Frequently asked questions
What is the primary function of a text-to-speech API?
It converts written text into audio files via a cloud API. You send text in a JSON body, and receive an audio response. This enables newsrooms to convert articles and transcripts into audio assets for multi-channel distribution and accessibility compliance without manual recording.
How does the API process input text?
The API processes text via a POST request to the speech endpoint. It accepts up to 4096 characters per request. Processing is sequential, taking roughly real-time duration for long texts (about 25 characters per second). Short phrases may take a few seconds. Short phrases may take a few seconds due to queueing under load.
Are the voices human or synthetic?
The voices are synthetic, generated from licensed recordings of human speakers. Every audio file includes an inaudible watermark marking it as synthetic speech. The eight presets are based on LibriTTS-R and LibriVox public-domain audiobook readings, not direct human voice cloning.
Does the API store the input text?
No. Input text is used only to produce the audio and is not used for training. Generated audio is returned in the response and not stored. Only request metadata, such as time, characters, and status, is kept for billing purposes.
Can the API handle long news articles?
Yes, but you must split text into parts of up to 4096 characters and join the resulting audio files. The API processes requests sequentially, so under load, a request may wait in a queue before synthesis begins.
What is the difference between tts-1 and tts-1-hd?
Both models use the same engine and price. The only difference is the model identifier in the request. You can choose either "tts-1" or "tts-1-hd" without any change in cost or processing behavior.
Is the audio output compressed?
Compression depends on the selected response format. You can choose mp3, which is compressed, or wav, opus, flac, and pcm, which are uncompressed or lossless. The API returns the audio file in the format you specify in the request.
How is the language detected?
If the language is not set, the API guesses it from the script. Latin script text defaults to English. For non-English Latin-script text, you must set the language parameter explicitly. Supported languages include English, Spanish, German, and others.
What happens if the text exceeds 4096 characters?
The API accepts a maximum of 4096 characters per request. You must split longer texts into smaller parts, send each part separately, and then join the resulting audio files on your end to create the complete audio asset.
Can I use the API for live news updates?
The API does not support streaming or WebSocket connections. Requests are processed sequentially, and short phrases take a few seconds. It is better suited for batch processing articles and transcripts rather than live, instant audio generation for breaking news.
Is the API available via WebSocket?
No. The API uses standard HTTP POST requests. There is no WebSocket support, no streaming, and no real-time partial audio delivery. You send a request and receive the complete audio file in the response.
Does the API support phoneme customization?
No. The API does not support SSML, phoneme dictionaries, or custom pronunciation lexicons. You can adjust speed and select from eight voice presets, but you cannot customize individual phoneme pronunciations or add emotion tags.
How are voices licensed?
Voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. The eight presets include four female and four male voices, such as Aria, Nova, Felix, and Atlas, trained on these sources.
Is there a limit to the number of requests?
Yes. Each API key is limited to 60 requests per minute. Before the first top-up, the limit is 10 requests per minute. Failed requests are free: if synthesis fails, the charge is returned in full.
Definition and Core Function
Text-to-speech synthesis converts written text into audible speech. For newsrooms, this API transforms articles and transcripts into audio assets, supporting accessibility compliance and multi-channel distribution strategies without manual recording.
Text-to-speech synthesis definition
The process generates synthetic speech from input text using predefined voice models. It eliminates the need for human narrators to produce standard news content, allowing product teams to scale audio production efficiently.
API-driven audio generation
Teams send text via a JSON payload to a cloud endpoint and receive an audio file. The system handles voice selection, language detection, and format encoding, delivering ready-to-use audio assets for digital platforms.
- Input: JSON body with text, voice, and format parameters.
- Output: Standard audio files (mp3, wav, opus, flac, pcm).
Cloud-based processing model
Processing occurs on remote servers, requiring only an internet connection and a valid API key. Input text is used solely for generation and is not retained for training, ensuring privacy while enabling on-demand audio creation.
Technical Mechanism
The API accepts plain text input up to 4,096 characters per request, treating spaces and punctuation as countable units. Longer articles are split into manageable segments for synthesis.
Users select from eight voice presets, such as 'atlas' for neutral news delivery or 'sage' for instructional content. These voices support 23 languages, with language detection defaulting to English for Latin script if unspecified. If the text is in a different language, you must set the language parameter explicitly.
Input text processing
The system processes up to 4,096 characters per request, counting spaces and punctuation. Longer articles are automatically split into segments, ensuring precise character limits while maintaining audio continuity.
Users select from eight voice presets, such as 'atlas' for neutral news delivery or 'sage' for instructional content. These voices support 23 languages, with language detection defaulting to English for Latin script if unspecified.
- Max 4,096 characters per request
- Spaces and punctuation count toward limits
- Automatic segmentation for longer texts
- 23 supported languages
Voice selection presets
Eight distinct voice profiles cater to specific newsroom needs, from the warm narration of 'aria' to the deep documentary tone of 'orion'. Each voice is derived from licensed recordings, ensuring consistent quality.
Voice selection is passed via the voice field in the JSON payload. The API handles language adaptation, allowing any voice to speak any supported language without additional configuration.
- 8 presets: aria, nova, luna, sage, felix, milo, atlas, orion
- Licensed from LibriTTS-R and LibriVox
- Any voice supports any of 23 languages
- JSON
voicefield for selection
Audio file generation
Synthesis produces audio files in formats including mp3, wav, opus, flac, and pcm. Each output includes an inaudible watermark identifying it as synthetic speech, aiding content verification.
Failed requests are free: if synthesis fails, the charge is returned in full. Audio files are returned directly in the response and are not stored by the service, minimizing data retention risks.
- Formats: mp3, wav, opus, flac, pcm
- Inaudible synthetic watermark included
- Failed requests are free
- Audio returned in response, not stored
Industry Application
The API enables newsroom product teams to convert article text and broadcast transcripts into professional audio assets efficiently. By sending JSON payloads via POST https://api.playmomoo.com/v1/audio/speech, editors generate consistent voiceovers from written content without manual recording.
This automation supports multi-channel distribution strategies, allowing simultaneous publishing of audio versions alongside text articles. The system handles standard news scripts and interview transcripts, ensuring uniform audio quality across all published materials.
Teams can leverage this capability to maintain brand voice consistency while reducing production overhead for daily content updates.
News article audio conversion ['The API accepts plain text inputs up to 4096 characters, converting editorial content into audio files. Newsrooms can process entire articles by splitting longer texts into manageable segments, joining the resulting audio files seamlessly.', "This process supports various voice presets like Atlas for neutral news delivery or Felix for energetic promo content, ensuring the audio matches the article's tone.", 'Editors can select formats such as mp3 or opus to optimize bandwidth and compatibility with existing CMS platforms.'] ['Convert articles and transcripts into audio assets', 'Supports standard editorial text workflows', 'Generates audio via JSON API requests']
Accessibility compliance tools ['Generating audio from text provides an accessible format for readers with visual impairments or reading difficulties. This aligns with accessibility standards by offering an alternative consumption method for published content.', 'The synthetic speech includes an inaudible watermark, marking the audio as generated, which aids in transparency. This feature supports compliance requirements without requiring complex infrastructure.', 'Teams can integrate the API directly into their publishing workflow, ensuring every new article has a corresponding audio version automatically.'] ['Supports visual impairment accessibility', 'Inaudible watermarking for transparency', 'Automated audio generation for compliance']
Digital distribution channels ['Audio assets generated by the API can be distributed across websites, podcasts, and social media platforms. This multi-channel approach maximizes the reach of news content without additional recording costs.', "The API's rate limit is 60 requests per minute (10 per minute before the first top-up). Requests are processed sequentially, so under load a request may wait in a queue.", 'Generated audio files are returned directly in the response, enabling immediate integration into digital distribution pipelines.'] ['Distribute audio across websites and podcasts', 'Supports high-volume batch processing', 'Immediate integration with digital pipelines']
Voice Technology Basis
The synthetic voices are built upon licensed recordings from the LibriTTS-R dataset (CC BY 4.0). These models derive their acoustic characteristics from public-domain audiobook readings sourced from LibriVox, ensuring a legally compliant foundation for commercial newsroom use without requiring additional licensing for the underlying voice samples.
Licensed recording sources
The underlying acoustic models are trained on the LibriTTS-R dataset, which carries a Creative Commons BY 4.0 license. The voice timbres originate from public-domain audiobook readings available via LibriVox. This provides a clear, verifiable chain of custody for the voice assets used in generation, avoiding the ambiguity often associated with AI voice cloning.
- LibriTTS-R dataset (CC BY 4.0)
Synthetic speech characteristics
Each of the eight presets (e.g., Aria, Felix, Atlas) represents a distinct vocal profile optimized for specific tonal requirements, from neutral news reading to energetic promos. The voices do not mimic specific celebrities or real-world individuals but offer consistent, professional presets based on licensed recordings. They support 23 languages natively, allowing news teams to produce localized content from a single API key.
Inaudible watermarking
Every audio file generated through the API contains an inaudible digital watermark. This marks the output as synthetic speech, providing an additional layer of transparency for news organizations. This feature aids in content verification and distinguishes AI-generated audio from human-recorded broadcasts, supporting editorial standards in digital distribution.
Data Handling Process
The API processes input text exclusively to generate the requested audio file. This text is not retained for model training purposes, ensuring that the specific content of your news articles or transcripts remains private to the synthesis operation.
Text input usage scope
Input text is used solely for the immediate conversion into audio. Once the synthesis is complete, the text itself is not stored in long-term training datasets, preserving the integrity of your editorial content.
Audio output retention policy
Generated audio files are returned directly in the HTTP response and are not stored by the API service after delivery. Your team retains full control over the resulting audio assets for distribution or archival.
Metadata storage for billing
Request metadata, including timestamp, character count, and status, is retained strictly for billing and usage tracking. This data ensures accurate charge calculation and provides transparency regarding API consumption without retaining the actual content.
Integration Requirements
Integration requires a valid API key for authentication, sent via the Authorization header. Requests are structured as JSON payloads to the speech endpoint.
Connectivity relies on standard HTTPS protocols. No proprietary SDK is required; standard HTTP libraries suffice for integration into existing newsroom pipelines.
API key necessity
Each account holds one active key. Keys are issued upon account creation or regeneration. The key authenticates all requests to the API base URL.
A new key immediately invalidates the previous one, ensuring secure access control for newsroom development teams.
- Sent in the Authorization header as a Bearer token
- One active key per account
- Regenerate anytime to revoke previous access
JSON request structure
Requests follow the standard OpenAI-compatible JSON format. The body includes model, input text, voice selection, and response format.
Fields include model (tts-1 or tts-1-hd), input (string up to 4096 characters), voice (one of eight presets), and response_format (mp3, wav, opus, flac, pcm).
- Standard JSON body
- Compatible with OpenAI SDKs by changing base_url
- Supports optional
languageandspeedparameters
Network connectivity needs
Standard HTTPS connectivity is required. No special ports or firewalls beyond standard web access are needed.
Requests are processed sequentially. Long texts should be split into chunks under 4096 characters to ensure reliable delivery.
- HTTPS required
- Standard web ports
- Synchronous processing
Performance Characteristics
Performance depends on text length and system load. Short phrases require a few seconds, while longer articles generate at roughly real-time speed.
Generation speed metrics
Synthesis takes a few seconds for short phrases. For longer texts, generation approximates real-time at approximately 25 characters of text per second of audio.
Queue behavior explanation
Requests are processed sequentially. Under load, a request may wait in a queue before synthesis begins.
Real-time approximation
Long texts generate audio at a rate comparable to playback speed. This allows newsrooms to plan ingestion pipelines that keep pace with editorial production without significant backlog.

Limitations and Constraints
The API imposes a hard limit of 4,096 characters per request. For longer news articles or transcripts, your integration must split the text into chunks, generate audio for each segment, and join the resulting files. This ensures consistent processing but adds a step to your pipeline.
Unlike proprietary consumer apps, this service does not stream partial audio chunks. The entire audio file is returned only after synthesis completes. This synchronous model simplifies integration but requires clients to handle potential latency during generation.
Character limit per request
Each POST request accepts a maximum of 4,096 characters. Text exceeding this threshold must be divided by the client application. The system does not automatically concatenate segments; your code must manage the assembly of the final audio asset from multiple outputs.
- Maximum input size: 4,096 characters
- Requires client-side splitting for longer articles
- Segments must be joined manually
No streaming support
The API returns a complete audio file upon success. There is no WebSocket endpoint or HTTP streaming for partial audio delivery. Clients must wait for the full response, which may take several seconds for longer texts due to sequential processing.
No SSML capability
The engine does not support Speech Synthesis Markup Language (SSML). You cannot control prosody, pause durations, or phoneme pronunciation via markup. Voice selection is limited to the eight available presets without fine-grained tonal adjustments.
Security and Privacy
Input text is utilized exclusively for audio synthesis and is not repurposed for model training. Generated audio files are returned in the response and are not stored on our servers, ensuring transient data handling.
Data usage for training
Your input text serves only to generate the specific audio file requested. It is not used to improve or train the underlying TTS engine, preserving content confidentiality for sensitive news articles.
Key security practices
Each account holds one active API key. If you issue a new key, the previous one becomes invalid immediately, allowing instant revocation if compromised. This ensures tight control over access.
- Single active key per account
- Instant invalidation upon key rotation
Request encryption standards
All API requests are transmitted over encrypted connections. Request metadata, including timestamp, character count, and status, is retained strictly for billing purposes and remains separate from the audio content.
Scalability for Newsrooms
Newsrooms require consistent audio output during high-traffic periods. The API processes requests sequentially, meaning generation speed approximates real-time performance of approximately 25 characters per second. This linear processing ensures predictable output without complex infrastructure scaling.
Volume handling capacity
The system handles bulk article conversions by splitting texts exceeding 4,096 characters into manageable segments. Each segment is synthesized independently and joined by the client application.
Concurrent request limits
Each API key supports a limit of 60 requests per minute. During the initial phase, this limit is 10 requests per minute until the first top-up is processed. Requests are queued sequentially under load, ensuring no data loss but potentially increasing total processing time for large batches.
Cost scaling efficiency
Pricing scales linearly with usage at $10 per million characters. There are no minimum spend requirements or monthly seat fees. Failed requests incur no charge, ensuring that only successful audio generations contribute to your operational costs.
Comparison to Manual Methods
Manual audio production for newsrooms typically requires hiring voice talent, booking studio time, and managing post-production edits for each article. This process introduces significant latency and variable costs, making it difficult to keep up with breaking news cycles or update existing content quickly.
The API shifts this workflow from a project-based model to an on-demand utility. By sending text via HTTP requests, teams eliminate the logistical overhead of scheduling and editing, enabling immediate audio asset generation directly from the content management system.
Efficiency gains
Generating audio via API takes seconds for short phrases, compared to hours or days for traditional recording and editing workflows. This speed allows newsrooms to publish audio versions of articles almost simultaneously with the text, meeting modern reader expectations for immediate accessibility.
Longer articles can be processed by splitting them into 4,096-character segments and joining the resulting audio files. This parallelization significantly reduces the total time to produce a full article’s audio asset compared to sequential human recording and editing processes.
- Eliminate studio booking and talent coordination.
- Generate audio in seconds, not days.
- Instant updates without re-recording entire segments.
Consistency improvements
Human narration varies in tone, pace, and quality across different sessions and readers. The API ensures a uniform audio style for all content, maintaining brand integrity across thousands of articles. Each voice preset delivers a consistent auditory experience, regardless of which article is being processed.
Unlike human readers who may mispronounce technical terms or vary their delivery, the synthetic voices provide predictable pronunciation and pacing. This consistency is crucial for accessibility compliance, ensuring that screen reader users experience a standardized audio format across the entire site.
- Uniform tone across all articles.
- Predictable pronunciation and pacing.
- No variability between recording sessions.
Resource optimization
Newsrooms no longer need to maintain dedicated audio production teams or expensive studio infrastructure for every new article. The API model converts fixed overhead into a variable cost of $0.01 per 1,000 characters, allowing budgets to scale directly with content volume.
Staff can focus on journalism and product development rather than audio logistics. The API handles the heavy lifting of synthesis, allowing small teams to produce high volumes of audio content without proportional increases in headcount or operational complexity.
- Convert fixed overhead to variable costs.
- No need for dedicated audio teams.
- Scale production without hiring.

Future Compatibility
The API leverages standard OpenAI-compatible request structures, ensuring immediate compatibility with existing OpenAI SDKs by simply adjusting the base URL. This allows newsroom engineering teams to integrate TTS capabilities into their current workflows without requiring custom client libraries or significant refactoring of their codebases.
While no proprietary SDK is offered, the consistent JSON interface supports rapid adaptation as standard libraries evolve. This approach ensures that future SDK updates from third-party providers remain directly usable, maintaining long-term integration stability for automated news pipelines.
SDK evolution support
Current integration relies on the standard OpenAI SDK interface. By using the official SDKs with a modified base_url, newsrooms can adopt new SDK versions as they are released without waiting for a specific vendor update. This reduces technical debt and ensures that standard authentication and request patterns remain valid as the broader ecosystem evolves.
Format updates
The API currently supports mp3, wav, opus, flac, and pcm formats. If new audio codecs become industry standards for digital distribution, the endpoint can be extended to include them. The JSON request structure allows for easy addition of new response_format options without breaking existing integrations.
Model enhancements
The system currently offers two models: tts-1 and tts-1-hd. Both use the same underlying engine and pricing structure. Future model iterations would likely follow this dual-tier approach, potentially offering improved fidelity or lower latency while maintaining the existing request shape. Existing integrations will remain compatible unless the fundamental JSON structure changes.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. News outlets will amplify your launch, turning product updates into public conversation for your entire audience.
23 languages
Any voice speaks any of these languages.
