Text-to-Speech API for Voice Assistants and Chatbots

By Auralux editorial team · Updated 2026-11-19

Auralux provides a straightforward text-to-speech API designed for development teams building voice-enabled applications. We handle the heavy lifting of audio generation so you can focus on conversation design and user experience. Send simple text, receive an audio file, and integrate seamlessly with your existing stack using our OpenAI-compatible endpoint.

This service is ideal for teams creating voice assistants and chatbots who need predictable costs and reliable performance. You get a simple JSON endpoint, 23 supported languages, and eight distinct voice presets. With character-based pricing and no monthly subscriptions, you pay only for what you use, making it easy to scale your voice features without complex billing surprises.

Text-to-Speech API for Voice Assistants and Chatbots

Frequently asked questions

What is a TTS API?

A TTS API is a cloud service where you send text and receive an audio file. You POST JSON to the endpoint, specifying the model, voice, and format. It simplifies audio generation so your team can focus on conversation design without managing complex synthesis infrastructure or hardware.

How does text-to-speech work for chatbots?

Send the user’s text to our endpoint with your chosen voice and language. We return an audio file (mp3, wav, etc.) that your bot plays. For long responses, split the text into 4096-character chunks, generate each part, and join the audio files seamlessly for the user.

Which voices are available for voice assistants?

We offer eight presets: Aria, Nova, Luna, and Sage (female); Felix, Milo, Atlas, and Orion (male). Each has a distinct tone, from warm narration to neutral news delivery. All voices support 23 languages, allowing you to maintain consistent character voices across different linguistic markets easily.

Do I need to manage audio files?

No. The API returns the audio file directly in the response. We do not store the generated audio. You only keep request metadata for billing. This means you handle storage and delivery to your users, ensuring you have full control over your voice-enabled application’s assets.

Can I use the OpenAI SDK with this API?

Yes. Our endpoint matches the OpenAI speech request shape. Simply change the SDK’s base_url to https://api.66mmr.info/v1 and use your Auralux key. This allows you to leverage existing OpenAI SDK integrations without rewriting your codebase, making migration or integration straightforward for development teams.

How are characters counted?

We charge per character, including spaces and punctuation. One million characters cost $10. Text is limited to 4096 characters per request. If you exceed this, split your input. This transparent metric helps you predict costs accurately based on the volume of text your bot processes.

What happens if synthesis fails?

If synthesis fails, the charge is returned in full. You are not billed for failed requests. This ensures you only pay for successful audio generation. You can retry the request or adjust parameters like speed or voice without financial penalty if the initial attempt encounters an error.

Is the audio streaming in real-time?

No, we do not support streaming or WebSocket audio delivery. You receive the complete audio file after generation. For long texts, we process requests sequentially, so latency depends on queue load. Typical speed is roughly real-time, but short phrases may take a few seconds to generate.

Do you support SSML?

No, we do not support SSML. You control pacing and emphasis by adjusting the speed parameter (0.25–4.0) or splitting text into separate requests. This simplifies the integration for teams who prefer standard text inputs over complex markup languages, keeping the API straightforward and easy to implement.

Can I clone my own voice?

No, we do not offer custom voice cloning. You choose from our eight licensed presets based on LibriTTS-R and LibriVox recordings. If you need a specific voice, you must select the closest preset. We do not upload or train models on your personal voice data.

Which languages are supported?

We support 23 languages, including English, Spanish, German, French, Japanese, and Hindi. Any voice can speak any supported language. If you use Latin script for non-English text, set the language parameter explicitly, as it defaults to English. Languages like Vietnamese and Thai are not supported.

Is there a free trial?

Yes, signing in with Google gives you $0.50 credit (about 50,000 characters) valid for 7 days. Email/password accounts start without trial credit. This allows you to test the API’s quality and integration speed before committing to a balance top-up via USDT (TRC-20) for ongoing usage.

How do I pay with USDT?

Top up using USDT on the TRON network (TRC-20). The minimum deposit is $5, and credits are added automatically within minutes. We do not accept cards, PayPal, or subscriptions. Your balance does not expire, and you can issue a new API key anytime if needed.

What is the rate limit?

You are limited to 60 requests per minute per key. Before your first top-up, the limit is 10 requests per minute. Requests are processed sequentially, so high load may cause queuing. This ensures fair usage for all teams building voice assistants without requiring complex infrastructure scaling on your end.

First request

curl

curl https://api.66mmr.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.mp3

How it works

Sound wave illustration for: How it works

Auralux transforms your text into natural-sounding speech in three simple steps. Paste your script, select a voice, and receive an audio file ready for playback. We handle the complex audio generation so your team can focus on conversation design and user experience.

Send text, receive audio

Send text, receive audio

Input plain text up to 4,096 characters. Our engine synthesizes the audio and returns the file directly in the response. You don’t need to manage storage or processing pipelines; the audio is generated on demand and delivered instantly.

  • Up to 4,096 characters per request
  • Audio returned in the response body
  • No storage management required

Simple JSON endpoint

Interact with our API via a standard POST request to /v1/audio/speech. Send your text, voice selection, and format preference in a clean JSON body. The response contains the audio file, ready for immediate playback in your voice assistant or chatbot.

  • POST https://api.66mmr.info/v1/audio/speech
  • JSON request body
  • Standard HTTP response format

OpenAI SDK compatible

If you are already using the OpenAI SDK, you can switch to Auralux by simply changing the base_url to https://api.66mmr.info/v1. The request shape is identical, making integration fast for existing development teams without rewriting their codebase.

  • Use official OpenAI SDKs
  • Change base_url only
  • Identical request shape

Why choose us for voice apps

Sound wave illustration for: Why choose us for voice apps

Predictable character pricing means you only pay for what you use, with no surprise overages. Our rate is $0.01 per 1,000 characters, making budgeting straightforward for scaling voice assistants.

We skip monthly subscriptions entirely. Whether you generate a few minutes or thousands of hours, your cost scales linearly. This model fits teams of any size without locking you into rigid tiers.

Predictable character pricing

You pay $0.01 per 1,000 characters, including spaces and punctuation. Failed requests are free, so you never pay for errors. This transparency helps you forecast costs accurately as your voice app grows.

Balances are topped up via USDT (TRC-20) and never expire. You control your spending precisely, with no minimums or hidden fees attached to your usage.

  • $0.01 per 1,000 characters
  • Failed requests are free
  • No hidden fees

No monthly subscriptions

We do not lock you into monthly plans or seat limits. Your account is flexible, allowing you to scale up or down without administrative overhead.

This approach suits development teams who prefer operational expenditure over fixed commitments, ensuring you only spend when you actually generate audio.

Fast synthesis speed

Long texts process at roughly real-time speed, generating about 25 characters per second. Short phrases may take a few seconds due to queue handling under load.

Requests are processed sequentially, ensuring consistent quality. You can split longer texts into 4,096-character chunks to optimize throughput and reduce latency for your users.

Supported languages

Sound wave illustration for: Supported languages

Our API supports 23 languages, covering most major markets for voice assistants and chatbots. Any voice can speak any of these languages, giving you flexibility in conversation design without needing multiple voice models.

You can explicitly set the language using an ISO code for precision, or let the system guess based on the text script. Latin script text defaults to English if unspecified.

23 languages supported

We support Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish.

This range covers the vast majority of use cases for global voice applications, ensuring your users hear natural-sounding speech in their preferred language.

Auto-detection with fallback

If you do not provide a language parameter, the API guesses the language from the script. For non-English Latin-script text, we recommend setting the language field explicitly to avoid misidentification.

This simplifies your integration code while maintaining accuracy for standard inputs.

No Vietnamese or Thai yet

We do not currently support Vietnamese, Indonesian, Thai, Ukrainian, Romanian, Czech, or Bengali. If your application targets these regions, you may need a fallback solution.

  • Vietnamese
  • Indonesian
  • Thai
  • Ukrainian
  • Romanian
  • Czech
  • Bengali

Voice selection

Sound wave illustration for: Voice selection

Choose from eight distinct voice presets designed to fit various conversation styles, from warm narrators to energetic promo voices. Each voice is optimized for clarity and natural flow in voice assistant interactions.

Select between female and male options to match your brand’s personality. Our voices are built on licensed recordings, ensuring high-quality synthetic speech without the need for custom voice cloning.

8 distinct voice presets

We offer eight carefully crafted voice presets to suit different application needs. Whether you need a calm instructional guide or a deep documentary narrator, there is a voice for every use case.

The selection includes aria, nova, luna, sage, felix, milo, atlas, and orion. Each preset provides a consistent tone that works well across voice assistants and chatbots.

  • 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

Female and male options

Our voices span a range of genders and tones, giving you the flexibility to choose the best fit for your user experience. You can mix and match voices across different languages without restriction.

Every voice can speak all 23 supported languages, allowing you to maintain consistent character voices even in multilingual applications.

Licensed recordings

Our voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. This foundation ensures high-quality audio synthesis while respecting licensing terms.

Each generated file carries an inaudible watermark that marks it as synthetic speech.

Audio formats

Sound wave illustration for: Audio formats

We support multiple audio formats to fit your specific infrastructure needs. You can choose between MP3 for efficient web delivery, WAV for uncompressed quality, or OPUS and FLAC for specialized use cases.

MP3 for web players

MP3 is the standard for web-based voice assistants. It offers a good balance between file size and audio quality, making it ideal for playback in chatbots or mobile apps.

  • Small file size reduces bandwidth costs
  • Wide compatibility with all web players
  • Ideal for real-time voice responses

WAV for high quality

Choose WAV when you need uncompressed audio for processing or archiving. It preserves the full fidelity of the generated voice without any lossy compression artifacts.

  • Uncompressed PCM audio
  • Best for post-processing
  • No quality loss during playback

OPUS and FLAC options

OPUS provides high-quality audio at very low bitrates, perfect for low-bandwidth environments. FLAC offers lossless compression, reducing file size without sacrificing any audio detail.

  • OPUS: Efficient for mobile networks
  • FLAC: Lossless compression
  • Both support high fidelity
Audio formats

Developer experience

Sound wave illustration for: Developer experience

The API uses a standard REST endpoint compatible with existing OpenAI SDKs. You simply send a POST request with your text and receive an audio file. This straightforward design means your team can integrate voice capabilities without learning a proprietary protocol or managing complex state.

Standard REST endpoint

Send text to POST https://api.66mmr.info/v1/audio/speech and receive audio instantly. The JSON body mirrors the OpenAI speech endpoint, so you only need to change the base_url in your existing SDK configuration. This reduces integration time and lets you keep your current development workflows intact.

  • POST /v1/audio/speech endpoint
  • JSON request body
  • OpenAI SDK compatible

Bearer token auth

Authentication is handled via a simple Authorization: Bearer header. You get one active key per account, which you include in every request. This keeps your integration clean and secure without requiring OAuth flows or session management for each synthesis call.

Instant key rotation

If your key is compromised or you need a fresh start, you can issue a new key at any time. The old key stops working immediately, so you don’t have to wait for propagation or manage stale credentials in your deployed voice applications.

Pricing transparency

Sound wave illustration for: Pricing transparency

Our pricing model is straightforward: you pay only for the characters you process. There are no monthly subscriptions, no tiered plans, and no hidden fees. This simplicity helps large development teams forecast costs accurately without worrying about sudden bill spikes from unexpected usage spikes.

Every character counts toward your total, including spaces and punctuation. Failed requests are free, so you never pay for errors. This ensures that your budget reflects actual successful audio generation, making it easy to scale your voice assistant without complex accounting.

$0.01 per 1000 characters

The base rate is $10 per one million characters, which equals $0.01 per 1,000 characters. This rate applies to both the 'tts-1' and 'tts-1-hd' models, as they use the same engine and pricing structure. You can calculate your costs directly based on your application's expected text volume.

Spaces and punctuation count as characters, so ensure your input text includes all visible symbols. For example, a 100-character sentence costs $0.001. This predictable unit cost allows you to integrate billing logic directly into your user experience or internal dashboards.

  • Base rate: $0.01 per 1,000 characters
  • Applies to both tts-1 and tts-1-hd models
  • Spaces and punctuation count toward the total

No hidden fees

There are no monthly subscriptions, seat fees, or minimum spend requirements. You can use the API at any volume, from a few requests a day to thousands per minute, without changing your cost structure. The balance does not expire, so you can top up when convenient.

We do not charge for API calls themselves, only for the characters processed. Request metadata is kept for billing, but input text is not used for training, so you do not pay for data usage rights. This transparency extends to our payment methods: USDT on TRON only, with no VAT invoices or credit card processing fees.

  • No monthly subscriptions or seat fees
  • No minimum spend requirements
  • Balance does not expire

Failed requests are free

If a synthesis request fails, the charge is returned in full. You only pay for successfully generated audio files. This protects you from paying for errors caused by network issues or invalid inputs. Failed requests are free, ensuring your balance reflects only usable output.

This policy works in tandem with our rate limits. Even if you hit the 60 requests per minute limit and requests queue, you are not charged until the audio is successfully generated. If the service is temporarily unavailable, you are not billed for the wait.

  • Full refund on failed requests
  • Charge only applies to successful synthesis
  • Protects against network or input errors

Technical limits

Sound wave illustration for: Technical limits

Each API call accepts a maximum of 4096 characters, so you will need to split longer scripts into manageable chunks before sending them to the endpoint. Spaces and punctuation count toward this limit, so plan your text segmentation accordingly.

Your API key allows up to 60 requests per minute. If you exceed this rate, new requests will be queued and processed sequentially, which may introduce latency during high-volume generation tasks.

4096 character limit

Keep individual payloads under 4096 characters to avoid truncation. For longer dialogues, your application should handle the splitting logic and reassemble the resulting audio files.

  • Maximum input length: 4096 characters per request
  • Spaces and punctuation are counted toward the limit

60 requests per minute

The standard rate limit is 60 requests per minute per key. This applies to all successful and failed synthesis attempts. New accounts start with a lower limit of 10 requests per minute until the first top-up.

Queue handling under load

Requests are processed one after another. Under heavy load, your requests may wait in a queue, meaning generation time will exceed the typical real-time speed of 25 characters per second. Short phrases may take a few seconds to generate even when idle.

Get started

Sound wave illustration for: Get started

Start by signing up via Google for an instant $0.50 trial credit, valid for 7 days. This gives you 50,000 characters to test the API without entering payment details or waiting for manual approval.

Sign up with Google

Click the sign-up button and authorize access. Your account is created immediately on 66mmr.info, granting you a unique API key and access to the dashboard.

Get your API key

Your key appears on the dashboard, ready for use. You can rotate it anytime; the old key stops instantly. Use it with the OpenAI SDK by pointing base_url to https://api.66mmr.info/v1.

  • Use Authorization: Bearer tts-... in your headers
  • One active key per account
  • Instant invalidation upon rotation

Try the demo

Send a POST request to /v1/audio/speech with your key. Receive an audio file in MP3, WAV, or other formats. Use the provided curl example to test voice synthesis in minutes.

  • Send text up to 4096 characters
  • Receive audio file directly in response
  • Test with any of 8 voices

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. This model simplifies budgeting for large teams, as expenses are tied strictly to character count rather than unpredictable minute-based tiers.

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. TRON developer hub: TRC-20
  3. LibriTTS-R corpus (OpenSLR 141)
  4. openai-node on GitHub
  5. LibriVox public-domain audiobooks
  6. FFmpeg documentation
  7. openai-python on GitHub
  8. curl documentation
  9. IETF RFC 7845: Ogg Opus
  10. Tether (USDT)
  11. MDN: Web Speech API
  12. OpenAI API reference: create speech

More on this site