Human Sounding Text to Speech That Keeps Callers Engaged

By Lvl2 editorial team · Updated 2026-11-19

AuricText is a cloud text-to-speech API built for startup teams running IVR systems. We deliver text-to-speech audio via a cloud API, stripping away enterprise complexity so you can prototype and deploy voice interfaces fast. This page explains how our straightforward API fits into your call center workflows, from simple prompts to complex multi-step interactions.

You will find details on our 8 distinct voice personalities, 23 supported languages, and transparent pricing based on character usage. Learn how to integrate our OpenAI-compatible endpoint, manage your API key, and handle long messages efficiently. Whether you are building a new system or optimizing an existing one, this guide covers the essentials for reliable, production-ready voice generation.

Human Sounding Text to Speech That Keeps Callers Engaged

Frequently asked questions

How natural does the speech sound?

Our voices are based on licensed recordings from LibriTTS-R and LibriVox public-domain audiobook readings. They are designed to be natural and human-sounding, keeping callers engaged. Every generated file carries an inaudible watermark that marks it as synthetic speech, ensuring transparency while maintaining high audio quality for your IVR system.

Can I use these voices for commercial calls?

Yes. Our voices are based on licensed recordings (LibriTTS-R, CC BY 4.0) and public-domain audiobook readings. You can use them for commercial calls without additional licensing fees. The API delivers clear, natural-sounding audio suitable for IVR systems, and there are no restrictions on the volume of characters you generate.

What happens if the API is slow?

Requests are processed one after another, so under load a request may wait in a queue. Typical speed is roughly real time on long texts. Short phrases take a few seconds. If the API is slow, your application should handle the delay gracefully.

Do I need to install special software?

No. You can use the official OpenAI SDKs by changing the base URL to https://api.lvl2.info/v1, or use plain HTTP. We do not provide an official SDK of our own, nor mobile SDKs or WordPress plugins. Just send a POST request with your API key and receive an audio file.

How do I handle long customer messages?

Split your text into parts of up to 4096 characters and join the audio files. The API accepts up to 4096 characters per request. For longer texts, you must handle the splitting logic in your application and concatenate the resulting audio files to create a seamless experience for your callers.

Are there limits on characters per call?

Yes, each API request accepts up to 4096 characters. Spaces and punctuation count as characters. If your message exceeds this limit, you must split the text into multiple requests. Failed requests are free, so you can retry or split without worrying about unnecessary charges for incomplete audio generation.

Can I mix different voices in one call?

Yes, by sending multiple requests. You can specify different voices for each request, such as using Aria for greetings and Felix for promotions. Since we do not support SSML, you must manage voice switching in your application logic by stitching together the audio files returned from each individual API call.

What formats work best with phone systems?

We support mp3, wav, opus, flac, and pcm formats. For phone systems, wav or pcm are often preferred for compatibility, while mp3 offers smaller file sizes. You can specify the desired response_format in your request to ensure the audio is delivered in the format your IVR system expects.

How quickly is audio generated?

Typical speed is roughly real time on long texts, about 25 characters per second. Short phrases take a few seconds. Requests are processed one after another, so under load a request may wait in a queue. This ensures consistent quality, though latency may increase during high traffic periods.

Do I get refunds for failed requests?

Yes. If synthesis fails, the charge is returned in full. You are only billed for successful audio generation. This means you can test your integration confidently, knowing that any errors in processing or server-side failures will not result in unnecessary deductions from your account balance or trial credit.

Can I change voices mid-campaign?

Yes. You can switch voices by simply changing the voice parameter in your API requests. We offer 8 presets, including female voices like Aria and Nova, and male voices like Felix and Milo. You can use different voices for different segments of your campaign without any additional setup.

Is there a limit on concurrent calls?

We do not specify a hard limit on concurrent calls, but requests are processed one after another. Under load, a request may wait in a queue. With a rate limit of 60 requests per minute per key, you can manage concurrency by batching requests or using multiple API keys if needed.

How do I track my usage costs?

Your account dashboard shows your balance, characters left, and usage history. Pricing is $10 per 1 million characters. You can monitor your spending in real-time. Failed requests are free, so your usage costs reflect only successful audio generations, giving you accurate visibility into your IVR system's operational expenses.

What if I need a new language quickly?

We support 23 languages, including English, Spanish, German, French, and Japanese. Any voice can speak any of these languages. If you need a new language, just specify the ISO code in your request. The system guesses the language if not set, but setting it explicitly ensures accuracy for non-English texts.

First request

curl

curl https://api.lvl2.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

Why startups choose natural voices for IVR

Sound wave illustration for: Why startups choose natural voices for IVR

Robotic menus drive callers away. Natural voices reduce abandonment by making automated interactions feel human, keeping users on the line longer when they need help.

When callers hear clear speech, they are more likely to resolve their issue without human intervention.

Reducing call abandonment rates

Reducing call abandonment rates

Monotone IVRs frustrate users, causing early hang-ups. Natural voices maintain engagement, ensuring callers stay until their query is resolved.

Startups can prototype and deploy voice interfaces with simple API calls.

  • Lower drop-off rates with human-like intonation
  • Higher completion rates for complex queries

Improving customer satisfaction scores

Clear audio helps callers understand instructions.

Consistent quality across 23 languages ensures every caller feels valued, regardless of their native tongue.

Use warm narrators for complex instructions or upbeat voices for promotions to match your brand's tone.

  • Clear, expressive audio reduces cognitive load
  • Consistent quality across 23 languages

Simplifying voice script design

Natural voices handle natural phrasing well. You don’t need rigid, robotic scripts; simple text works effectively.

Use the API to generate audio for any text up to 4096 characters. This flexibility lets you write conversational scripts without worrying about technical constraints.

Split longer texts into parts if needed, and join the audio seamlessly for your IVR system.

  • Write conversational scripts, not rigid commands
  • Flexible text input up to 4096 characters

Choose from 8 distinct voice personalities

Sound wave illustration for: Choose from 8 distinct voice personalities

AuricText offers eight distinct voice presets designed to match the tone of your IVR flow. Each voice is derived from licensed recordings.

Whether you need a warm narrator for complex instructions or an upbeat voice for promotions, you can select the perfect tone to keep your customers engaged from the first ring.

Warm narrators for complex instructions

Aria provides a clear, warm tone ideal for guiding callers through multi-step menus. Luna offers a soft, storytelling quality that makes detailed information easier to digest.

Use these voices when explaining policies or complex processes where clarity reduces confusion and call-back rates.

  • Aria: Warm, clear narrator
  • Luna: Soft, storytelling tone

Upbeat voices for promotional calls

Nova brings a bright, energetic energy to outbound promotional calls, keeping listeners attentive. Felix adds a punchy, promotional rhythm perfect for time-sensitive offers.

These voices help maintain high energy levels during short, impactful messages where engagement is critical.

  • Nova: Bright, upbeat delivery
  • Felix: Energetic, promo-focused

Calm tones for support queues

Milos conversational style feels friendly and approachable for support queues, reducing caller anxiety. Sage delivers a calm, instructional tone that works well for holding messages or status updates.

Atlas provides a neutral, reliable sound for news-style announcements, while Orion offers a deep, documentary feel for premium or serious contexts.

  • Milo: Friendly, conversational
  • Sage: Calm, instructional
  • Atlas: Neutral, news-style
  • Orion: Deep, documentary

Global support for diverse customer bases

Sound wave illustration for: Global support for diverse customer bases

AuricText supports 23 languages, including Arabic, Chinese, English, French, German, Hindi, Japanese, Korean, Spanish, and Turkish. Each voice can speak any of these languages, allowing you to standardize your audio production pipeline across multiple markets.

We handle language detection automatically: if you do not specify a language, the system guesses based on the script, defaulting to English for Latin text. For non-English Latin-script text, we recommend setting the language explicitly to ensure accuracy. This reduces configuration errors when integrating with global call centers.

23 languages supported natively

Our API natively supports 23 languages, covering major global markets. This includes Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish. You can switch voices or languages per request, enabling dynamic IVR menus that adapt to the caller's preferred language.

  • Full support for 23 distinct languages.
  • Native rendering without external translation steps.
  • Switch languages dynamically per API call.

Automatic language detection

The API automatically detects the language from the input script. If the language field is omitted, the system infers the language, defaulting to English for Latin scripts. For non-English Latin-script text, we recommend explicitly setting the language parameter to prevent misidentification. This simplifies integration for startups managing multilingual IVR flows without complex routing logic.

  • Automatic detection from input script.
  • Latin text defaults to English.
  • Explicit language setting recommended for clarity.

Consistent quality across dialects

Each of the 8 voice presets maintains consistent audio quality across all supported languages. Whether you are using Aria for warm narrations or Felix for energetic promos, the synthetic speech remains clear and natural. This consistency ensures that your customer experience does not degrade when switching between languages or voice profiles, keeping callers engaged throughout the call.

  • Consistent audio quality across all 23 languages.
  • Unified voice presets for all markets.
  • Natural sound maintained regardless of language.

Simple API integration for developers

Sound wave illustration for: Simple API integration for developers

Drop-in compatibility with existing OpenAI SDKs makes integration straightforward. Simply point your client to our base URL and use standard keys.

OpenAI-compatible endpoint structure

Use the familiar POST /v1/audio/speech endpoint. Pass your text, select a voice from our eight presets, and receive an audio file. Our request shape mirrors the industry standard, so you can reuse existing code with minimal changes.

  • Standard JSON body with model, input, voice, and response_format fields.
  • Supports all major OpenAI SDKs by updating the base_url to https://api.lvl2.info/v1.
  • Compatible with standard HTTP clients for simple scripts.

Fast response times for short queries

Short phrases are generated in a few seconds, keeping your IVR prompts snappy. For longer texts, the system processes requests sequentially, typically achieving real-time speeds of about 25 characters per second.

  • Instant audio for short prompts under 4096 characters.
  • Predictable processing times for standard use cases.

Reliable audio generation pipelines

We handle the heavy lifting of audio synthesis so you can focus on your call flow. Failed requests are free, ensuring you only pay for successful generations. For texts exceeding 4096 characters, split them into parts and join the resulting audio files seamlessly.

  • Failed requests are refunded immediately.
  • Automatic handling of long texts via client-side splitting.
  • Support for MP3, WAV, OPUS, FLAC, and PCM formats.

Transparent pricing for growing volumes

Sound wave illustration for: Transparent pricing for growing volumes

Pricing is straightforward: $10 per 1 million characters, or $0.01 per 1,000 characters. Spaces and punctuation count as characters, so your cost scales exactly with your script length. Failed requests are free, ensuring you only pay for successful audio generation.

Pay per character usage model

There are no monthly plans, seats, or minimum spend requirements. You pay only for what you use, making it ideal for startups with variable call volumes.

  • $0.01 per 1,000 characters
  • Spaces and punctuation count
  • Failed requests are free

No minimum monthly spend

Our pay-as-you-go model means no surprises on your invoice. Top-ups via USDT on the TRON network are credited automatically within minutes, allowing you to maintain control over your budget without committing to long-term contracts.

Free trial credits included

Sign in with Google to receive $0.50 in trial credit (about 50,000 characters) valid for 7 days. This lets you test the API with real IVR workflows before committing funds. Email-based accounts start without trial credit.

  • $0.50 credit for Google sign-ins
  • Valid for 7 days
  • Approximately 50,000 characters
Transparent pricing for growing volumes

Technical reliability for production use

Sound wave illustration for: Technical reliability for production use

AuricText delivers consistent audio quality tailored for production IVR systems. Each request generates speech that maintains consistent quality across different voice personalities.

Consistent audio quality standards

Our engine uses licensed recordings from LibriTTS-R and LibriVox. Whether you select a warm narrator or a neutral news voice, the output remains stable and reliable for customer-facing applications.

  • Licensed source material ensures consistent vocal texture
  • Uniform quality across all 8 voice presets

Standard format support (MP3, WAV)

We provide standard audio formats compatible with most IVR platforms. You can request MP3 for efficient bandwidth usage or WAV for uncompressed clarity. Other formats like OPUS, FLAC, and PCM are also available depending on your system's requirements.

  • MP3 and WAV support out of the box
  • Optional OPUS, FLAC, and PCM formats
  • Compatible with standard telephony systems

Predictable processing times

Short phrases typically generate within a few seconds. For longer texts, processing occurs in real-time, roughly 25 characters per second. Requests are processed sequentially, so under high load, a request may wait in a queue, but generation remains stable and error-free.

  • Short queries: few seconds generation time
  • Long texts: ~25 characters per second
  • Sequential processing ensures stable queues

Data privacy for sensitive conversations

Sound wave illustration for: Data privacy for sensitive conversations

When your IVR handles billing details or personal information, knowing where that text goes matters. Our API processes your input solely to generate the required audio file. Once the synthesis is complete and the audio is delivered to your system, the original text is not retained for any other purpose. This ensures that sensitive customer data remains transient during the conversion process, reducing exposure risks in your call flow architecture.

Text used only for generation

Your input text is consumed exclusively for the purpose of creating the audio output. We do not store the actual content of your scripts after the request is fulfilled. This strict limitation means your voice scripts, whether they contain account numbers or personal narratives, are not kept in long-term archives for analysis or reuse.

No voice training on your data

Unlike some services that use customer data to improve their models, we do not use your input text or generated audio to train our voice engines. The voices you hear are based on licensed recordings from LibriTTS-R and LibriVox, not on your company’s specific conversations. Your proprietary content remains yours, with no risk of it influencing the general behavior of our synthetic voices.

Secure key management

Access to your audio generation is controlled via a single active API key per account. This key is used for authentication in every request you send to the endpoint. If you suspect any unusual activity or simply want to refresh your security posture, you can issue a new key at any time. The previous key stops working immediately, ensuring that no unauthorized systems can continue generating audio using old credentials.

  • One active key per account
  • Instant revocation of old keys
  • Standard Bearer token authentication

Start building your voice interface today

Sound wave illustration for: Start building your voice interface today

Get started in minutes by signing up on our /get-api-key/ page. Choose Google or email authentication to receive your unique API key instantly. Your dashboard displays your key, balance, and a ready-to-use cURL example, so you can make your first call without writing boilerplate code.

Our documentation covers the standard OpenAI-compatible endpoint structure, making it easy to integrate with your existing stack. Use the official OpenAI SDKs by simply changing the base URL, or send raw HTTP requests. Clear examples help you generate audio files for your IVR prompts quickly.

Quick API key setup process

Sign up using Google or an email address to receive your key immediately. The dashboard shows your balance, characters left, and a working cURL example so you can test the API right away.

  • Google or email sign-up
  • Instant key generation
  • Dashboard with balance and usage stats

Developer-friendly documentation

The docs explain the JSON request shape, supported models, and voice options. Since the endpoint mirrors OpenAI’s structure, you can use existing OpenAI SDKs by updating the base URL to https://api.lvl2.info/v1.

  • OpenAI-compatible endpoints
  • Standard JSON bodies
  • Ready cURL examples

Email support for technical queries

For integration questions or billing details, reach out via email. We do not offer live chat or phone support, but our team is available to help you resolve technical issues efficiently.

  • Email support only
  • No live chat
  • Technical and billing assistance

Frequently asked questions

Sound wave illustration for: Frequently asked questions

Our voices are synthesized from licensed recordings and sound natural for IVR flows. They lack SSML or emotion tags but deliver clear, human-like speech.

Each audio file contains an inaudible watermark identifying it as synthetic, which is standard for this type of technology.

How natural does the speech sound?

Our voices are synthesized from licensed recordings and sound natural for IVR flows. They lack SSML or emotion tags but deliver clear, human-like speech.

Can I use these voices for commercial calls?

Yes. The voices are based on licensed recordings, and you can use the generated audio commercially without additional restrictions.

What happens if the API is slow?

Requests are processed sequentially. Under load, your request may wait in a queue. Typical speed is roughly real-time (25 characters per second), but short phrases may take a few seconds.

Do I need to install special software?

No. You only need an API key and an HTTP client. The endpoint is compatible with the OpenAI SDK structure.

How do I handle long customer messages?

Split text into parts of up to 4,096 characters. Generate audio for each part and join the files on your end.

Are there limits on characters per call?

Each API request supports up to 4,096 characters. You can combine multiple requests for longer messages.

Can I mix different voices in one call?

Yes. Call the API multiple times with different voice parameters and concatenate the resulting audio files.

What formats work best with phone systems?

We support MP3, WAV, Opus, FLAC, and PCM. MP3 or WAV are standard for most IVR systems.

How quickly is audio generated?

Typical speed is roughly real-time. Short phrases may take a few seconds. Processing is sequential, so load affects speed.

Do I get refunds for failed requests?

Yes. If synthesis fails, the charge is returned in full. Failed requests are free.

Can I change voices mid-campaign?

Yes. Just update the voice parameter in your next API call. There is no locking mechanism.

Is there a limit on concurrent calls?

The limit is 60 requests per minute per key (10 per minute before the first top-up). There is no hard limit on concurrency, but requests queue if you exceed the rate.

How do I track my usage costs?

Check your dashboard at /get-api-key/ to see balance, characters used, and usage history. Pricing is $0.01 per 1,000 characters.

What if I need a new language quickly?

We support 23 languages. Set the language parameter or let the API guess from the script. No setup is required.

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. Its predictable pacing ensures that complex instructions are delivered clearly, reducing the chance of customer confusion.

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. OpenAI API reference: create speech
  3. curl documentation
  4. LibriVox public-domain audiobooks
  5. LibriTTS-R corpus (OpenSLR 141)
  6. openai-python on GitHub
  7. Tether (USDT)
  8. IETF RFC 7845: Ogg Opus
  9. Node.js documentation
  10. FFmpeg documentation
  11. LibriTTS-R paper (arXiv 2305.18802)
  12. W3C WAI: making audio and video accessible

More on this site