Description
About Speech Synthesis Test
Introduction
The DeviceHub Speech Synthesis Test is a free online Web Speech API voice lister and sample speaker, without installing third-party TTS apps for a quick check, without claiming every OS ships the same voice pack, and without presenting speech output as a medical or clinical diagnosis of hearing, speech, or language ability. Engineers verifying speechSynthesis.getVoices(), QA confirming a Speak button works after a user gesture, educators teaching utterance rate and pitch, support desks proving a browser has zero installed voices, and inclusive-design teams spot-checking whether TTS is available before building read-aloud features all open one HTTPS page, wait for voices, and click Speak. DeviceHub states limits plainly: speechSynthesis voices differ by OS; speaking may need a user gesture; prefers-* tools reflect OS/browser settings, not all AT; this is not a full accessibility audit; and it never replaces WAVE, axe, screen-reader testing with NVDA/JAWS/VoiceOver, or professional audits. Permissions stay none in DeviceHub’s typed union, engine audio playback still uses your speakers. Never claim clinical outcomes from sample TTS. Web Speech synthesis lets pages speak text through OS-installed voices, which is useful for quick read-aloud experiments and capability checks, yet it is not a screen reader, not NVDA/JAWS/VoiceOver certification, and not a medical evaluation of hearing or speech. DeviceHub Speech Synthesis Test lists voices when the engine provides them and speaks sample text after you click Speak, emphasizing that voices differ by OS and that browsers often require a user gesture before audio starts. Product teams building lightweight read-aloud, QA verifying empty voice lists on locked-down images, educators teaching SpeechSynthesisUtterance, and support desks proving Safari versus Chrome pack differences all benefit from one HTTPS page with honest limits. This category never replaces WAVE, axe, or professional audits; prefers-* reflect OS/browser settings, not all AT; contrast math tools are not crawlers; text-spacing is a preview; forced-colors varies by platform; and DeviceHub tools are informational, not clinical diagnoses. Pair with Accessibility Preferences, Keyboard Focus Test, and Text Spacing Test as adjacent education. Permissions stay none in the typed union.
What this tool does
Speech Synthesis Test enumerates available voices when the engine provides them, lets you select one, and speaks sample text via speechSynthesis after you trigger Speak. It does not install system voices, does not upload utterance text for advertising when you use the page, and does not evaluate pronunciation disorders clinically. Guidance covers voiceschanged timing, gesture requirements, and OS differences (Windows vs macOS vs mobile). Typed permissions remain none. Closing the tab should cancel ongoing speech where the engine allows. Informational only, not clinical. The UI enumerates voices, exposes language tags when present, and speaks sample or user-entered text via speechSynthesis. It explains voiceschanged races where early getVoices() returns empty, gesture gates, and cancellation via cancel(). It does not install neural voice packs, does not grade pronunciation clinically, and does not upload utterances for advertising when you use the page. Guidance separates synthesis from speech recognition (listening APIs are out of scope here). Typed permissions remain none, your speakers still play audio locally. Closing the tab should stop speech where engines allow. Informational capability demo, not audiology.
When to use it
Use when building read-aloud features, when voice lists are empty in privacy browsers, when teaching Web Speech API limits, when QA needs gesture-related repro steps, and when comparing Chrome versus Safari voice packs. Prefer real screen-reader testing for AT compatibility. Prefer Accessibility Preferences for visual prefers-* hubs. Avoid medical/speech-therapy claims. Avoid expecting DeviceHub to download premium neural voices. Use when prototyping read-aloud, when enterprise images lack voices, when documenting gesture requirements for Speak buttons, when comparing mobile silent switches and desktop mute, and when teaching Web Speech limits versus real AT. Prefer actual screen-reader testing for name/role/value. Prefer Color Contrast Checker and Keyboard Focus Test for visual/keyboard work. Avoid clinical speech-therapy claims. Avoid expecting DeviceHub to fetch premium cloud TTS voices.
How it works
The Web Speech API’s speechSynthesis interface exposes getVoices() and speak(SpeechSynthesisUtterance). Many browsers populate voices asynchronously and require a user gesture before audio. DeviceHub wires those patterns educationally. Voice quality and language coverage vary by OS pack. Typed permissions stay none, speaker hardware is still required on your side. Not a substitute for NVDA/JAWS/VoiceOver audits. Not WAVE/axe for document structure. Not clinical assessment. Local playback; minimize pasting sensitive text into the sample box on shared screens. speechSynthesis.getVoices() returns SpeechSynthesisVoice objects; speak() queues utterances. Many engines populate voices asynchronously and block audible speak() until a user gesture. DeviceHub mirrors those constraints educationally. Rate, pitch, and volume are hints the OS may clamp. Typed permissions stay none. Secure HTTPS applies. OS voice packs differ radically, empty lists are valid evidence. This does not prove your ARIA live regions work. Not WAVE/axe. Not clinical hearing tests. Avoid pasting secrets into the text box on shared screens; local playback can still be overheard. Enterprise environments sometimes strip optional voice packs to reduce image size; an empty list is then correct capability evidence for that browser on that image, not a DeviceHub outage. Conversely, cloud-managed Chromebooks may ship a different voice set than personal Windows laptops, capture device class in bugs. Rate and pitch sliders, when present, are hints; some OS voices ignore extremes. Cancelling speech mid-utterance matters for overlapping announcements in real apps, practice cancel() expectations here, then still verify your production live regions with screen readers. DeviceHub will not claim Web Speech equals WCAG “compatible AT” coverage. Prefer multilingual sample strings when you ship localized products, knowing missing language voices return fallbacks. Keep preferring professional audits for structure. Keep stating results are not clinical audiology or speech-language pathology findings.
Step-by-step instructions
- Open Speech Synthesis Test over HTTPS and wait for the voice list (listen for voiceschanged if needed).
- Select a voice and language if multiple appear.
- Click Speak after a user gesture; raise system volume if silent.
- Note empty lists as OS/browser capability evidence, not necessarily a DeviceHub defect.
- For page structure and names, still run WAVE/axe or professional AT testing, TTS sample ≠ full audit.
- Label results as Web Speech capability checks, not medical or clinical diagnoses.
Common problems
No gesture before speak. Voices empty until voiceschanged. Assuming Safari matches Chrome packs. Treating TTS as screen-reader certification. Clinical claims about hearing. Pasting secrets into sample text. Expecting DeviceHub to fix OS voice downloads. Confusing synthesis with speech recognition. Replacing professional audits. Mobile silent switch on iOS. Background tab suspension stopping speech. No user gesture. Empty voices until voiceschanged. Assuming cross-OS voice parity. Treating TTS as SR certification. Clinical hearing claims. Secrets in sample text. Expecting DeviceHub to install voices. Confusing synthesis with recognition. Skipping professional audits. iOS silent switch. Background tab suspension. Enterprise policies stripping voice packs. Filing medical forms from a Speak demo. Replacing axe with TTS screenshots. Autoplay policies blocking speak without gesture. Headphones unplugged mid-demo. Assuming Android Chrome matches iOS Safari voices. Using TTS as the only “accessibility test” on a PR. Pasting customer PII into the box. Expecting DeviceHub to download Azure or Google cloud voices. Confusing synthesis with dictation/recognition permissions. Claiming hearing diagnoses. Skipping keyboard and contrast checks entirely because “we have Speak.”
Privacy explanation
Speech Synthesis Test speaks text you enter or sample strings locally through the OS voice engine. DeviceHub does not need to upload utterances for a basic voice list check. Avoid dictating passwords or personal health details into the box on shared displays. Permissions stay none in the typed union. Informational, not medical or clinical records. Cancel speech and clear text after classroom demos. Utterances play locally through OS voices. DeviceHub does not need to upload text for a basic voice listing. Avoid speaking passwords, health details, or personal data on shared displays. Permissions stay none in the typed union. Informational, not medical or clinical records. Cancel speech and clear fields after demos. Overheard speech on shared floors is a social privacy risk even when DeviceHub uploads nothing, use headphones for sensitive demos. Informational Web Speech checks only, not medical records. Clear text fields after class.