Frequently asked questions
What is TTS in the context of language learning?
Text-to-speech (TTS) converts written text into natural-sounding audio files. For language learning applications, it provides consistent pronunciation models and auditory reinforcement for vocabulary and grammar. Our API delivers high-quality synthetic speech, allowing product teams to integrate reliable audio into their platforms without managing complex infrastructure or hardware.
How many languages are supported by this API?
The API supports 23 languages, including English, Spanish, French, German, Chinese, Japanese, Arabic, Hindi, and Russian. Any voice can speak any of these languages. If the language is not specified, the system guesses it from the script, defaulting to English for Latin text.
Can I use the OpenAI SDK with this endpoint?
Yes. The endpoint uses the same request shape as the OpenAI speech endpoint. You can use official OpenAI SDKs by changing the base_url to https://api.play93k.com/v1 and providing your LinguaVoice API key. This allows seamless integration for teams already familiar with the OpenAI SDK structure.
What is the character limit for a single request?
Each API request supports up to 4096 characters of input text. For longer texts, you should split the content into parts of this size and join the resulting audio files. This ensures consistent processing and avoids potential timeouts or errors during synthesis.
Are voices customizable for specific learning styles?
The API offers eight preset voices with distinct tones, such as calm, energetic, or storytelling styles. However, there is no custom voice cloning or emotion tagging. You select from fixed presets like Aria, Nova, or Felix, which are based on licensed recordings from LibriVox public-domain audiobooks.
Does the API support streaming audio responses?
No, the API does not support streaming, WebSocket, or real-time partial audio. You send a text payload and receive the complete audio file in the response. This synchronous model simplifies integration, though it means you must wait for the full file generation before playback.
How are failures handled in the billing process?
Failed requests are free. If synthesis fails, the charge is returned in full to your account balance. This ensures you only pay for successfully generated audio files, protecting your budget from wasted credits due to temporary processing errors or invalid inputs.
What formats can I receive the audio in?
You can receive audio in mp3, wav, opus, flac, or pcm formats. Choose the format that best fits your application’s playback requirements. MP3 is suitable for web playback, while WAV and FLAC offer higher quality for offline learning modules or archival purposes.
Is there a minimum spend required to start?
No, there is no minimum spend or monthly plan. You can start with the free trial credit of $0.50 or top up via USDT on the TRON network with a minimum of $5. Balance does not expire, and you pay only for what you use.
Do I need to pay VAT on my purchases?
No, there are no invoices with VAT. Payments are processed via USDT on the TRON network, and no traditional payment methods like cards or PayPal are used. This simplifies billing for international teams by avoiding complex tax jurisdictions and invoice management associated with traditional payment processors.
How long does the free trial credit last?
The free trial credit of $0.50 (approx. 50,000 characters) is valid for 7 days. It is granted when you sign in with Google. Accounts created with email and password start without this trial credit, so consider using Google sign-in to test the API immediately.
Can I mix different voices in one lesson?
Yes, you can mix voices by making separate API requests for each segment and joining the audio files. Each request uses one of eight presets. This allows you to create diverse lessons with different narrators, such as using a calm voice for instructions and an energetic one for dialogue.
Is the audio generated in real-time?
Generation is roughly real-time on long texts, processing about 25 characters per second. Short phrases may take a few seconds. Requests are processed sequentially, so under load, a request may wait in a queue. It is not instant, but efficient for batch processing.
How do I change my API key if it is compromised?
You can issue a new API key at any time from your account dashboard. When a new key is issued, the old one stops working immediately. This ensures quick response to security concerns without needing to wait for manual support or verification processes.
First request
curl
curl https://api.play93k.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.mp3How TTS enhances language learning
Text-to-speech integration allows language learning applications to provide immediate auditory feedback, reinforcing vocabulary retention through repeated, accurate pronunciation. By converting text directly into audio, learners encounter natural speech patterns that static text cannot convey, bridging the gap between recognition and comprehension.
The API supports 23 languages, enabling diverse learners to access consistent, high-quality audio models. This ensures that educational content remains accessible across different linguistic backgrounds without requiring complex infrastructure or external voice actors for every new lesson.
Contextual audio for vocabulary retention
Integrating TTS into lesson modules allows learners to hear words in context, improving memory retention. The API’s ability to generate audio from text ensures that vocabulary is always pronounced correctly, reducing cognitive load during study sessions.
Learners can replay specific phrases as needed, reinforcing their understanding of phonetics and intonation. This repetitive exposure is critical for mastering new languages, as it allows for self-paced practice without relying on human instructors for every interaction.
- Immediate auditory feedback for vocabulary items
- Support for repetitive listening exercises
- Consistent pronunciation across all lessons
Pronunciation models for diverse learners
The API offers eight distinct voice presets, each with a unique tone suitable for different learning styles. From warm narrators to neutral news anchors, these voices provide varied auditory experiences that keep learners engaged.
Each voice is based on licensed recordings, ensuring high-quality synthetic speech that sounds natural. This diversity allows educators to tailor the audio experience to the specific needs of their learners, whether they prefer calm instructional tones or energetic promo styles.
- Eight distinct voice presets
- Licensed recordings for high quality
- Varied tones for different learning styles
Consistent voice identity across lessons
Maintaining a consistent voice identity across all lessons helps learners build a familiar auditory environment. Whether using Aria for storytelling or Atlas for news-based learning, the same voice can be used throughout a course, enhancing continuity.
This consistency is crucial for language acquisition, as it reduces the cognitive effort required to adapt to new voices. Learners can focus on the content rather than adjusting to different vocal characteristics, leading to a more effective learning experience.
- Unified voice identity across courses
- Reduced cognitive load for learners
- Enhanced continuity in learning modules
Technical specifications for product teams
Product teams can integrate LinguaVoice using a standard RESTful endpoint compatible with existing OpenAI SDKs. By changing the base URL to https://api.play93k.com/v1, developers can use official OpenAI clients with minimal code changes, leveraging the same request shape for seamless adoption.
Each synthesis request accepts a JSON body containing the text, voice selection, and output format. The API returns audio files directly, allowing immediate playback or storage in your application's media library without additional processing steps.
RESTful endpoint compatible with OpenAI SDKs
Integration requires only a POST request to https://api.play93k.com/v1/audio/speech. The endpoint accepts standard JSON payloads, enabling developers to use familiar OpenAI SDKs by simply adjusting the base URL configuration.
- Standard JSON request body
- Compatible with official OpenAI SDKs
- Base URL: https://api.play93k.com/v1
Support for MP3, WAV, OPUS, FLAC, and PCM formats
The API supports MP3, WAV, OPUS, FLAC, and PCM output formats via the response_format field. This flexibility allows teams to optimize file sizes for mobile data usage or select uncompressed formats for high-fidelity educational content.
- MP3, WAV, OPUS, FLAC, PCM
- Optimize for mobile or high fidelity
Speed control from 0.25x to 4.0x for pacing
Adjustable speed between 0.25x and 4.0x allows educators to tailor audio pacing for diverse learners. Slower speeds aid pronunciation practice, while faster rates support advanced listening exercises, ensuring the audio matches the lesson's difficulty level.
- Speed range: 0.25–4.0
- Tailor pacing for learners
Available voices for educational content
LinguaVoice provides eight voice presets derived from licensed LibriTTS-R (CC BY 4.0) recordings and LibriVox public-domain audiobook readings. These voices are designed for educational consistency, allowing product teams to maintain a recognizable audio identity across diverse language learning modules without the overhead of custom voice design.
Each generated file includes an inaudible watermark identifying it as synthetic speech. The voices support all 23 available languages, ensuring that narration, news segments, and interactive prompts remain linguistically accurate regardless of the voice preset selected.
Aria: Warm narrator for storytelling lessons
Aria offers a warm, clear tone ideal for immersive storytelling. Her delivery is well-suited for vocabulary retention exercises where narrative context aids memory. As a female preset, she provides a gentle, engaging baseline for beginner to intermediate learners.
Use Aria for reading passages, dialogues, and contextual examples where a consistent, approachable voice helps sustain student attention over longer sessions.
- Warm, clear narrator
- Ideal for storytelling
- Supports all 23 languages
Atlas: Neutral tone for news-based learning
Atlas delivers a neutral, professional tone optimized for news-based learning modules. His voice minimizes emotional bias, making him suitable for factual reporting, current events, and academic content where clarity and neutrality are prioritized over expressive flair.
He serves as a reliable default for instructional materials where the focus remains on the linguistic content rather than the speaker's personality.
- Neutral, professional tone
- Optimized for news
- Clear and unbiased delivery
Felix: Energetic voice for promotional content
Felix brings an energetic, upbeat quality to promotional content and call-to-action segments. His dynamic delivery captures attention effectively in short-form audio, such as quiz introductions, achievement notifications, or feature highlights within the learning app.
His male preset style adds variety to the audio landscape, preventing listener fatigue in apps that mix instructional and promotional audio.
- Energetic, upbeat style
- Great for promotions
- High engagement potential
Language support and coverage
Language learning applications require precise audio generation across diverse linguistic contexts. Our API supports 23 languages, enabling developers to create immersive vocabulary and pronunciation exercises without managing complex infrastructure for each dialect.
23 languages including English, Spanish, and Chinese
We support 23 languages, including English, Spanish, Chinese, Arabic, Japanese, and German. Every voice can speak any of these languages, allowing flexible casting for educational content.
- English, Spanish, Chinese, Arabic, French, German, Japanese, Korean, Hindi, Portuguese, Russian, Italian, Turkish, Dutch, Swedish, Danish, Norwegian, Finnish, Polish, Greek, Hebrew, Swahili, Malay
Automatic language detection via script
The API guesses the language from the input script if not explicitly set. Latin-script text defaults to English, while specific scripts trigger the corresponding language.
Explicit language setting for Latin-script ambiguity
For non-English Latin-script text, you must set the language field to ensure accurate pronunciation. This prevents the default English bias from affecting words like 'café' or 'naïve'.
- Unsupported: Vietnamese, Indonesian, Thai, Ukrainian, Romanian, Czech, Bengali
Integration workflow
Integration begins with authenticating requests using a Bearer token generated from your account. This single key authorizes all endpoints and can be rotated instantly via the dashboard, ensuring secure access for your language learning platform's backend services.
Authentication with Bearer token keys
Include Authorization: Bearer tts-... in every HTTP header. Use the key issued at /get-api-key/. Rotate keys immediately if compromised; the old key stops working instantly.
JSON body structure for speech requests
Send a POST request to https://api.play93k.com/v1/audio/speech. The body requires model ("tts-1" or "tts-1-hd"), input (text up to 4096 characters), voice (e.g., "aria", "atlas"), and response_format (mp3, wav, opus, flac, pcm). Optional fields include speed (0.25–4.0) and language (ISO code).
Use the OpenAI SDK by setting base_url to https://api.play93k.com/v1. This allows direct compatibility with existing client libraries without custom parsing logic.
- Model: "tts-1" or "tts-1-hd"
- Input: String, max 4096 chars
- Voice: One of 8 presets
- Format: mp3, wav, opus, flac, pcm
Handling rate limits and queue times
Limit is 60 requests per minute per key (10/min before first top-up). Requests process sequentially. Long texts are split into 4096-character chunks and joined. Expect queue delays under load. Typical speed is ~25 chars/sec for real-time generation.

Pricing structure for scaling teams
Costs scale linearly with usage, eliminating waste from unused capacity. At $10 per million characters (including spaces and punctuation), teams pay only for what they generate. The free trial provides $0.50 credit for 7 days, while permanent balances are topped up via USDT (TRC-20) with no expiration.
Pay-per-character model at $10 per million characters
Billing is precise: every character in your input text counts toward the total. This structure suits variable lesson volumes, from small vocabulary sets to extensive course libraries, ensuring predictable expenses without overpaying for idle licenses.
- Rate: $0.01 per 1,000 characters
- Minimum top-up: $5 USDT
No monthly subscriptions or seat fees
There are no recurring base fees, no per-seat charges for your team, and no minimum spend requirements. You retain full control over capital allocation, topping up only when necessary. Balances do not expire, allowing flexible budgeting for long-term educational projects.
Failed requests are free of charge
If synthesis fails, the charge is returned in full. This protects your budget from technical glitches or invalid inputs, ensuring you only pay for successfully delivered audio files.
Data privacy and usage rights
LinguaVoice treats your educational content as transactional data rather than training material. When you send text to our API, that input is processed solely to generate the corresponding audio file for your language learning application. We do not retain the semantic content of your lessons to improve our underlying voice synthesis models, ensuring that proprietary curriculum data remains distinct from our training datasets.
Input text is not used for training models
Your lesson text is processed exclusively for the immediate generation of the audio output. It is not ingested into our model training pipeline, preserving the uniqueness of your course material.
Generated audio is returned in response and not stored
The synthesized audio file is delivered directly in the API response and is not persistently stored on our servers after delivery. We only retain request metadata—such as timestamp, character count, and status—for billing and usage verification purposes.
Inaudible watermark marks synthetic speech
Every audio file generated by our API contains an inaudible digital watermark. This marker clearly identifies the output as synthetic speech, which is useful for transparency in educational contexts and helps distinguish AI-generated audio from human-recorded voiceovers.
Performance metrics
The API generates audio at approximately real-time speed for extended texts, producing roughly 25 characters of text per second of audio. This ratio ensures predictable generation times for lesson content, allowing product teams to estimate loading durations accurately without significant overhead for standard vocabulary exercises or reading passages.
Real-time generation speed on long texts
For texts approaching the 4096-character limit, synthesis typically completes in a few seconds for short phrases. Longer inputs scale linearly with duration, maintaining a consistent processing rate that aligns with natural speech pacing, which helps maintain user engagement during interactive language drills.
Queue processing under high load
Requests are processed sequentially. Under heavy load, a request may wait in a queue before synthesis begins, potentially extending the total response time beyond the base generation speed. This linear processing model ensures stability but means concurrent users may experience variable latency depending on current server demand.
4096 character limit per request
Each API call accepts a maximum of 4096 characters in the input field. For lessons exceeding this length, the client must split the text into parts, generate separate audio files, and join them. This constraint keeps individual request processing fast and predictable, simplifying error handling and retry logic for developers.
Why choose this API for education
LinguaVoice removes administrative friction for educational product teams. With billing handled via USDT on the TRON network, top-ups are credited automatically within minutes, eliminating the delays of traditional banking or payment processors. There are no monthly subscriptions, seat fees, or minimum spend requirements to manage.
Simplified billing with USDT top-ups
Payments are made exclusively in USDT (TRC-20) with a $5 minimum. Unlike credit cards or PayPal, this method avoids invoices with VAT. Your balance does not expire, allowing you to scale usage without recurring billing cycles.
- USDT on TRON network only
- Automatic credit within minutes
- No monthly subscriptions or seat fees
- Balance never expires
No enterprise contracts required
You can integrate immediately without signing complex agreements. There is no need for enterprise contracts, SLAs, or uptime guarantees to begin production usage. This flexibility allows language learning platforms to test voice quality and integrate the API directly into their curriculum workflows without legal overhead.
Direct email support for technical issues
When technical questions arise, support is available via email. There is no phone number or live chat to navigate; you receive direct assistance for integration issues. This approach ensures that product teams get specific, documented answers without waiting on hold.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. We must design intuitive learning paths that adapt to each user's unique pace and style for maximum engagement.
23 languages
Any voice speaks any of these languages.
