Keyboard Tester
Test every key on your keyboard in the browser — see what each key reports, how many register at once, your auto-repeat timing, and which keys are stuck. Nothing is uploaded.
Keyboard Tester
Rollover — how many keys register at once
Hold at least two ordinary keys at once to start measuring.
Last key
Press any key. Nothing you type is stored or sent anywhere.
Held right now
Auto-repeat timing
Hold one key down until it starts repeating.
Coverage
Press keys to work through the layout.
Stuck keys
No key has been held longer than 5 seconds.
Recent presses
Keys a web page may never see
Fn— resolved inside the keyboard's firmware. It produces no event on any operating system, so no browser tester can show it.F11andF12— full screen and developer tools. Most browsers act on these before the page does.PrintScreen— handled by the operating system. On Windows it commonly arrives as a key-up with no key-down.MetaLeft/MetaRight— the Windows or Command key opens the Start menu or Spotlight, which may take it first.- Combinations the OS owns outright — Cmd+Q, Alt+F4, Ctrl+Alt+Del — cannot be intercepted by any web page, on any site.
How to Use the Keyboard Tester
Pick the layout in front of you
ANSI for a US board, ISO for UK and most of Europe, JIS for Japanese. This decides which keys the diagram draws and what the coverage total counts — it does not change how keys are detected.
Work across the board once
Every key you press turns grey and stays grey, so the coverage panel is a checklist rather than a score. Anything still unmarked at the end is either untested or genuinely dead.
Mash a handful of keys together
Hold as many ordinary keys at once as you can. The rollover panel reports the most that registered together — as a floor, since it can only see what you actually pressed. Modifiers are counted separately, because the keyboard sends them separately.
Hold one key to read your repeat settings
Keep a key down until it starts repeating and leave it a moment longer. You will get the delay before repeating starts and the rate once it does — two separate operating-system settings, measured rather than looked up.
What a Keyboard Test Can and Cannot Tell You
A keyboard tester looks like it should be trivial: press a key, light it up, done. The reason most of them mislead you is that a browser sits three layers away from the hardware. What arrives at a web page has already passed through the keyboard's own firmware, the USB report format, and the operating system's input stack, and each of those layers can change, delay or swallow a keypress. A test that ignores those layers gives confident answers to questions it cannot actually see, and the most common of those answers is a rollover figure.
Rollover is a floor, not a specification
Rollover is the number of keys a board can report simultaneously, and it matters to gamers and fast typists because a keyboard that drops the fifth key silently loses inputs during a burst. The number quoted almost everywhere is six, and there is a precise reason for it: the USB keyboard boot-protocol report is eight bytes, of which one is a modifier bitfield, one is reserved, and six are key slots. Six is a report format, not a wiring limit — which is why a keyboard that advertises full n-key rollover works by sending a different, larger report instead.
Here is what no tester can do: measure the ceiling. All any of them observe is the most keys you happened to press at once. Press four and the honest answer is "at least four" — the fifth might have worked perfectly. This tool therefore reports a lower bound and labels it as one, and it treats exactly six as unsettled rather than conclusive, because six is precisely the figure a boot-protocol board and an n-key board both produce when you have only pressed six.
The number other testers print as your keyboard's rollover is really the number of keys you pressed. It is a fact about your hands, presented as a fact about your hardware.
Modifiers are sent a different way
The second thing to get right is that Ctrl, Shift, Alt and Meta are not ordinary keys as far as the report is concerned. Each has its own bit in the modifier byte, so holding all four consumes none of the six slots. A tester that counts every held key together will tell you a standard 6KRO keyboard manages ten-key rollover the moment you rest two fingers on the modifiers — flattering, and wrong. Counting them separately is a one-line difference in code and a large difference in whether the answer means anything.
Layouts across Asia, and why the diagram has three
Singapore, Malaysia, the Philippines, Indonesia and Vietnam all standardise on the US ANSI layout at retail, which is why a keyboard bought in Sim Lim or Low Yat looks identical to one bought in Texas. That uniformity is convenient right up until you import — an ISO board ordered from Europe arrives with an extra key wedged between the left Shift and Z, and a taller Enter that takes its space from the row above. Japan is the region's other case: JIS boards carry five keys ANSI has never had, three of them dedicated to switching between kana and Latin input, and a JIS keyboard attached to a machine configured for ANSI will report those keys as nothing at all.
There is also a climate argument for testing a keyboard periodically rather than only when it fails. Membrane keyboards in humid, un-airconditioned rooms — which is most of maritime Southeast Asia for most of the year — degrade gradually rather than suddenly: a switch starts dropping one press in twenty long before it dies outright. That failure is nearly invisible while typing, because you correct the typo without registering why you made it. It is obvious in a press history that shows twenty presses producing nineteen events, which is why this tool keeps one.
What it measures about your computer, not your keyboard
The auto-repeat panel is the one measurement here that describes your operating system rather than your hardware. Two settings govern it — how long a key must be held before repeating begins, and how quickly it repeats afterwards — and they are separate values that people routinely conflate. Reading them back by measurement rather than from a settings pane is useful precisely because it reflects what your applications actually receive, delays and all.
10 Facts About Keyboards and Key Events
The USB keyboard boot-protocol report is eight bytes: one modifier byte, one reserved byte and six key slots. That is where the six-key rollover ceiling comes from — it is a report format, not a wiring limit.
Ctrl, Shift, Alt and Meta ride in a separate bitfield, one bit each. They do not consume any of the six slots, so a keyboard can report four modifiers plus six ordinary keys and still be a plain 6KRO board.
KeyboardEvent.code names the physical key, not the character. The key that types “a” on a French AZERTY board still reports KeyQ, because code uses US-layout positions as its naming scheme.
keyCode — the number most old tutorials use — is deprecated in the UI Events specification. code and key replaced it, and they answer two different questions.
The Fn key produces no keyboard event at all. It is resolved inside the keyboard's own firmware before anything reaches the operating system, which is why no browser-based tester can show it.
A full-size ANSI board has 104 keys, ISO has 105, and JIS has 109 — the same core with different additions.
The single key that makes ISO 105 rather than 104 sits between the left Shift and Z, and reports as IntlBackslash.
JIS adds five keys ANSI has never had: 変換, 無変換, かな, ¥ and ろ — reported as Convert, NonConvert, KanaMode, IntlYen and IntlRo.
Auto-repeat is two separate operating-system settings: the delay before repeating starts, and the rate once it does. Averaging them together reports a speed nobody has configured.
The location property is how a page tells left Shift from right Shift, and the number row from the numpad: 0 standard, 1 left, 2 right, 3 numpad.
Frequently Asked Questions
- Because a lower bound is the only honest answer. Every measurement here is of what your keyboard reported while you happened to be pressing keys — if you press four at once, no test on earth can tell whether the fifth would have registered. Other testers print a bare number; that number is really "the most keys you pressed", dressed up as a hardware specification. Press more keys together and the figure rises.
- Because the keyboard does not send them the same way. The USB boot-protocol report carries the eight modifier keys as single bits in a dedicated byte, separate from the six slots that ordinary keys compete for. Counting them together makes a standard 6KRO keyboard look like a 10-key-rollover board as soon as you hold two modifiers, which is why this tool counts them separately and says so.
- Probably not — several keys are taken by the browser or the operating system before a web page sees them. F12 opens developer tools, F11 toggles full screen, the Meta key opens the Start menu or Spotlight, and PrintScreen is handled by the OS (on Windows it commonly arrives as a key-up with no key-down). The Fn key never generates an event at all. These are listed under the keyboard so you can tell an intercepted key from a dead one.
- No. Everything runs in your browser, there is no network request of any kind, and nothing is written to storage — not even your layout choice. Reload the page and the whole session is gone. This matters more than usual for a keyboard tester, because testing one means typing on it.
- No, and the tool clears it for you. If you change window or tab while holding a key, the key-up is delivered to whatever took focus and never reaches this page. Left uncorrected, the key would stay lit and then be reported as stuck — a hardware fault invented out of an alt-tab. Every held key is released when the page loses focus, so click back on the page and press it again.
- Ghosting is when a keyboard's switch matrix cannot distinguish certain three-key combinations, so a key you never touched registers — or, more often on modern boards, a key you did press is blocked. This tool cannot separate a blocked key from a key you simply did not press, because both look identical from the browser: no event arrives. What it can do is show you the combinations that do register, which is how you find the blocked ones by hand.
- No. Switch layout at the top — ANSI, ISO and JIS are all drawn — and the coverage count follows your choice. The underlying detection does not depend on it either way, because
codereports the physical key position rather than the character produced. The layout picker exists so the diagram matches the board in front of you and the coverage total is the right number. - It reads back your operating system's two keyboard settings: how long you must hold a key before it starts repeating, and how fast it repeats after that. It is a useful check when typing feels sluggish or runaway, and it is measured rather than read from a config file, so it reflects what your applications will actually experience. It measures the OS, not the keyboard hardware.
- Neither label fits an intermittent key well. Stuck means held down with no release, which is what the stuck panel detects. A key that registers only sometimes is usually a dirty or worn switch, and the way to see it here is the key-press history: press the suspect key twenty times and watch whether every press appears. A switch that drops one press in five is visible in that list long before it is obvious while typing.
- On a laptop, yes — the built-in keyboard reports the same events as an external one, though laptops usually lack a numpad and the coverage total will reflect that. On a phone or tablet with an on-screen keyboard, no: software keyboards frequently report no useful
codeat all, and many send a single generic value for every key. Connect a physical keyboard and it behaves normally.
Related News
You may be interested in these recent stories from our newsroom.
No related news yet for this tool. Our editorial team publishes new pieces every week.
Browse all news →Pick up where you left off
Stored only in this browser — never sent to our servers.