Frequently asked questions
What is SpeechSynthesisUtterance?
SpeechSynthesisUtterance is a browser API that converts text to speech locally using the user’s device resources. It requires no network connection and is ideal for simple, short audio feedback. However, it lacks the consistent quality and specific voice presets available in cloud-based solutions, making it less suitable for branded e-commerce experiences.
Is SpeechSynthesisUtterance free?
Yes, the browser API is free and built into modern web browsers. It incurs no API costs or character limits. Unlike our cloud API, which charges $0.01 per 1,000 characters, browser TTS uses device processing power, making it cost-effective for low-volume, non-critical audio notifications.
Does SpeechSynthesisUtterance support SSML?
No, the standard SpeechSynthesisUtterance interface does not support SSML markup. It only accepts plain text input. If you need precise control over pauses, pitch, or emphasis, you must rely on cloud TTS APIs that offer more granular audio generation controls beyond basic text input. However, our API does not support SSML.
Can I change the voice in SpeechSynthesisUtterance?
You can select from voices installed on the user’s device, but availability varies by browser and operating system. You cannot guarantee a specific voice like Aria or Felix. Cloud TTS offers 8 consistent presets (e.g., Nova, Atlas) that sound identical across all devices and browsers.
Why is cloud TTS more consistent?
Cloud TTS uses dedicated servers to generate audio, ensuring the same voice quality and tone regardless of the user’s device. Browser TTS depends on local engine implementations, which can vary significantly between Chrome, Safari, and Firefox, leading to inconsistent brand audio experiences for your clients.
Does cloud TTS support more languages?
Our cloud API supports 23 languages, including Arabic, Chinese, Japanese, and Swahili. Browser support varies widely; some devices lack voices for less common languages. For global e-commerce, cloud TTS ensures consistent language support across all user devices without relying on local OS language packs.
Is cloud TTS slower than browser TTS?
Cloud TTS generates audio at roughly real-time speed (25 characters per second). Short phrases may take a few seconds due to network latency and queueing. Browser TTS is often faster for short text because it processes locally, but cloud TTS offers higher fidelity and consistency.
Can I use both APIs together?
Yes, you can combine both. Use SpeechSynthesisUtterance for immediate, low-priority UI feedback like button clicks. Use cloud TTS for critical content like product descriptions or customer support messages where voice consistency and quality matter. This hybrid approach balances performance with brand consistency.
Which is better for e-commerce?
Cloud TTS is better for e-commerce due to consistent voice quality across all devices. Browser TTS varies by OS, risking inconsistent brand experience. Cloud TTS also supports 23 languages and specific voices like Atlas for news or Luna for storytelling, enhancing customer engagement.
Does SpeechSynthesisUtterance work offline?
Yes, SpeechSynthesisUtterance works offline if the device has the necessary voice data installed. Cloud TTS requires an internet connection to send text to the API and receive the audio file. For offline product demos or kiosk modes, browser TTS is the only viable option.
Can I clone voices in SpeechSynthesisUtterance?
No, SpeechSynthesisUtterance does not support voice cloning. It only uses voices pre-installed on the device. Our cloud API also does not offer custom voice cloning for customers, but it provides 8 high-quality presets based on licensed recordings, ensuring a professional and consistent audio experience.
How do I detect the language in browser TTS?
Browser TTS language detection is automatic but unreliable. Cloud TTS allows you to explicitly set the language ISO code or guess it from the script. This explicit control ensures the correct voice is used, especially for non-English Latin-script text, avoiding mispronunciation errors.
Is cloud TTS secure?
Input text is used only to produce audio and is not used for training. Generated audio is returned in the response and not stored. Request metadata is kept for billing. We do not offer enterprise contracts or SLAs, but data handling is transparent and minimal. We also do not offer phone support.
Can I store cloud TTS audio?
The API returns the audio file in the response but does not store it on our servers. You are responsible for storing the generated audio files if needed. This allows you to cache frequent product descriptions or marketing copy to reduce API usage and costs.
Defining SpeechSynthesisUtterance
SpeechSynthesisUtterance is a built-in browser API that converts text to audio directly on the user’s device. It requires no external network calls or API keys, making it a zero-cost solution for basic UI notifications or simple accessibility features.
The audio quality and available voices depend entirely on the operating system and browser combination. This means a voice sounding natural on Chrome for Windows may sound robotic or unavailable on Safari for iOS.
Browser-native API
This API leverages the device’s native speech engine, eliminating server latency for short phrases. However, it lacks programmatic control over audio formats or background processing, limiting its use in complex e-commerce workflows.
Dependent on OS
Since the engine is OS-bound, consistency across devices is not guaranteed. Digital agencies must test voice output on multiple platforms to ensure brand audio meets client expectations before deployment.
- Voice availability varies by OS version
- Latency depends on device hardware
- No standard format output (usually WAV)
Limited voice selection
Users are restricted to the voices installed on their specific device. Unlike cloud APIs that offer 8 distinct presets, browser voices are static and often lack the professional polish required for high-stakes product narrations.
Cloud TTS API Overview
Kiriya Editorial’s cloud TTS API processes text server-side, delivering high-fidelity audio files via a standard HTTP endpoint. Unlike browser-based synthesis, this approach offloads the heavy lifting from the user’s device to dedicated infrastructure, ensuring that complex voice models run smoothly regardless of the client’s hardware capabilities.
Server-side processing
Text is sent to our API endpoint and synthesized on our servers. The result is a ready-to-play audio file (mp3, wav, opus, flac, or pcm) returned immediately. This eliminates the variability of local machine resources, ensuring consistent performance even for longer texts up to 4096 characters.
Consistent quality
Because synthesis happens in a controlled cloud environment, the audio output remains identical across all devices. Whether your client accesses the store on a low-end Android phone or a high-end desktop, the voice quality—from the warm narration of Aria to the deep documentary tone of Orion—remains unchanged.
Programmable control
You gain precise control over the audio generation process. Adjust speed between 0.25 and 4.0, select from 8 distinct voice presets, and specify the output format. The API supports standard JSON bodies, making it easy to integrate into your existing e-commerce workflows or custom applications.
Comparison of Voice Quality
Browser speech synthesis relies on the host operating system, meaning voice quality varies significantly between Windows, macOS, and mobile devices. A voice that sounds natural on a desktop may sound robotic or disjointed on a different platform. This inconsistency can undermine brand reliability for agencies delivering uniform client experiences across diverse hardware.
Consistency across devices
Cloud TTS processes audio on servers, ensuring the same output regardless of the user’s device. This eliminates the variability of local OS voices, providing a predictable baseline for all listeners. For digital agencies, this consistency is critical when delivering audio features to online stores with mixed client demographics.
Multiple voice options
The browser API offers a limited, OS-dependent set of voices. In contrast, the cloud API provides eight distinct presets, including warm narrators like Aria or energetic promo voices like Felix. Agencies can select specific tones that align with brand guidelines, rather than accepting whatever the device provides.
- 8 distinct voice presets available
- Consistent tone across all platforms
- Specific character matching for brands
Professional sound
Cloud-generated audio uses licensed recordings from LibriVox, resulting in higher fidelity than standard system voices. The output is clean and clear, suitable for product descriptions or announcements. While browser voices suffice for basic notifications, cloud audio meets the professional standards expected in commercial e-commerce environments.
Language Support Differences
Browser native speech synthesis relies on the user’s operating system and installed language packs. If a client’s device lacks a specific language pack, the API silently falls back to a default voice or fails entirely. This creates inconsistent audio experiences across different devices and regions.
In contrast, our cloud API processes text server-side, ensuring consistent quality regardless of the client’s OS. We support 23 languages, allowing agencies to deploy uniform audio experiences for global online stores without worrying about local voice installations.
Browser language limits
The browser API is strictly bound by the user’s system settings. If a user’s device does not have a German voice pack installed, the API cannot generate German speech, even if the text is in German. This dependency makes cross-regional consistency difficult to guarantee without extensive user-side testing.
- Tied to OS language packs
- Silent fallback to default voice
- Inconsistent across devices
Cloud multi-language support
Our API supports 23 languages, including Arabic, Chinese, Hindi, and Swahili. Every voice preset can speak any of these languages, providing a unified set of voices for diverse customer bases. This eliminates the need for managing multiple voice packs across different operating systems.
Specific language code
To ensure accurate pronunciation, you should explicitly set the language field in your request using the ISO code (e.g., 'fr' for French). Without this, the API guesses the language from the script, which may default to English for Latin text, leading to mispronunciation of non-English phrases.
- Use ISO codes (e.g., 'ja', 'de')
- Prevents default English fallback
- Ensures correct pronunciation
Control and Customization
The browser’s SpeechSynthesisUtterance offers minimal control, relying on the user’s operating system settings. In contrast, our cloud TTS API provides precise, programmatic control over every aspect of the generated audio file, ensuring consistent results across all client devices without manual OS configuration.
SSML support in cloud
Our API does not support SSML. This is a missing feature. While you cannot use XML markup for fine-grained pronunciation or emphasis, the API’s simplicity avoids the complexity of parsing SSML documents for basic voice-enabled store features.
Speed adjustment
You can adjust the speaking rate between 0.25 and 4.0 times normal speed. This allows digital agencies to tailor the pacing for different content types, from slow, clear instructions to rapid promotional updates, without re-recording.
Format selection
Choose from mp3, wav, opus, flac, or pcm formats. This flexibility lets you optimize for web performance (mp3/opus) or high-fidelity requirements (wav/flac), ensuring the audio files fit your e-commerce platform’s specific technical constraints.
Performance and Speed
Browser speech synthesis feels instant because processing happens locally on the user’s device. However, the actual audio playback speed is fixed by the operating system’s voice engine, offering no programmatic control over rate or pitch without complex workarounds. This creates inconsistent user experiences across different browsers and devices.
In contrast, our cloud API processes text on the server. Generation typically takes a few seconds for short phrases, scaling to roughly real-time for longer texts (about 25 characters per second). While you cannot stream partial audio, the consistency in audio quality is guaranteed regardless of the client’s hardware. Note that streaming is not supported.
Browser latency
Latency is minimal since the browser generates audio locally. However, the inability to adjust speed dynamically means you cannot optimize for faster consumption or accessibility needs without reloading the voice settings. This fixed pace can feel sluggish for quick UI notifications or lengthy read-aloud features.
- Immediate start time
- Fixed playback speed
- Dependent on device CPU
Cloud generation time
Requests are processed sequentially. Short phrases take a few seconds, while longer texts generate at approximately real-time speed (about 25 characters per second). For texts exceeding 4096 characters, you must split the input and join the resulting audio files, adding slight overhead to the total generation time.
- Real-time generation speed
- 4096 character limit per request
- No streaming support
Queue handling
Under heavy load, requests may wait in a queue since processing is sequential. Your account is limited to 60 requests per minute (or 10 before the first top-up). If you exceed this limit, subsequent requests will fail, so your implementation should handle retries or backoff strategies.
- 60 requests per minute limit
- Sequential processing
- Retry logic required
Use Cases for Each
The browser API serves well for transient interactions, such as reading error messages or confirming a cart update. It requires no network call, ensuring immediate feedback without delaying the user interface or incurring API costs.
Cloud TTS excels for persistent audio assets. Agencies can generate high-quality, consistent voiceovers for product descriptions or promotional banners once, storing the audio files directly in the store’s media library for reliable playback across all devices.
Simple UI notifications
Use the browser’s native engine for brief, context-specific alerts like form validation or button confirmation. Since these messages are short and transient, the slight variance in voice quality is acceptable, and the zero-cost, offline-ready nature keeps the experience lightweight.
This approach avoids network overhead, ensuring the notification plays instantly even on slow connections, which is critical for maintaining user flow during critical checkout steps.
Complex audio products
For longer content like detailed product guides or marketing stories, the cloud API ensures professional consistency. You can generate the audio once and host it, avoiding the variability of different operating systems and browsers that affects the browser API.
The ability to select specific voices like 'Aria' for narration or 'Atlas' for news-style updates allows agencies to match the brand’s tone precisely, delivering a polished auditory experience that matches the visual design.
Multi-language stores
The browser API relies on OS-level language packs, which may vary by device. The cloud API supports 23 languages with dedicated voices, ensuring that a French customer hears a native voice, while a Japanese customer hears a distinct, high-quality Japanese voice.
By explicitly setting the language parameter, you guarantee the correct pronunciation and tone for each market, preventing the awkward mix of voices that can occur when the browser guesses the language incorrectly.

Cost Implications
The browser’s SpeechSynthesis API incurs zero direct costs, making it ideal for low-volume notifications or simple audio cues where budget is the primary constraint. However, quality variance across operating systems can affect user perception without additional engineering effort.
Our cloud API charges $0.01 per 1,000 characters, with failed requests billed at zero. This predictable model scales with actual usage, ensuring you only pay for delivered audio files rather than infrastructure overhead.
There are no monthly subscriptions or minimum spend requirements. Accounts start with a $0.50 trial credit valid for seven days. Additional credits are added via USDT (TRC-20), with a $5 minimum top-up, credited automatically within minutes.
Free browser API
The browser’s SpeechSynthesis API incurs zero direct costs, making it ideal for low-volume notifications or simple audio cues where budget is the primary constraint. However, quality variance across operating systems can affect user perception without additional engineering effort.
Pay-per-use cloud
Our cloud API charges $0.01 per 1,000 characters, with failed requests billed at zero. This predictable model scales with actual usage, ensuring you only pay for delivered audio files rather than infrastructure overhead.
Volume discounts
There are no monthly subscriptions or minimum spend requirements. Accounts start with a $0.50 trial credit valid for seven days. Additional credits are added via USDT (TRC-20), with a $5 minimum top-up, credited automatically within minutes.
- No monthly plans or seats
- No minimum spend required
- Trial credit: $0.50 for 7 days
Technical Implementation
Choosing between browser-native synthesis and a cloud API requires distinct integration paths. Agencies must decide whether to manage client-side audio playback or handle server-side generation and delivery.
The decision impacts how voice assets are stored, cached, and played within the customer’s shopping experience.
JavaScript integration
Web Speech API uses built-in browser objects like SpeechSynthesisUtterance. Implementation requires checking window.speechSynthesis support and handling voice loading asynchronously. It is ideal for simple UI notifications but lacks programmatic control over playback timing or volume per instance.
Cloud TTS requires fetching audio files via JavaScript, then creating Audio elements. This allows precise control over when and how audio plays, enabling complex sequences like multi-step checkout announcements.
- Check
SpeechSynthesisavailability before usage - Handle voice loading delays
- Manage
Audioobject lifecycle for cloud files
HTTP requests
Cloud TTS relies on POST requests to https://api.kiriya.info/v1/audio/speech. The JSON body includes model, input, voice, and response_format. Responses return binary audio data, typically mp3 or wav.
Clients must handle rate limits (60 requests/minute) and implement retry logic for failed synthesis. Text exceeding 4096 characters requires splitting logic before sending requests.
- POST to
/v1/audio/speech - Include
Authorization: Bearerheader - Parse binary audio response
SDK compatibility
The cloud API mimics the OpenAI speech endpoint structure. Agencies can use official OpenAI SDKs by simply changing the base_url to https://api.kiriya.info/v1 and providing the Kiriya API key.
This allows reuse of existing OpenAI integration patterns without writing custom HTTP parsers for audio handling, streamlining development for teams already familiar with OpenAI’s ecosystem.
bullets
Data Privacy Considerations
Understanding where data lives is critical for digital agencies managing client voice features. The choice between browser and cloud synthesis hinges on privacy requirements, from local processing to cloud storage policies.
Agencies must verify how input text and generated audio are handled to ensure compliance with client data agreements and internal security standards.
Local processing
The browser Web Speech API processes text directly on the user’s device. This means input text never leaves the client’s browser, offering maximum privacy for sensitive UI notifications or simple reads.
Since no network request is made for synthesis, there is zero risk of server-side data retention or third-party exposure during playback.
Cloud data usage
Our API receives input text to generate audio files. Input text is used solely for synthesis and is not used for training. Generated audio is returned in the response and not stored.
Request metadata, including time, character count, and status, is retained strictly for billing purposes. This ensures transparent accounting without preserving the actual content of your synthesized speech.
Training data policies
Unlike some free browser engines that may use usage data to improve models, our API does not use input text for training. This guarantees that client content remains proprietary and is not ingested into public model datasets.
This separation ensures that your audio outputs are distinct from any general-purpose training data, maintaining the exclusivity of your voice assets.
When to Choose Cloud TTS
When building voice features for online stores, consistency defines professionalism. Browser voices vary by device, operating system, and browser version, creating a disjointed user experience. A cloud API ensures the same voice, tone, and quality whether the customer is on iOS, Android, or desktop.
This reliability is critical for brand identity. Agencies can guarantee that promotional audio matches the client’s brand guidelines precisely, regardless of the end-user's hardware. This control eliminates the variability inherent in local system voices, ensuring a cohesive auditory experience across all touchpoints.
Brand consistency
Cloud TTS delivers uniform audio output across all devices. Unlike browser-native voices which change based on the user’s OS and browser, a cloud API ensures the same voice profile is used everywhere. This predictability allows agencies to maintain a strict auditory brand identity.
For digital agencies, this consistency reduces support queries related to audio quality and ensures marketing messages are delivered with the intended tone. It eliminates the guesswork of which voice a client will hear, allowing for precise creative direction.
- Uniform voice output across iOS, Android, and desktop.
- Predictable audio branding for consistent customer experience.
- Eliminates device-specific voice variations.
High volume needs
For stores processing thousands of product descriptions or notifications daily, the cloud API offers scalable reliability. While browser APIs are limited by the client’s device, cloud services handle high throughput with predictable costs. The API supports 60 requests per minute per key, with queue handling for bursts.
Billing is straightforward: $0.01 per 1,000 characters. Failed requests are free, protecting your budget. This pay-per-use model suits agencies managing multiple clients with varying volumes, without the overhead of licensing multiple software seats or managing complex infrastructure.
- Pay-per-use model: $0.01 per 1,000 characters.
- Failed requests are free.
- Handles bursts via queueing; 60 req/min per key.
Professional audio
Cloud TTS provides eight distinct voice presets, from warm narrators to energetic promo voices. Each voice supports 23 languages, ensuring global stores can maintain high-quality audio in their target markets. The audio files include an inaudible watermark, identifying them as synthetic.
Formats like MP3, WAV, FLAC, and Opus allow for optimized delivery. Unlike browser speech which is often compressed and lower fidelity, cloud-generated audio offers studio-grade clarity. This professionalism enhances the perceived value of the online store, especially for premium product descriptions.
- 8 distinct voice presets for various tones.
- Support for 23 languages with accurate pronunciation.
- Multiple output formats (MP3, WAV, FLAC, Opus).
When to Choose Browser TTS
Browser-native speech synthesis remains the pragmatic choice for lightweight, ephemeral interactions where absolute audio fidelity is secondary to speed and simplicity. It eliminates infrastructure overhead, making it ideal for brief UI cues or accessibility announcements that do not require brand-specific voice consistency.
The primary advantage lies in its zero-cost structure. Since the API is built into modern operating systems, there are no per-character fees, no monthly subscriptions, and no API key management required. This makes it economically viable for low-volume, sporadic usage patterns.
Simple use cases
For digital agencies building basic online stores, browser TTS handles straightforward tasks like reading out-of-stock notifications or guiding users through a simple form. The API requires minimal JavaScript integration and works seamlessly with standard DOM events.
It is particularly effective for accessibility features such as screen reader enhancements or reading short, static labels. Since the text is processed locally, there is no network latency, ensuring immediate playback for brief strings of text.
- Brief UI notifications
- Simple form guidance
- Static label reading
- Basic accessibility enhancement
Low budget
Unlike cloud APIs that charge $0.01 per 1,000 characters, browser TTS incurs no direct costs. This is crucial for agencies managing tight margins on small projects or prototypes where usage volume is unpredictable.
There are no API keys to track, no rate limits to monitor, and no balance checks required. The absence of billing infrastructure reduces both development time and ongoing operational complexity for low-stakes audio features.
- Zero per-character cost
- No API key management
- No rate limits
- No billing overhead
Offline capability
Since the speech engine relies on the user’s operating system, audio generation continues even without an internet connection. This ensures that critical store features remain functional during network outages or in remote retail environments.
This reliability is valuable for kiosks or mobile apps where connectivity might be unstable. The browser API leverages existing OS-level voice packs, ensuring consistent behavior regardless of the client’s internet service provider.
- Works without internet
- No network latency
- Reliable in poor connectivity
- Leverages OS voice packs
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. The following examples illustrate how to implement each approach using standard tools available to modern web developers.
23 languages
Any voice speaks any of these languages.
