Frequently asked questions
What is text-to-speech?
Text-to-speech (TTS) is a cloud API that converts input text into audio files. You send text via JSON, and receive synthesized speech. It replaces static audio files with dynamic voice generation, allowing IVR systems to deliver fresh, relevant messages without pre-recording every possible scenario.
How does TTS work in IVR?
Your IVR system sends user input or menu options to our POST endpoint. The API generates audio in formats like mp3 or wav. The IVR plays the returned audio file to the caller. This enables dynamic menu updates and personalized greetings without manual recording for every variation.
What are the benefits of TTS?
TTS reduces storage costs by eliminating thousands of pre-recorded files. It allows instant updates to menu structures and supports 23 languages. With 8 distinct voices, you can maintain brand consistency while offering natural-sounding interactions. Failed requests are free, ensuring cost efficiency for high-volume call routing.
Which voices are best for IVR?
For IVR, use atlas (neutral, news) for clear instructions, or milo (friendly, conversational) for customer service. Felix (energetic) suits promos, while sage (calm) works for support. All voices support all 23 languages, allowing consistent tone across international call centers without needing separate voice actors.
How much does TTS cost?
Pricing is $10 per 1 million characters, or $0.01 per 1,000 characters. Spaces and punctuation count. Failed requests are free. There are no monthly plans, seats, or minimum spends. You pay only for what you generate, making it scalable for startups with variable call volumes.
Is TTS better than pre-recorded audio?
TTS offers flexibility and lower storage costs. Pre-recorded audio is static; TTS adapts to real-time data. TTS has slightly higher latency than playing a cached file but provides dynamic content. For IVR, the ability to update menus instantly without re-recording outweighs the minor generation delay.
What languages are supported?
We support 23 languages: Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish. Any voice can speak any language. Vietnamese, Indonesian, Thai, and others are not supported.
How fast is the synthesis?
Synthesis is roughly real-time, generating about 25 characters per second. Short phrases take a few seconds. Requests are processed sequentially, so under load, a request may wait in a queue. This speed is suitable for IVR menus where callers expect immediate, though not instant, responses.
Can TTS handle long texts?
Yes, but you must split text into parts of up to 4096 characters per request. The API returns separate audio files for each part. Your system should join these files to play a continuous message. This ensures reliable synthesis for longer hold messages or detailed instructions.
Is TTS accessible?
TTS supports 23 languages and offers clear, distinct voices like atlas and sage, aiding comprehension. Voices are based on licensed recordings. Generated files carry an inaudible watermark. While we do not offer SSML, the clear audio output supports screen readers and standard accessibility protocols in IVR flows.
What is the API endpoint?
The endpoint is POST https://api.sotmal.info/v1/audio/speech. Send a JSON body with model, input, voice, response_format, and speed. Use the same request shape as OpenAI’s speech endpoint. Other endpoints include GET /v1/voices and GET /v1/models for configuration details.
How do I integrate TTS?
Include your Bearer token in the Authorization header. Send text via the POST endpoint. Receive audio in your chosen format (mp3, wav, opus, flac, pcm). Use the OpenAI SDK by changing the base_url, or use plain HTTP. Split long texts manually before sending.
What is the character limit?
Each request supports up to 4096 characters of input text. If your message exceeds this, split it into multiple requests. Generate audio for each part, then join the files in your IVR system. This ensures reliable synthesis without truncation or errors in the audio output.
Can I customize the voice?
You cannot customize voice timbre or clone voices. You choose from 8 presets: aria, nova, luna, sage, felix, milo, atlas, orion. You can adjust speed (0.25–4.0) and specify a language. Emotion tags and SSML are not supported. The voice remains consistent across all requests.
Is there a free trial?
Yes, signing in with Google gives $0.50 credit (approx. 50,000 characters) valid for 7 days. Email/password accounts start without trial credit. Failed requests are free. This allows testing latency and quality before committing to USDT top-ups for production IVR usage.
How do I pay for TTS?
Top up using USDT on the TRON network (TRC-20). Minimum deposit is $5. Credits are automatic within minutes. No cards, PayPal, or invoices. Balance does not expire. You can issue a new API key at any time, invalidating the old one immediately.
What is USDT?
USDT (Tether) is a stablecoin pegged to the US dollar. We accept it on the TRON network (TRC-20) for API credits. This allows fast, low-cost global payments without banking delays. Transfers are credited automatically, enabling immediate usage for your IVR system after deposit.
Is data stored?
Input text is used only to produce audio and is not used for training. Generated audio is returned in the response and not stored. We keep request metadata (time, characters, status) for billing purposes. No personal data is retained beyond what is needed for service operation.
Can I use TTS for hold music?
TTS is designed for speech, not music. Use voices like aria or nova for spoken hold messages. For background music, pre-recorded audio is better. TTS excels at delivering dynamic information, such as wait times or menu options, rather than ambient musical tracks.
What is the difference between tts-1 and tts-1-hd?
Both models use the same engine and price. The difference is audio quality: tts-1-hd offers higher fidelity. Both accept the same parameters and voices. Choose tts-1-hd if audio clarity is critical for your IVR customers, otherwise, tts-1 provides identical functionality at the same rate.
Definition of TTS in IVR
Text-to-speech (TTS) in Interactive Voice Response (IVR) systems replaces static audio files with dynamically generated speech. This API converts input text into audio files, allowing call centers to deliver up-to-date information without pre-recording every possible message.
By injecting dynamic content directly into the call flow, startups can reduce operational costs and maintain 24/7 availability. The system handles voice synthesis, ensuring clear communication for high-volume call routing across multiple languages.
Automated Voice Generation
Our API acts as a cloud-based engine that receives text and returns an audio file. It supports models like tts-1 and tts-1-hd, generating speech in formats such as mp3 or wav. This automation eliminates the need for manual recording sessions for every message variation.
- Cloud-based API endpoint
- Multiple output formats (mp3, wav, opus)
- No manual recording required
Dynamic Content Injection
Instead of playing static files, the IVR sends variable text—such as account balances or wait times—to the TTS API. The system generates the audio on the fly, ensuring callers hear current data. This supports 23 languages, adapting to the caller's preference when the language parameter is set.
- Real-time text processing
- Support for 23 languages
- Variable data integration
Synthesis Process
The synthesis process involves sending a POST request with the text, voice choice, and language code. The API returns the audio file, which the IVR plays immediately. For longer texts, the system splits the input into parts under 4096 characters to ensure consistent quality.
- POST request structure
- Character limit of 4096
- Split and join for long texts
How It Works
Our API accepts text payloads via a standard HTTP POST request. You specify the desired voice, language, and audio format within the JSON body. The system processes up to 4,096 characters per request, ensuring consistent input handling for IVR menu prompts or dynamic status updates.
Once submitted, the engine synthesizes the audio. Requests are processed sequentially, meaning latency depends on current queue load. Longer texts are automatically split into manageable segments to maintain stability, then joined seamlessly for playback in your call routing system.
Text Input Process
Send a JSON payload to the speech endpoint. Include the text string, model selection (tts-1 or tts-1-hd), and the target voice preset. You can also define the output format (mp3, wav, opus) and adjust speech speed between 0.25x and 4.0x to fit your IVR timing constraints.
- Supports ISO language codes for precise voice selection
- Max input length is 4,096 characters per request
- Optional speed adjustment from 0.25 to 4.0
Audio Generation Steps
The engine converts your text into synthetic audio files. Each request is charged based on character count, including spaces and punctuation. Failed requests incur no cost, protecting your budget during integration testing or high-volume spikes.
For texts exceeding the limit, the API splits the input into parts. Each segment is synthesized individually and returned as separate audio files, which your application can concatenate before playing to the caller.
API Response Handling
Upon completion, the API returns the generated audio file directly in the response body. You can store this file or play it immediately to your telephony platform. The response includes metadata for billing verification, allowing you to track usage accurately.
Use standard HTTP clients or the OpenAI SDK (with adjusted base URL) to handle responses. This simplicity reduces integration overhead for startups building custom IVR solutions without needing proprietary SDKs.
Benefits for Call Centers
Text-to-speech integration eliminates the need for human operators to read static scripts, significantly lowering per-call labor costs. By generating audio dynamically, call centers avoid the overhead of recording, storing, and managing thousands of unique audio files for different scenarios.
This approach ensures consistent service delivery around the clock without requiring shift staff. The API handles the heavy lifting, allowing startups to maintain a professional voice presence even during off-hours or high-volume periods.
Reduced Operational Costs
Dynamic voice generation reduces the need for large teams of voice actors or call center agents to read scripts. With a cost of $0.01 per 1,000 characters, you pay only for what you use, avoiding fixed monthly salaries for repetitive reading tasks.
- Pay-as-you-go model eliminates fixed labor costs for script reading.
- No studio time or recording equipment expenses.
- Reduced storage costs for libraries of pre-recorded prompts.
24/7 Availability
The API operates continuously, ensuring that your IVR system never goes silent due to staff breaks or holidays. Customers receive natural-sounding responses at any hour, improving satisfaction without requiring overnight human staffing.
- Consistent voice quality day and night.
- No downtime for staff shifts or holidays.
- Immediate response generation without queue delays for human agents.
Scalability
As your call volume grows, the API scales instantly without hiring new voice talent. Whether you handle 100 or 100,000 calls, the synthesis engine processes requests sequentially, maintaining consistent output quality and cost structure.
- Handle traffic spikes without hiring temporary staff.
- Uniform voice across all languages and regions.
- Easy to add new menu options without new recordings.
Voice Selection Criteria
Choosing the right voice is critical for IVR clarity and user retention. Our API provides eight distinct presets, ranging from the warm narration of Aria to the neutral news style of Atlas, ensuring your automated menus sound professional and understandable across different call center use cases.
Tone and Clarity
Select voices based on the context of the message. Use Sage for instructional guides or Orion for documentary-style updates. Each voice is engineered for high intelligibility, reducing customer confusion and repeat calls without needing complex configuration.
- Aria for warm narration
- Atlas for neutral news
- Sage for instructional content
- Orion for deep documentary tone
Language Support
Support 23 languages including English, Spanish, German, and Japanese. Any voice can speak any supported language, allowing you to maintain a consistent brand voice across global markets. Explicitly set the language parameter for non-English Latin text to avoid detection errors.
- 23 supported languages
- Universal voice capability
- Explicit language parameter recommended
Brand Alignment
Align your IVR personality with your startup’s identity. Whether you need the upbeat energy of Nova for promotions or the calm reliability of Milo for support, our presets allow you to tailor the auditory experience to your specific brand guidelines.
Technical Requirements
To integrate our TTS API into your IVR system, you need a stable internet connection capable of handling HTTPS requests to https://api.sotmal.info/v1. The system requires standard HTTP headers, including your unique Bearer token for authentication, ensuring secure and reliable data transmission between your telephony platform and our synthesis engine.
Integration is straightforward since the endpoint mirrors the OpenAI speech API structure. You can use existing OpenAI SDKs by updating the base_url to our domain, or send raw JSON payloads. This compatibility allows startups to leverage familiar development workflows without learning a proprietary protocol, reducing integration time and complexity.
Internet Connectivity
Ensure your telephony platform has reliable outbound internet access. Our API relies on standard HTTPS connections, so any firewall or proxy configurations must allow traffic to api.sotmal.info on port 443. Dropped connections during synthesis may result in incomplete audio segments being played to the caller.
- Stable outbound HTTPS access required
- Port 443 must be open
- Proxy compatibility recommended
API Integration
Send a POST request to /v1/audio/speech with your text, voice selection, and format preferences. The API returns the audio file directly in the response body, allowing your IVR system to buffer the audio immediately. We support standard formats like MP3, WAV, and OPUS, ensuring broad compatibility with most PBX systems.
Authentication is handled via a Bearer token in the Authorization header. Each account has one active key, which can be rotated instantly if compromised. This simple model reduces overhead compared to complex OAuth flows, making it ideal for high-volume IVR deployments.
- POST to
/v1/audio/speech - Supports MP3, WAV, OPUS
- Bearer token authentication
- OpenAI SDK compatible
Latency Considerations
Synthesis typically takes about one second per 25 characters of text. For IVR menus, this means short prompts are generated nearly instantly, but longer descriptions may introduce a noticeable pause. Requests are processed sequentially, so under high load, queuing may add further delay.
To minimize perceived latency, keep menu options concise. Avoid sending large blocks of text in a single request. Splitting longer messages into smaller chunks can help maintain a responsive user experience, though each chunk incurs its own generation overhead.
- ~1 second per 25 characters
- Sequential processing under load
- Short prompts recommended
- Split long texts for speed
Use Cases in IVR
Our TTS API transforms static IVR trees into dynamic conversational interfaces by generating natural voice menus in real time. Instead of rigid, pre-recorded prompts, the system generates audio from contextual text, allowing for precise control over tone and pacing via the speed parameter.
Menu Navigation
Generate dynamic routing options that adapt to user context, such as account numbers or time of day. The API handles up to 4096 characters per request, enabling complex multi-level menus without latency spikes typical of large audio files.
- Dynamic branch labels
- Adaptive language switching
- Real-time menu updates
Status Updates
Deliver instant order or ticket statuses using the tts-1 or tts-1-hd models. By setting the language parameter explicitly, the system ensures clear enunciation of technical terms or reference numbers across all 23 supported languages, reducing misinterpretation in high-volume call centers.
Hold Messages
Replace monotonous hold loops with varied, branded content. Use voices like sage for calm instructional tones or felix for energetic updates. This reduces perceived wait times and operational costs by eliminating the need to produce and store hundreds of unique audio files.

Comparison with Pre-Recorded Audio
Pre-recorded audio requires manual updates for every message change, while our API generates fresh audio on demand. This eliminates the need for re-recording sessions or managing a library of static files.
Instead of storing hundreds of WAV files for different scenarios, you send text to the API and receive the audio instantly. This reduces storage overhead and keeps your IVR content current without manual intervention.
Flexibility
Static recordings are fixed; if a phone number or holiday schedule changes, you must re-record the entire file. Our API allows you to inject dynamic data like account balances or wait times directly into the text before synthesis.
You can update messages instantly across all languages. If a policy changes, you update the source text once, and the API regenerates the audio for all supported voices and languages without human effort.
- Dynamic data injection (names, dates, balances)
- Instant global updates across all voices
- No re-recording for minor text changes
Storage Efficiency
Traditional IVR systems store thousands of audio files, consuming server space and increasing management complexity. Our API returns audio on the fly, removing the need for large local storage archives.
You only pay for what you generate. Failed requests are free, and you do not pay for idle storage of unused menu options. This model scales efficiently with call volume.
- Zero local storage for generated audio
- Pay-per-character instead of per-file
- No costs for idle menu options
Personalization
Pre-recorded messages are generic. With our API, you can personalize greetings by inserting the caller’s name or specific account details into the text before synthesis.
This creates a more engaging experience without needing multiple pre-recorded variations for each possible customer scenario. The audio is generated uniquely for each call in real-time.
- Caller name insertion
- Real-time account balance reading
- Unique audio per call instance
Implementation Steps
Setting up your IVR system with our TTS API is a straightforward three-step process designed for rapid deployment. By following these implementation steps, startups can replace static menu recordings with dynamic, natural-sounding voice interactions without significant infrastructure changes.
The process begins with securing access, selecting the appropriate voice profiles for your brand, and verifying integration through concrete examples. This ensures your call routing system delivers clear, cost-effective audio responses immediately.
API Key Acquisition
Navigate to /get-api-key/ to create an account via Google or email. You will receive a unique Bearer token (tts-...) immediately. This key authenticates your requests to https://api.sotmal.info/v1 and allows you to monitor your balance and usage directly on the dashboard.
- Sign up via Google or email/password
- Receive a unique Bearer token instantly
- Monitor balance and character usage in real-time
Voice Choice
Select from eight presets like aria for clear narration or atlas for neutral news-style menus. Ensure the voice matches your brand’s tone. Remember that any voice can speak any of the 23 supported languages, so set the language parameter explicitly for non-English Latin text to avoid detection errors.
- Choose from 8 distinct voice presets
- Match tone to brand identity
- Set language parameter for accuracy
Integration Testing
Send a test POST request to /v1/audio/speech with sample text. Verify that the response returns a valid audio file in your chosen format (mp3, wav, etc.). Check that the speed parameter (0.25–4.0) adjusts the playback rate correctly for IVR timing constraints.
- Send test POST request
- Verify audio file response
- Adjust speed for timing
Common Challenges
IVR systems demand low latency to maintain a natural conversational flow. Because requests are processed sequentially, heavy load can introduce queue delays, making short phrases take a few seconds to generate. This latency is critical for real-time user experience in call routing.
The API processes text in batches, meaning under significant load, synthesis may wait in a queue. For high-volume call centers, this sequential processing must be accounted for in your IVR design to prevent audible pauses that disrupt the caller experience.
Latency Issues
Requests are processed one after another, so under load, a request may wait in a queue. While typical speed is roughly real time (25 characters per second), short phrases often take a few seconds to generate. This sequential processing can introduce noticeable delays if many concurrent calls are active.
- Sequential processing causes queuing under high load
- Short phrases may take several seconds to synthesize
- Typical speed is ~25 characters per second
Character Limits
Each API request accepts up to 4,096 characters. For longer scripts, you must split the text into parts, generate audio for each, and join the files. This adds complexity to your IVR logic but ensures no data loss due to payload size constraints.
- Maximum 4,096 characters per request
- Split long texts into multiple parts
- Join generated audio files manually
Language Detection
- Latin text defaults to English
- Explicit language code required for non-English Latin scripts
- 23 supported languages require correct tagging
Future Trends
As IVR systems evolve, API-driven TTS allows for real-time adaptation to customer data, moving beyond static scripts. This shift enables dynamic voice menus that respond instantly to user input, reducing hold times and improving first-call resolution rates without manual intervention.
AI Advancements
Future integrations will likely combine TTS with advanced NLP for deeper context understanding. While our current API handles dynamic text injection efficiently, upcoming AI layers may automate voice selection based on sentiment or urgency, optimizing the caller journey further.
Natural Voices
The demand for human-like intonation in IVR is growing. Our current models, trained on licensed recordings, provide clear, natural speech across 23 languages. As synthesis engines improve, we expect even more nuanced prosody.
Emotion Recognition
While our current preset voices do not support real-time emotion tags, future TTS APIs may integrate with voice stress analysis. This would allow the system to detect caller frustration and adjust tone or speed dynamically, enhancing the customer experience during high-volume call routing.
Conclusion
Our API provides a cost-effective, scalable solution for IVR systems, enabling startups to deploy natural-sounding voice menus without overpaying for pre-recorded assets or complex infrastructure.
Summary of Benefits
By leveraging dynamic text-to-speech, you gain flexibility in content updates, reduce storage needs, and maintain consistent brand alignment across 23 languages. The API’s straightforward pricing and lack of monthly fees make it ideal for high-volume call routing.
Recommendation for Startups
For startups prioritizing speed to market and cost efficiency, this TTS API offers a reliable foundation. It balances latency and character efficiency, allowing you to scale voice menus without significant engineering overhead.
Next Steps
Visit /get-api-key/ to secure your API key and start with $0.50 in free trial credit. Review the documentation to integrate the speech endpoint into your IVR system and begin testing voice selection and latency.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. For longer scripts, split the text into parts, generate audio files, and join them programmatically.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- TRON developer hub: TRC-20
- curl documentation
- W3C WAI: making audio and video accessible
- ISO 639 language codes
- LibriVox public-domain audiobooks
- OpenAI API reference: create speech
- Node.js documentation
- MDN: Web Speech API
- openai-python on GitHub
- LibriTTS-R corpus (OpenSLR 141)
- openai-node on GitHub
