Skip to main content
Documentation menu
On this page

Documentation

Every tool on this site, explained.

81 tools across six categories, each with the steps to run it, what its results actually mean, and an honest note on what a browser cannot measure. No sign-in, nothing to install.

  • 81 tools documented
  • 6 categories
  • 8 need a permission

Introduction

Real Test Keyboard is a set of browser tools for checking hardware you already own — keyboards, screens, speakers and phones — plus GPU and rendering benchmarks, a row of everyday calculators and a handful of short games. This documentation covers all 81 of them.

Each tool page follows the same shape: what the tool is, how to run it, what the result means, what it can and cannot detect, and the problems people actually hit. Where a tool reports a number, the page says what that number includes — a browser measures through several layers, and pretending otherwise would make the figures look more precise than they are.

Tip

Press / to jump straight to a search box. On this page it focuses the hero search above, which looks through all 81 tools by name, category or what they measure; on a category or tool page it focuses the menu filter in the sidebar.

Browse by category

Six categories, mirroring the site's own navigation. Each one lists every documented tool it contains.

Quick start

The shortest path from "something feels wrong" to a result you can act on.

  1. Start with the broadest test for the part that is misbehavingKeys not typing? Open the keyboard test on the home page and press every key once. Marks on the screen? Start with the dead pixel test. Silence? Speaker test. The broad test tells you whether there is a fault at all before you go looking for its cause.
  2. Give the test what it needsClick once inside the tool so it has keyboard focus, go fullscreen for anything visual, and allow the permission prompt if one appears. A surprising share of "it does not work" is one of those three.
  3. Read the result against the page, not against a memoryEvery documentation page has a What the results mean section written for that specific tool, including the thresholds it uses. That is the difference between a number and a diagnosis.
  4. Narrow it down with the focused toolOnce the broad test finds something, the related tools at the bottom of each page are the next step — a suspect key goes to the broken key detector, a suspect panel to the ghosting test, and so on.
  5. Keep the evidenceTools that produce a report offer Copy, Download or Share. A report with real event data attached is worth far more to a support desk or a seller than a description of the symptom.

How a doc page reads

Every tool page runs in the same order, so once you have read one you know where to look on the rest.

DocumentationCategoryToolWhat it isHow to use itWhat the result meansTroubleshootOpen tool

What the badges mean

A badge is a factual claim about that tool, so it only appears when it is true of that page:

BadgeWhat it tells you
AvailableLive on this site right now, at the route printed on the page.
Browser ToolRuns as a web page — no app, no driver.
No InstallNothing to download, no extension, no account.
Client-SideThe test runs entirely on your device; its data is not uploaded.
Requires PermissionCannot start until your browser grants access — a microphone, or motion sensors on iOS. 8 tools carry this badge.
DesktopNeeds a physical keyboard or mouse to tell you anything useful.
MobileBuilt for phone and tablet hardware, so a desktop will read empty.
Plays SoundProduces audio, so speakers or headphones are part of the test.

Troubleshooting

Five problems that come up across more than one tool. Anything specific to a single tool is on that tool's own page.

A key or click is not registering

Almost always focus, a reserved key, or software sitting in front of the browser.

  • Click once inside the test area first — a page with no keyboard focus receives no key events.
  • Tab, Escape and the function keys belong to the browser. Turn on Capture keys where a tool offers it.
  • Media, fn-layer and macro keys are often handled inside the keyboard's firmware and never reach the operating system, so no browser tool can see them.
  • Close overlays, macro utilities and game launchers: anything that hooks input can consume it before the browser does.

Keyboard documentation

A display test looks wrong

Panel tests are only as honest as what the browser is allowed to draw.

  • Go fullscreen and dim the room. Reflections and desktop chrome are mistaken for panel faults more often than anything else.
  • Browser zoom, display scaling and any GPU-level image processing all change what you see. Reset zoom to 100% first.
  • A frame rate the browser reports is the rate it is being served at, which a compositor or a battery-saver profile can cap below the panel's own refresh rate.
  • Check on a second pattern before concluding: a mark that moves with the pattern is content, a mark that stays put is the panel.

Display documentation

No sound, or the microphone is not detected

Output routing and input permission are two separate problems with two separate fixes.

  • Browsers block audio until you interact with the page. Press the play or start control rather than expecting sound on load.
  • Check the operating system's output device. Connecting a headset often switches the default device without moving the app you were using.
  • Microphone tests need an explicit permission. If you dismissed the prompt, re-allow the site in the address-bar site settings — the prompt does not usually appear twice.
  • A Bluetooth headset can expose a different profile for playback and for recording; the browser sees the profile, not the radio.

Audio documentation

A phone sensor test shows nothing

Three things have to line up: the hardware, the browser API, and permission.

  • Sensor readings require a secure context. On http:// the API is simply not there.
  • iOS asks for motion and orientation access on a tap, and only on a tap — use the button the page provides rather than waiting for a prompt.
  • Desktop machines usually have no gyroscope, accelerometer or magnetometer at all, so an empty reading there is correct, not broken.
  • Vibration is unavailable on iOS Safari and is ignored by some Android browsers in low-power mode.

Mobile documentation

The numbers change between runs

Timing in a browser is measured through several layers, and every one of them adds jitter.

  • Close other tabs and heavy applications, then take the best of several runs rather than the first one.
  • Browsers deliberately coarsen high-resolution timers as a privacy measure, so sub-millisecond figures are approximations by design.
  • A reaction or latency figure is the whole chain — display, input device, operating system, browser and you — not one component of it.
  • Any tool that keeps a history keeps it in this browser's local storage, so a fresh browser or private window starts from zero.

Keyboard documentation

What leaves your device

Worth being precise about, because "runs in your browser" is often said loosely. The tests themselves are computed on your device: keystrokes, audio, sensor readings and calculator inputs are processed in the page and not sent anywhere. Settings, personal bests and high scores are written to your browser's local storage, on the device that produced them — which is why they do not follow you to another browser, and why clearing site data removes them for good.

The site is still an ordinary web page, so loading it fetches its own HTML, CSS, fonts and images the way any site does. Where a tool needs something beyond that, its documentation page says so in Limits worth knowing rather than leaving you to guess.

Important

A permission prompt is the browser asking, not an account request. Microphone tests need microphone access to read an input level at all, and iOS gates motion sensors behind a tap. Deny it and only that test stops working.

Documentation FAQ

Do I need to install anything to use these tools?

No. Every tool documented here runs as a web page in the browser you are reading this in. There is no download, no extension and no account, which is also why each tool is limited to what a browser is allowed to see.

Is my test data uploaded anywhere?

The tools run in your browser and the tests themselves do not send your keystrokes, audio, sensor readings or calculator inputs to a server. Where a tool remembers settings, a personal best or a high score, it stores that in your browser's own local storage on that device. Each documentation page states what its tool stores.

Why does a tool ask for a permission?

Microphone tests need microphone access, and motion or orientation tests on iOS need an explicit tap to allow sensor access. The browser asks, not the site, and denying it stops only that one test. Any tool that needs a permission carries a Requires Permission badge on its documentation page.

How accurate are the measurements?

Accurate enough to compare runs and spot faults, not laboratory instrumentation. A browser measures through the operating system, the browser itself and its own deliberately coarsened timers, so a latency or refresh figure describes the whole chain rather than one component. Every page that reports a number explains what that number includes under Limits worth knowing.

Can I use the tools on a phone or tablet?

Most of them, yes. The Mobile category is built for phone hardware specifically. Keyboard tests need a physical keyboard to be meaningful, and the display tests are more useful on the larger screen you are actually judging.

Does the documentation match the tools exactly?

That is the point of it. Each page was written against the code that ships for that tool, so the thresholds, formulas and scoring rules printed here are the ones the tool applies. If a page describes something the tool does not do, that is a bug in the page and worth reporting.

Where do I start if I just want to check a keyboard?

Open the keyboard test on the home page, press every key once, and read the legend. The Quick start section above walks through it in four steps, and the Keyboard documentation covers the deeper tests once you know something is wrong.

Something in the documentation looks wrong. How do I report it?

Use the contact page and say which documentation page and which line. A correction to the documentation is a small change; a tool that behaves differently from its documentation is the more useful thing to hear about.

Still stuck? Contact us with the page and the symptom.

Start with the tool you need

All 81 tools are free, run in this browser, and need no account. Pick a category, or open the keyboard test and press a key.