Frequently asked questions
What is the base_url for the API?
The API base URL is https://api.s9999.info/v1. You use this endpoint for all requests, including the standard speech generation and voice listing endpoints.
Which SDKs are compatible?
The official OpenAI SDKs work directly with our API. You only need to change the base_url to https://api.s9999.info/v1 and use your Speakable key. Other HTTP libraries like requests or axios also work perfectly since the request shape matches the standard OpenAI speech endpoint.
How do I handle long texts?
Split your input text into parts of up to 4096 characters. Send each part as a separate request to the speech endpoint. Then, join the resulting audio files in your application. This ensures you stay within the per-request character limit while generating full content.
What are the rate limits?
You are allowed 60 requests per minute per API key. If your account has not been topped up yet, the limit is lower at 10 requests per minute. Exceeding these limits will result in temporary throttling until the window resets.
Can I use the API with Python?
Yes, you can use Python with the official OpenAI SDK or any standard HTTP library. Simply configure the base_url to https://api.s9999.info/v1 and pass your Bearer token. The SDK handles the JSON body construction and audio response decoding automatically.
How do I get my API key?
Sign up on the /get-api-key/ page using Google or an email and password. Once registered, your dashboard displays your active API key. You can issue a new key at any time, which immediately invalidates the previous one for security.
What happens if I exceed limits?
If you exceed the 60 requests per minute limit, your requests will be throttled. Failed synthesis requests are free, so if a request fails due to limits or errors, the charge is returned in full. You do not lose credit for failed attempts.
Is the API synchronous?
Yes, the API is synchronous. You send a POST request and receive the audio file in the response. There is no WebSocket or streaming endpoint. Requests are processed one after another, so under load, you may experience slight delays in the queue.
Do I need to install a library?
You do not need a proprietary library. You can use the official OpenAI SDK or any standard HTTP client. If using the OpenAI SDK, just update the base_url. No specific Speakable library is required to integrate the API into your application.
How do I specify a language?
Include the optional 'language' field in your request body with the ISO code, such as 'en' or 'de'. If omitted, the system guesses the language from the script, defaulting to English for Latin text. Explicitly setting it ensures correct pronunciation for non-English languages.
Can I change the voice mid-request?
No, each request generates audio for a single voice. To change voices, you must send separate requests for each segment of text. You can choose from eight preset voices like aria, nova, or felix, but each audio file corresponds to one voice only.
What is the max input length?
The maximum input length is 4096 characters per request. This limit applies to the text string you send in the 'input' field. If your content is longer, you must split it into chunks and process them individually before joining the audio files.
Are responses streamed?
No, responses are not streamed. The API returns the complete audio file in the HTTP response. There is no support for real-time partial audio or WebSocket connections. You receive the full file once synthesis is complete, which may take a few seconds for short phrases.
How do I test the API?
Use the ready curl example provided on the /get-api-key/ dashboard page. You can also use the OpenAI SDK by setting the base_url. Your free trial includes $0.50 credit (about 50,000 characters), allowing you to generate audio to verify functionality.
First request
curl
curl https://api.s9999.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.s9999.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.s9999.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
Speakable uses standard Bearer token authentication. Include your API key in the request header as Authorization: Bearer tts-.... This token grants access to all endpoints, including speech synthesis and voice metadata.
Bearer token usage
Pass the key in the Authorization header for every request. The format is identical to OpenAI’s authentication method, allowing you to swap the base_url and use existing SDKs seamlessly.
Key management
Each account holds one active key. You can generate a new key from your dashboard at any time. This provides a simple way to rotate credentials without managing multiple tokens.
Revoking old keys
When you issue a new key, the previous one stops working immediately. There is no delay; the old token is effectively revoked the moment the new one is created, ensuring tight security control.
Endpoints
The API exposes three core endpoints for interaction. The primary speech synthesis endpoint accepts text and configuration to return audio. Two discovery endpoints provide metadata about available voices and model configurations, allowing developers to build dynamic UIs that adapt to available assets.
POST /v1/audio/speech
Send a JSON body with the text to convert. The endpoint accepts the input string, up to 4096 characters, along with the chosen model, voice, and optional response_format. Authentication requires a Bearer token in the header. On success, it returns the audio file directly; on failure, the request is free.
- Endpoint: POST https://api.s9999.info/v1/audio/speech
- Body: JSON with model, input, voice, format, and speed
- Auth: Authorization: Bearer <key>
GET /v1/voices
Retrieve the list of eight available voice presets along with their supported languages and formats. This endpoint helps you determine which voice suits your content type, whether it is the warm narration of aria or the deep documentary tone of orion.
- Endpoint: GET https://api.s9999.info/v1/voices
- Returns: Voice IDs, names, and supported languages
GET /v1/models
Fetch the available model configurations. Currently, this includes tts-1 and tts-1-hd, which share the same engine and pricing but differ in quality settings. Use this to verify model availability before synthesis requests.
- Endpoint: GET https://api.s9999.info/v1/models
- Returns: Model IDs and details
Request Parameters
Configure the synthesis engine, input constraints, and voice parameters to generate high-quality audio files. The API accepts standard text inputs, allowing you to select the model, voice, and output format in a single request.
Model selection (tts-1, tts-1-hd)
Choose between "tts-1" and "tts-1-hd". Both use the same engine and carry identical pricing.
Input text limits
Each request accepts up to 4096 characters for the input field. Spaces and punctuation count toward this limit. For longer texts, split the content into chunks of this size and join the resulting audio files.
Voice and language fields
Select one of eight preset voices (e.g., aria, nova, atlas). While the system guesses the language from the script, explicitly setting the language ISO code is recommended for non-English Latin text to ensure accuracy.
Response Formats
The API returns audio files in formats specified by the response_format parameter. Choose MP3 for broad compatibility, WAV for lossless quality, or OPUS and FLAC for efficient compression.
MP3 output
MP3 is one of the available formats, offering a balance of quality and file size. It is ideal for web playback and general distribution where universal player support is required.
WAV output
WAV provides uncompressed, lossless audio. Use this format when you need the highest fidelity for editing or archival purposes, accepting larger file sizes.
FLAC and OPUS options
FLAC offers lossless compression for smaller file sizes without quality loss. OPUS provides efficient lossy compression, suitable for low-bandwidth environments while maintaining clear speech intelligibility.
Rate Limits
The API enforces a strict limit of 60 requests per minute per key.
60 requests per minute
Each API key is allowed a maximum of 60 requests within any given minute. This rate applies uniformly across all endpoints, including speech generation and voice listing. Developers should monitor their usage to avoid temporary throttling during peak integration periods.
10 requests before top-up
New accounts start with a lower limit of 10 requests per minute until the first top-up is processed. This allows developers to test the integration quickly before committing to a USDT (TRC-20) deposit. Once the minimum $5 is credited, the limit increases to 60 requests per minute immediately.
Queue behavior under load
Requests are processed sequentially. If the queue fills up, additional requests will wait their turn rather than failing immediately. For long texts, which take roughly 25 characters per second to generate, this queuing ensures audio files are produced in the correct order.

SDK Integration
Integrating Speakable is straightforward because the API follows the same request structure as the OpenAI speech endpoint. You can use existing OpenAI SDKs by simply changing the base_url to https://api.s9999.info/v1 and providing your site-specific key. This allows indie developers to leverage familiar code patterns without learning a new library.
For those preferring lightweight solutions, plain HTTP requests work identically. Send a JSON body with model, input, voice, and response_format to the /v1/audio/speech endpoint. The API returns the audio file directly, ensuring no vendor lock-in and minimal dependency overhead for your startup’s audio features.
OpenAI SDK base_url
Configure your OpenAI SDK client to point at https://api.s9999.info/v1. Pass your tts-... bearer token in the authorization header. The request shape remains identical to the standard speech endpoint, so your existing integration logic requires only this single configuration change.
This approach lets you use the official SDK for authentication and retries while accessing Speakable’s voices and pricing. It is ideal for projects already using OpenAI tools that need consistent, high-quality TTS without switching libraries.
- Set
base_urltohttps://api.s9999.info/v1 - Use standard
Authorization: Bearer tts-...header - Maintain existing OpenAI SDK compatibility
Plain HTTP requests
If you prefer not to use the OpenAI SDK, send a standard POST request to /v1/audio/speech. Include your API key in the headers and the text payload in the JSON body. The response contains the generated audio file in your chosen format.
This method gives you full control over headers, timeouts, and retry logic. It is well-suited for microservices or environments where minimizing third-party dependencies is a priority.
- POST to
/v1/audio/speech - Include
Authorization: Bearerheader - Return audio in
mp3,wav,opus, orflac
No custom SDK required
We do not provide an official SDK. You are free to use any HTTP client or the OpenAI SDK as described. This ensures you are not tied to a proprietary library that might change or become deprecated.
By adhering to standard REST principles, your integration remains portable. You can easily switch HTTP libraries or languages without rewriting your core audio generation logic.
Error Handling
Handling failures correctly is critical for maintaining user experience during TTS generation. This section details how to interpret API responses and manage retry scenarios without exceeding rate limits or incurring unnecessary charges.
Common error codes
The API returns standard HTTP status codes. A 401 Unauthorized indicates an invalid or revoked Bearer token. A 429 Too Many Requests signals that you have exceeded the 60 requests per minute limit. Other 4xx errors typically reflect invalid parameters, such as an unsupported voice or model.
- 400: Bad Request (invalid JSON or parameters)
- 401: Unauthorized (invalid or revoked key)
- 429: Too Many Requests (rate limit exceeded)
Retry logic
Implement exponential backoff for transient errors like 503 Service Unavailable. Do not retry immediately, as requests are processed sequentially and queues may be full. Wait at least one second between attempts to avoid hammering the API and triggering stricter rate limits.
Failed request refunds
If synthesis fails completely, the charge is returned in full. You are only billed for successfully generated audio files. Failed requests do not consume your character quota, ensuring you only pay for usable output.
Audio Processing
Handling texts longer than 4096 characters requires splitting the input into valid segments. Each segment triggers a separate synthesis request, ensuring that the API constraints are respected without truncating meaningful content. This approach allows developers to generate audio for articles, tutorials, or lengthy documentation seamlessly.
The system processes these requests sequentially. While individual segments generate quickly, the total time scales with the text length. Under heavy load, requests may queue, but the deterministic nature of the split ensures consistent output structure for downstream processing.
Long text splitting
Split text into chunks of up to 4096 characters. Each chunk generates a distinct audio file. This limit applies to the input field per request, so longer narratives must be divided programmatically before sending to the endpoint.
- Maximum 4096 characters per request
- Spaces and punctuation count toward the limit
- Splitting is handled by the client or wrapper
Character limit enforcement
Ensure each submitted string does not exceed 4096 characters. Requests exceeding this limit will fail, incurring no charge. Accurate character counting includes all spaces and punctuation, which affects the billing calculation of $0.01 per 1,000 characters.
Audio joining strategy
Concatenate the returned audio files in the order of the original text segments. Since the API returns complete audio files for each segment, a simple concatenation of the binary data or a playlist file creates the final continuous output.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. This sequential processing guarantees that each audio file is generated correctly without race conditions.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- IETF RFC 7845: Ogg Opus
- TRON developer hub: TRC-20
- openai-node on GitHub
- Creative Commons Attribution 4.0
- FFmpeg documentation
- LibriVox public-domain audiobooks
- LibriTTS-R paper (arXiv 2305.18802)
- ISO 639 language codes
- W3C WAI: making audio and video accessible
- OpenAI API reference: create speech
- LibriTTS-R corpus (OpenSLR 141)
