compatibility Notes
The official OpenAI SDKs work correctly with this endpoint. We do not offer a proprietary SDK, ensuring you can stick with the libraries you already know. Note that streaming and SSML parameters are not supported in this implementation.
- Uses standard OpenAI SDK clients
- No proprietary client required
- Base URL: https://api.bet44.info/v1
Rate Limits
The API enforces a standard rate limit of 60 requests per minute for authenticated keys. This threshold applies uniformly across all endpoints, ensuring consistent throughput for both short phrase synthesis and bulk audio generation workflows. Developers should monitor their request volume to avoid temporary throttling during peak usage periods.
Requests exceeding the per-minute threshold are processed sequentially. If multiple synthesis tasks are queued, they are handled one after another, which may introduce latency during high-load scenarios. This deterministic queuing ensures that no requests are dropped, though response times may increase proportionally to the queue depth.
60 Requests Per Minute
Each API key is allowed a maximum of 60 requests per minute. This limit is calculated on a rolling window basis. Developers using the OpenAI SDK or direct HTTP requests should implement exponential backoff if they receive a 429 status code to prevent overwhelming the service.
The limit applies per key, not per account. If you rotate your key, the new key begins with a fresh counter (10 requests per minute). This allows for distributed load balancing by switching keys when necessary, provided the total volume does not exceed the aggregate capacity of your infrastructure.
- Standard limit: 60 requests/minute per key
- Rolling window calculation
- Applies to all endpoints (speech, voices, models)
Pre-Funding Limit
Accounts created with an email and password, or those that have not yet completed a top-up, are restricted to 10 requests per minute. This lower threshold prevents resource exhaustion during the initial testing phase before the user has committed funds to the account.
Once a top-up of at least $5 in USDT (TRC-20) is confirmed, the rate limit automatically increases to the standard 60 requests per minute. No manual support ticket is required for this transition. The change is reflected immediately in the API response headers.
- Pre-funding limit: 10 requests/minute
- Standard limit: 60 requests/minute
- Automatic upgrade after confirmed top-up
Queue Behavior
Requests are processed in the order they are received. If the server is under heavy load, additional requests may wait in a queue. This queue is internal and transparent to the client. The client receives a response only after the synthesis is complete or fails.
Longer texts may take more time to process, potentially occupying a queue slot for a longer duration. Short phrases are generated quickly, but even these may experience slight delays if the queue is full. Developers should account for this variability in their application logic, especially when generating audio in real-time or near-real-time applications.
- Sequential processing order
- Transparent internal queue
- No request dropping
Error Handling
The API returns standard HTTP status codes to indicate request success or failure. Understanding these codes allows for immediate identification of authentication issues, malformed requests, or server-side processing errors during speech synthesis.
Common HTTP Errors
A 401 Unauthorized error indicates an invalid or expired API key. Ensure the Authorization header uses the correct Bearer token format. A 429 Too Many Requests error occurs when exceeding the 60 requests per minute limit. Check your current usage via the dashboard to adjust request frequency.
- 400: Bad Request (invalid JSON or missing fields)
- 401: Unauthorized (invalid API key)
- 429: Too Many Requests (rate limit exceeded)
- 500: Internal Server Error (synthesis failure)
Retry Logic
Implement exponential backoff for transient errors like 500 or 503 status codes. For 429 errors, wait until the rate limit window resets before retrying. Failed synthesis requests are free, so retrying does not incur additional charges if the error persists.
Character Limit Errors
Requests exceeding 4096 characters return a 400 error. Split long texts into smaller segments before sending. Join the resulting audio files client-side to maintain continuity. This ensures predictable processing times and avoids queue delays for large inputs.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. Each voice is a deterministic preset, meaning the output remains consistent across identical inputs.
23 languages
Any voice speaks any of these languages.
