TTS API for Chatbots and Voice Assistants

By Predictz Voice editorial team · Updated 2026-11-18

Predictz Voice provides a reliable text-to-speech API tailored for startups building voice interfaces. Our catalog of 8 licensed voices supports 23 languages, ensuring your chatbots sound natural and consistent.

This page details the technical specifications, pricing, and integration workflow for developers. You will find concrete examples of JSON payloads, voice presets like Aria and Atlas, and clear limits for high-volume usage.

TTS API for Chatbots and Voice Assistants

Frequently asked questions

What makes a voice realistic for chatbots?

Predictz Voice uses eight presets based on licensed recordings from LibriTTS-R and LibriVox. These voices, such as Aria and Felix, are designed for natural, consistent speech in chatbots. Each generated file includes an inaudible watermark to mark it as synthetic. The engine supports 23 languages, ensuring your voice interface sounds professional and reliable for users.

How many characters can I send in one request?

You can send up to 4096 characters per request. This limit includes spaces and punctuation. For longer texts, you should split the content into parts of this size and join the resulting audio files. This approach ensures efficient processing while maintaining the quality of the synthesized speech output.

Which voices work best for voice assistants?

For voice assistants, Sage offers a calm, instructional tone, while Felix provides an energetic, promo style. Milo is friendly and conversational, and Atlas remains neutral for news-style updates. These eight presets cover various interaction styles, allowing you to choose the best fit for your specific chatbot personality and use case.

Can I use this API for commercial projects?

Yes, the API is designed for commercial use, including startups building voice interfaces. Input text is used only for production and not for training. Generated audio is returned in the response and not stored. You can integrate the API into your app without restrictions on the number of projects you build.

Does the API support streaming audio?

No, the API does not support streaming, WebSocket, or real-time partial audio. It returns the complete audio file after synthesis. Requests are processed sequentially, so under load, a request may wait in a queue.

How do I handle texts longer than 4096 characters?

Split your text into parts of up to 4096 characters each. Send each part as a separate request to the API. Then, join the resulting audio files on your end. This workflow ensures that long texts are processed efficiently and the final audio output is continuous and coherent.

Is there a limit on requests per minute?

Yes, the rate limit is 60 requests per minute per key. Before your first top-up, the limit is 10 requests per minute. If you exceed this limit, your requests may be delayed or rejected. Monitor your usage via the dashboard to ensure smooth operation for your chatbot or voice interface.

Can I clone my own voice for my brand?

No, custom voice cloning is not offered. You cannot upload your own voice or design a new voice. The API provides eight preset voices based on licensed recordings. These voices are consistent across all languages and projects, ensuring a standardized audio experience for your users without the need for custom cloning.

Do you offer SSML support for emphasis?

No, the API does not support SSML (Speech Synthesis Markup Language). You cannot use SSML tags for emphasis, pauses, or pronunciation adjustments. The API uses simple text input with optional language codes. For more control, you can adjust the speed parameter or choose different voices to achieve the desired tone.

What happens if a synthesis request fails?

If a synthesis request fails, the charge is returned in full. You are not billed for failed requests. This ensures that you only pay for successful audio generation. You can retry the request or adjust the input text and resubmit it without incurring additional costs for the failed attempt.

Are the voices based on real people?

Yes, the voices are based on licensed recordings from LibriTTS-R and LibriVox public-domain audiobook readings. Each generated file carries an inaudible watermark to identify it as synthetic speech, ensuring transparency.

Can I use the audio files in my app?

Yes, you can use the generated audio files in your app. The input text is used only for production and not for training. The audio is returned in the response and not stored by the API. You have full control over the generated audio files for your commercial or personal projects.

How is the language detected automatically?

If the language parameter is not set, the API guesses the language from the script. Latin script text defaults to English. For non-English Latin-script text, you should explicitly set the language parameter. The API supports 23 languages, including Arabic, Chinese, and Spanish, ensuring accurate synthesis for diverse user bases.

Do I need to pay for trial credits upfront?

No, you do not need to pay upfront for trial credits. Signing in with Google gives you $0.50 of credit, valid for 7 days. Accounts created with email and password start without trial credit. You can top up later using USDT on the TRON network, with a minimum spend of $5.

First request

curl

curl https://api.getpredictz.com/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 our TTS API

Sound wave illustration for: Why startups choose our TTS API

Predictz Voice provides a unified text-to-speech API designed for startups building voice interfaces. Our engine ensures consistent character across 23 languages, allowing chatbots to maintain a distinct brand identity regardless of the spoken language. This consistency is critical for user experience in global voice assistants.

Latency is optimized for real-time interaction, with synthesis speed roughly matching real-time generation. This ensures that chatbot responses feel immediate and natural, reducing user wait times during conversations.

Consistent character across languages

Consistent character across languages

Each of the eight voices retains its specific tone across all supported languages. Whether using Aria’s warm narrator style or Atlas’s neutral news delivery, the audio quality remains uniform. This prevents the disjointed feeling when a bot switches languages mid-conversation.

Low latency for chatbot responses

Requests are processed sequentially, typically generating audio at a rate of 25 characters per second for long texts. Short phrases may take a few seconds, but the system is optimized to minimize queue times under normal load. This speed supports responsive voice interfaces without noticeable delays.

Failed requests are free, ensuring you only pay for successful syntheses. This reliability helps maintain predictable performance metrics for your application.

Simple JSON integration

The API follows a standard JSON request shape, identical to the OpenAI speech endpoint. You send a POST request with the text, voice, and format, and receive the audio file directly. This simplicity reduces integration time for development teams.

Using the official OpenAI SDKs is straightforward; you only need to change the base_url to https://api.getpredictz.com/v1 and provide your API key. This compatibility leverages existing developer knowledge and tools.

  • POST endpoint at /v1/audio/speech
  • Standard JSON body format
  • Compatible with OpenAI SDKs
  • Direct audio file response

Voice catalog for voice assistants

Sound wave illustration for: Voice catalog for voice assistants

Predictz Voice provides eight distinct voice presets designed for consistent audio branding in chatbots and voice assistants. Each voice is derived from licensed recordings, ensuring legal clarity and a unique sonic identity for your application.

Aria: warm narrator tone

Aria delivers a clear, warm narration style ideal for onboarding flows or instructional content. Her tone remains engaging without feeling artificial, making her suitable for apps that require a trustworthy guide.

  • Warm and clear narrative style

Atlas: neutral news style

Atlas offers a neutral, professional delivery perfect for news updates or informational bots. His voice maintains a steady pace and neutral emotion, ensuring the focus remains on the information being conveyed to the user.

  • Neutral, reliable news delivery

Milo: friendly conversational

Milo provides a friendly, conversational tone that mimics natural human interaction. This preset works well for customer support chatbots where a approachable, relaxed voice helps reduce user friction during troubleshooting or queries.

  • Approachable, conversational style

Supported languages and scripts

Sound wave illustration for: Supported languages and scripts

Predictz Voice supports 23 languages, including English, Spanish, and German, allowing chatbots to serve global users without importing new voices for each market. Every voice preset can speak any supported language, ensuring a consistent auditory brand identity across your entire user base.

23 languages including English and Spanish

The API handles Arabic, Chinese, Danish, Dutch, Finnish, French, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Swahili, Swedish, and Turkish. Vietnamese, Indonesian, Thai, Ukrainian, Romanian, and Czech are not supported.

Automatic language detection

If you omit the language field, the system guesses the language from the text script. Latin script text defaults to English, so non-English Latin text should explicitly set the language parameter.

Explicit language code setting

Pass an ISO language code in the request body to ensure correct pronunciation and intonation. This prevents misidentification when the text contains mixed scripts or ambiguous phrasing.

Audio formats for playback

Sound wave illustration for: Audio formats for playback

The API accepts mp3, wav, opus, or flac formats via the response_format parameter. Your chatbot infrastructure determines the optimal choice. MP3 provides small file sizes for web delivery, while WAV offers uncompressed high fidelity for local processing. OPUS ensures efficient delivery with excellent compression ratios.

MP3 for web delivery

MP3 is the standard for web-based voice assistants. It balances audio quality with minimal bandwidth usage, ensuring quick response times for users on mobile networks. This format is ideal for reducing data costs in high-volume chatbot deployments.

WAV for high fidelity

Choose WAV when preserving the original audio quality is critical. This uncompressed format captures the full nuance of the synthesized voice, making it suitable for high-end voice interfaces or archival purposes where lossless audio is required.

OPUS for efficient delivery

OPUS provides superior compression efficiency compared to MP3. It maintains high perceptual quality at lower bitrates, making it excellent for efficient delivery in data-constrained environments or when optimizing for speed over limited bandwidth connections.

Pricing for high volume usage

Sound wave illustration for: Pricing for high volume usage

Pricing is transparent and linear: $0.01 per 1,000 characters, with spaces and punctuation counting toward the total. Failed requests are free, ensuring you pay only for successful synthesis. The standard $0.50 trial provides roughly 50,000 characters for seven days, allowing immediate validation of voice quality without financial commitment.

$0.01 per 1,000 characters

Cost scales directly with usage volume. At typical generation speeds, long texts process at roughly real-time, making high-volume voice interfaces predictable in expense. You are billed precisely for the characters sent, with no hidden fees for model selection between tts-1 and tts-1-hd.

  • Characters include all spaces and punctuation.
  • Failed requests incur zero charge.
  • Both tts-1 and tts-1-hd share the same rate.

Pay-as-you-go balance

Top-ups via USDT on the TRON network (TRC-20) are processed automatically within minutes. The minimum deposit is $5. Balances never expire, allowing you to maintain a standing credit for your chatbot's ongoing needs without recurring billing cycles.

  • Minimum top-up: $5.
  • Accepted currency: USDT (TRC-20).
  • Balance remains valid indefinitely.

No monthly minimums

There are no subscriptions, seats, or minimum spend requirements. This structure suits startups with variable traffic patterns, ensuring you do not pay for idle capacity during low-usage periods. You can scale up or down without contractual penalties.

Pricing for high volume usage

Technical constraints and limits

Sound wave illustration for: Technical constraints and limits

Predictz Voice is engineered for reliable, predictable performance in production chatbot environments. Understanding the specific technical boundaries of the API ensures your voice interfaces remain responsive and cost-effective without unexpected failures or latency spikes during peak usage.

4096 character input limit

Each API request accepts a maximum of 4096 characters, including spaces and punctuation. For longer dialogues or monologues, your application must split the text into chunks, synthesize each part, and join the resulting audio files seamlessly.

60 requests per minute

The standard rate limit is 60 requests per minute per API key. New accounts start with a lower limit of 10 requests per minute until the first top-up is processed. Exceeding this limit results in a 429 error.

Real-time synthesis speed

Synthesis typically runs at roughly real-time speed (25 characters per second). Short phrases may take a few seconds to process. Under load, requests are queued sequentially, so generation time may vary slightly depending on server traffic.

Integration with OpenAI SDKs

Sound wave illustration for: Integration with OpenAI SDKs

Predictz Voice uses the exact same request shape as the OpenAI speech endpoint, allowing startups to leverage existing OpenAI SDKs with minimal code changes. You simply point the SDK to our base URL and use the standard tts-1 model identifier.

Use tts-1 model

Use tts-1 for standard synthesis or tts-1-hd for higher fidelity. Both models use the same engine and pricing, so you can switch based on your quality needs without changing your code structure.

Change base_url easily

Update your SDK configuration to use https://api.getpredictz.com/v1 as the base URL. This directs all requests to our infrastructure while maintaining compatibility with official OpenAI client libraries.

Standard JSON body format

Send a JSON body with model, input, voice, and response_format. The Authorization header requires your tts-... bearer key. This standard format ensures consistent integration across Python, Node.js, and other supported languages.

Data privacy and licensing

Sound wave illustration for: Data privacy and licensing

Predictz Voice processes input text exclusively to generate audio output, ensuring your raw data remains isolated from our training pipelines. We do not use your submitted text to improve our models, preserving the confidentiality of your unique dialogue scripts and voice interfaces.

No training on input text

Your input text is used solely for synthesis and is not retained for model training. Generated audio is returned in the response and not stored, ensuring your data remains ephemeral unless you explicitly save it.

LibriVox licensed voices

All eight voice presets are derived from recordings licensed under CC BY 4.0 from the LibriTTS-R dataset, sourced from LibriVox public-domain audiobook readings. This ensures clear, commercial-friendly licensing for your startup's voice products without complex rights management.

Synthetic watermark included

Every generated audio file contains an inaudible watermark that explicitly marks the content as synthetic speech. This helps maintain transparency with end-users regarding the automated nature of the voice interface.

Get started with API key

Sound wave illustration for: Get started with API key

To begin, navigate to the /get-api-key/ page and authenticate using your Google account. This method instantly provisions your account on the Predictz Voice platform without requiring additional configuration steps.

Upon successful sign-in, you are immediately granted $0.50 in trial credit, valid for seven days. This provides approximately 50,000 characters of synthesis capacity, allowing you to validate the API's performance and voice quality before committing to paid usage.

Sign up with Google

Visit the registration page and select the Google sign-in option. The system creates a unique account identifier linked to your email, ensuring secure access to your dashboard and billing details.

Receive free trial credit

Your new account starts with $0.50 in credit. This allowance covers roughly 50,000 characters of text-to-speech generation. The credit expires seven days after account creation, providing a risk-free window to test integration workflows.

Copy your API key

After logging in, locate your unique API key on the dashboard. Use this key in the Authorization header (Bearer tts-...) for all API requests. You can generate a new key at any time, which instantly invalidates the previous one.

  • Key format: Bearer tts-...
  • Single active key per account
  • Instant revocation upon regeneration

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. This allows a single voice identity to maintain brand consistency across global markets without needing to retrain or swap voice assets for each new region.

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. Node.js documentation
  3. FFmpeg documentation
  4. LibriTTS-R paper (arXiv 2305.18802)
  5. Python documentation
  6. OpenAI API reference: create speech
  7. ISO 639 language codes
  8. TRON developer hub: TRC-20
  9. IETF RFC 7845: Ogg Opus
  10. W3C WAI: making audio and video accessible
  11. MDN: Web Speech API
  12. LibriVox public-domain audiobooks

More on this site