Frequently Asked Questions

Every value on this page comes from the browser's built-in cryptographic random API (Web Crypto, crypto.getRandomValues) rather than the usual Math.random(), and modulo bias is removed with rejection sampling — good enough for values that must be unpredictable, like passwords and UUIDs.

No. Everything is computed inside this browser and no value is ever sent to a server. For convenience, only the last tab you opened is remembered in this browser; generated passwords, codes and profiles are not stored.

No. Names, phone numbers, emails and addresses are fake data built from random combinations of common surnames, syllables and place-name fragments, unrelated to any real person. Use them only for testing screens and forms — never for identity checks or real sign-ups.

It is the number of guesses a brute-force attacker would need, on a log scale. Each extra bit doubles the possibilities. Generally, 60 bits or more is considered safe even against offline brute-force attacks. It rises as you add length or character types.

At the same entropy, something like ABc7#kQ2 is harder to remember and type by hand than river-fox-42-stamp. More words means more security — use five or more for important accounts.

Usually it's because the check options at the top are turned off (the setting is saved in your browser, so whatever you turned off last time stays off). Turn back on whichever you need: Misuse, spacing, repeats, punctuation, loanwords. If it still doesn't work, press the "Sample" button to see whether results appear, then try turning off browser extensions or doing a hard refresh (Ctrl+F5).

Quick check instantly catches only what matches the rules for common mistakes, spacing patterns, punctuation and loanwords. Nothing to download. Thorough check adds Hunspell (an open-source Korean dictionary of about 300,000 words) on top, filtering out typos the rules don't cover (like "됀다" or "안됫어") and wrong conjugations. Turning it on downloads the dictionary and engine (about 2.5MB) once; after that it runs from cache.

Misuse (되/돼, 안/않, 왠/웬, -든지/-던지, -로서/-로써, 몇일/며칠 and so on), Spacing (할수있다 → 할 수 있다, 안된다 → 안 된다 — bound nouns, auxiliary verbs and negative adverbs), Repeated typos (the same character typed twice, spaces before punctuation), Loanwords (메세지 → 메시지, 컨텐츠 → 콘텐츠, 악세사리 → 액세서리 and so on), and with Thorough check, anything flagged as Not in dictionary.

Not quite. Those checkers run morphological analysis and grammar rules on a server. This site is deployed statically with no server, so it can't call them directly (CORS and terms of use) and uses only dictionaries and rules that run in your browser. Thorough check greatly widens typo coverage, but context-aware spacing (e.g. fully correcting "나는학교에간다") is still weak, and it may flag expressions that are correct in context.

Nowhere. All checking happens inside this browser (thorough check in a background worker), and your text is never sent to a server or saved. The only things saved are the checkboxes at the top and whether thorough check is on.

Right after you press it, an Undo notice briefly appears at the top. Meanwhile, Ctrl+Z (⌘+Z on Mac) restores the original. Fixing items one at a time by clicking the underlines is safer.

The spelling dictionary is hunspell-dict-ko (GPL-3.0 / CC-BY-SA-4.0) and the checking engine is hunspell-wasm (a WebAssembly port of Hunspell, MPL/GPL/LGPL). The rule set was written by hand with reference to the National Institute of Korean Language's standards and public collections of Korean spelling rules.

They are the same thing. HSB(Hue·Saturation·Brightness) is the name Adobe uses, and HSV(Value) is another name for the same model; this tool treats them as one. HSL is different — in HSL a pure color sits at L (lightness) 50%, while in HSB it sits at B 100%.

The wheel's Hue angle is what the rules go by. Complementary sits directly opposite the base color (180°), analogous 30° either side, triadic 120° apart and tetradic 90° apart. Monochromatic keeps the hue and varies only saturation and brightness — the same approach as Adobe Color.

It is the luminance ratio between two colors. Body text needs 4.5 or higher (AA); large text and headings need 3.0 or higher to read comfortably. 7.0 or higher is AAA, very crisp. The first palette color is treated as the background and the rest as text.

It applies the color-vision matrices published by Machado et al. (2009) as-is, so it is an approximation. Color vision deficiency varies from person to person, so it can't reproduce exactly "what they see", but it is good enough to judge whether the two colors stay distinguishable. Red-green deficiency is the most common, affecting about 5% of Korean men, so always check palettes that put green and red side by side. Greyscale is the final test: if the colors still differ there, the palette doesn't rely on color alone. The preview only changes what's painted on screen; copied and saved HEX values are always the original colors.

It only appears when your browser supports the EyeDropper API (desktop Chrome, Edge and others). Otherwise the button is hidden — just paste a HEX value instead.

No. This site keeps no data on a server. Saved palettes and your last color are stored here only — not visible in private windows or on other devices.

No. Loading, rotating, cropping, filtering, resizing and saving all happen inside this browser with Canvas. Images are never sent over the network.

This editor keeps the original pixels and recomputes every frame from them in the order rotate → flip → filter → crop → resize. However much you move the sliders, no quality loss builds up before you save. (If you reopen a saved JPEG and edit it again, JPEG compression is applied once more at that point.)

It changes the image's width and height in pixels. Shrinking makes the file smaller; enlarging fills in pixels by interpolation, so it won't get sharper than the original. This is about screen pixels, not print DPI.

Even a 1–2° rotation leaves empty triangles in the four corners. To hide them, the image is zoomed in just enough to fill the frame. You can set the exact area again with Crop.

Most modern browsers can save WebP, but some environments fall back to PNG. Check the file extension after saving, and if you need WebP, try another browser such as Chrome or Edge.

The elements you added are merged into the image as pixels at that point, just like cropping. After merging they can no longer be selected, moved or resized, so finish rotating, flipping and fine-angle adjustments before decorating. Resizing is the exception: decorations scale along with the image.

No. The checkerboard is only a marker that the area is transparent — it is not saved. Saving as PNG or WebP keeps the transparency; saving as JPEG, which has no transparency, fills those areas with white.

Yes. Opening the page gives you a blank canvas right away. In the Decorate tab, dragging on an empty spot draws right away, and you can save as PNG. If you need a background, turn on "Fill with a background colour" in the Decorate tab.

QR Codeand the 1D barcodes EAN-13·EAN-8·UPC-A·UPC-E·Code 128·Code 39·ITFare read by this page itself. Where the browser has code detection built in — Chrome and Edge family browsers, and Android — that is used first, which also brings Data Matrix·PDF417·Aztec(driving licences, boarding passes, logistics labels). Firefox and Safari have no such feature, so only the seven formats above work there.

Browsers hand over the camera only on a page opened over https. Click the padlock in the address bar and check the camera permission is not set to "Block". Once you refuse, you are never asked again, so switch it to "Allow" and reload. Even with no camera, or a blocked one, an image file or a pasted screenshotstill works.

Yes. Drop the image in, or take a screenshot and press Ctrl+Vto paste it, and it is decoded right away. If it won't read, check that the white margin around the code is not cropped off, and that the code is not too small in the picture. Barcodes read far better when the bars run horizontally — crop it that way.

This page never jumps to a link on its own. The full address and its domainare shown first, and it opens in a new tab only when you click. Because a QR code hides the address from the eye, fake payment and delivery notices ("quishing") are common — so check whether the domain is one you recognise before you tap. javascript: and other dangerous forms are blocked from being clickable at all.

No. The camera feed is only shown on screen and decoded; it is never uploaded anywhere. Decoding also finishes inside this browser. Recent scans are saved in this browser for convenience and can be removed at any time with "Clear history" (clear them after use on a shared PC).

Put it on a business card or event sign and the scanning phone opens its save-contact screen right away. It uses the most widely compatible format, vCard 3.0 (tested with the default iOS and Android cameras). But the more fields you fill, the denser the code and the harder it scans — stick to name, phone and email and replace the rest with a single website link; it reads far more reliably. What you enter never goes to a server and stays only in this browser (clear it after use on a shared PC).

No. It just opens your messaging app with the number and message already filled in; the person scanning still has to tap send. It uses SMSTO:, which both Android and iOS understand.

It can, since the logo covers part of the code. QR codes have a feature called Error correction that recovers data even when part of the code is hidden, so when you add a logo, raise the level to H(30%) and keep the logo at 25% or less. Always test-scan with a real phone after saving.

Scanners read the pattern by the difference in brightness. Keep the foreground (pattern) dark and the background light, with enough contrast between them. Light grey on white, or pastel pairs, often fail. The "Black & white" button restores the defaults at any time.

It stores the network name, password and encryption type in the standard format (WIFI:T:WPA;S:name;P:password;;). The default Android camera and iPhone (iOS 11+) recognise it and offer to join right away. Because the password is stored as-is, use a guest network for QR codes posted in public.

For use as-is on screen or in print, choose PNG; to blow it up for business cards or banners, or edit it in a design tool, choose SVG (vector, so it stays sharp at any size). A raster logo (PNG/JPG) is embedded as an image inside the SVG.

No. The QR code is generated entirely inside this browser. Your last settings (content, colors, options) are saved in this browser for convenience; uploaded logo images are not saved.

With a mouse most people land between 200 and 300 ms, because that is roughly how long it takes a person to see light and move a finger. Frequent gamers often sit at 180-220 ms, and tapping on a phone tends to come out 20-50 ms slower because touch detection is added on top. Anything under 100 ms almost certainly means you guessed when the green would appear.

Not as an absolute value. Anything measured in a browser also contains the refresh rate of your monitor (up to 16.7 ms at 60 Hz), the polling interval of your mouse or keyboard, and the compositing delay of your operating system. The same person at the same moment scores faster on a 144 Hz gaming monitor than on a laptop screen. So comparing across devices means little; comparing yesterday and today on the same device is the useful comparison. This tool anchors the measurement as close as it can to the moment the screen was actually painted, which reduces the error but cannot remove those delays.

One round is luck. A single blink already takes more than 100 ms, so if the signal lands right then the whole score is ruined. In the other direction, if the wait drags on and you press early on a hunch that happens to be right, you get a number faster than human reaction allows. Averaging five rounds cancels much of both. The wait is drawn at random between 1.2 and 4.5 seconds each time, so there is no rhythm to memorise.

Yes. The measuring panel is a real button, so Space or Enter works, and on a phone you simply tap the screen. Each input method has its own delay, though, so do not compare mouse, keyboard and touch scores against each other. For anyone who has trouble telling the colours apart, the wording inside the panel changes too — "Wait" and "Press now!" — and the state is announced to screen readers.

In this browser only. Your nickname and the averages of your last ten runs stay in this device storage, and nothing is sent to a server. That is also why there is no global leaderboard — as long as the number is produced in a browser it can be faked, and a leaderboard nobody can police is worse than none. To clear your records, delete this site data in your browser.

Only the nickname you typed, your average and grade, the five individual times, the date you measured, and the address of this page. You can check exactly what it contains in the preview above the buttons. The image is drawn inside this browser and goes nowhere until you press Save, Copy or Share yourself.

This tool only measures the path to the Cloudflare edge serving this site. Official speed tests measure against dedicated infrastructure in many locations, so routes and servers differ and the numbers can disagree. If you need your line's exact top speed, use the official tester your ISP recommends.

It's how much your latency jumps around: the average difference in round-trip time between consecutive requests, the same definition the video and voice call standard RFC 3550 uses. Even with low average latency, high jitter makes video stutter and players teleport in games. Roughly: Under 10 ms is good; over 30 ms is noticeable on video calls. If it's especially high on Wi-Fi, suspect router distance or interference first.

No. For security reasons browsers can't send ICMP (real ping) directly. Instead, several short requests go to this site's server and the median round-trip time is shown. It's latency measured against this site, so it differs from your ping to other sites or game servers.

Not yet. Having the server connect to any address a user types could be abused (for port scans, or as a relay to attack other servers), so it will be added separately once its scope is safely limited.

That's normal. IP location isn't GPS — it is estimated from the city and ISP registered to the IP range, so it usually lands on an ISP gateway or a nearby city centre. On mobile data it can be tens of kilometres off. Put the other way round: a website cannot learn your exact address from your IP alone.

The IP itself is never sent anywhere. The location doesn't come from a lookup service either — the Cloudflare edge that received your request simply returns what it already knew. One exception: the map images (tiles) are fetched from openstreetmap.org — when the map is shown, the tile addresses covering the visible area are requested from there.

The Latency above is how long the browser took to send a real request and get an answer, so server processing and browser wait are mixed in. The TCP round-trip time here is what Cloudflare observed while establishing the connection, so it is closer to the line itself. Depending on how you connect, it may be unavailable.

No. Speed and latency results are only shown on this screen and never stored. Connection details likewise just reflect that moment's request and aren't logged.

While this page is open it always listens for keys — there's nothing to turn on. If one key still doesn't respond, it really isn't being detected — except that the Fn key, dedicated laptop keys (brightness, volume) and shortcuts the operating system grabs first (Win+L, Alt+Tab, etc.) never reach the browser, so this tool can't check them. Tab lights up, but its default browser behaviour (moving focus) is deliberately left alone so you can still leave the page.

A key is flagged when it is re-pressed within 30 ms of release — a person can hardly press the same key twice that fast on purpose. It turns yellow and shows the interval. Repeats the operating system generates while you hold a key (Auto-repeat) aren't chatter and aren't counted. One or two flags may be false alarms — watch whether the same key keeps getting flagged.

Hold several keys down at once and the Max simultaneous value climbs. Cheap membrane keyboards top out at 3–4 keys, while gaming keyboards register 6 or more, or every key. However, Ctrl, Alt and Shift combos can clash with browser shortcuts, and this tool deliberately doesn't block them, so testing with letter keys is more accurate.

Yes. Rather than the typed character, this tool uses the Physical key (event.code) value, so the same key lights up whatever your input language or keyboard layout. Note that the Korean/English toggle key to the right of the bottom-left Alt key arrives, depending on the device, as AltRight or a separate code.

It is a device to make you finish the check. Left alone, people only press the keys they already use often; this game picks one key at random on the keyboard layout, so you end up stepping on keys you would never think to test. Play a while and the Keys seen count fills itself in. A round lasts 30 seconds, and a key you fail to hit in time moves on to the next one — so a dead key can never freeze the game. Score and best score live only in this page's memory and vanish on refresh.

They aren't stored. Stats and input order live only in this page's memory and vanish on refresh. Nothing goes to browser storage or a server — the key sequence could contain passwords, so there's no reason to keep it.

No. It requires at least one year of continuous service and an average of 15+ hours a week over four-week periods. Beyond the first year it is prorated by day (e.g. 2 years 6 months = 30 days' pay × 2.5).

Average wage is the total pay received in the three months before leaving divided by the number of days — based on what you were actually paid (including bonuses and overtime). Ordinary wage is the amount set to be paid regularly and uniformly (base pay + fixed allowances), and is the basis for leave pay and overtime. Severance uses the average wage, but if that is below the ordinary wage, the ordinary wage is used.

If leave is still unused when the usage period (usually one year) ends, it is paid out on the next month's payday. However, if your company properly followed the leave-use promotion procedure and you still didn't take it, you may lose the right to claim the pay.

It is usually waived once 3 years have passed since the loan started. There are also exceptions when you switch to a new policy loan (refinancing), when the bank runs a waiver promotion, or where the law provides (e.g. some low-income finance products). Some products let you repay a set share each year fee-free even within 3 years.

With a savings plan, each monthly deposit is held for a different length of time. Only the first deposit earns interest for all 12 months; the last earns one month. So pre-tax interest on a 12-month plan is monthly amount × monthly rate × (12×13÷2), a little over half the annual rate measured against your total deposits. The "3.5% a year" is the rate for the time each deposit is held, not a return on everything you paid in. The after-tax return on principal in the results above is what you actually take home.

These forms are simplified estimates of the statutory rules. Actual amounts vary with company employment rules, retirement pensions (DC plans depend on contributions and returns), how bonuses are counted, taxes (retirement and earned income tax) and changes to social insurance rates. Check exact figures with your HR team, bank or a tax professional.

No. All calculations happen inside your browser. Only the last calculator you opened is remembered in this browser; inputs like salary or loan amounts are never saved or sent anywhere.

To leave the volume of your music untouched, it doesn't play a silent background track. Instead, "Keep screen on" is on by default so the screen stays awake, and if it briefly turns off, elapsed time is corrected against the real clock. Before each segment change, an alert that ramps up over the final 8 seconds with beeps and vibration plays, so you're unlikely to miss it even with music on.

Keeping the screen fully off while using other apps for a long time can make the browser slow the timer down. For outdoor runs, keep the screen on.

No. When on, it only shows your current pace (time per km) and distance while you run. Coordinates and your route stay in this browser's memory only and disappear when the workout ends or you leave the page. They are never sent to a server or kept on your device. (Only settings are saved in this browser — location never is.)

Better to turn it off. With no real movement, distance shows 0 and pace shows --:--. If you only need time-based intervals, just use it without GPS.

Run segments grow and then shrink, e.g. 1→2→3→2→1 minutes. Turn it on under "Fine-tune"; run time then steps up and down in multiples of the base value.

Yes. Run, rest, reps and every other setting are saved automatically in this browser and loaded next time. Save combinations you use often with the star (★) button to Favorites and load them with a single tap.

No, there are no sample files at all. Every sound is synthesized moment by moment with the browser's built-in Web Audio API. The piano and Otamatone run oscillators through filters, the guitar uses Karplus-Strong (a string-synthesis technique that feeds a short burst of noise into a decaying delay line) for plucked notes, and the drums wrap noise and oscillators in short envelopes to imitate hits.

Browsers block audio until you interact with the page once (the autoplay policy). Tap any key, string, pad or the neck once. On an iPhone, also check the silent switch on the side.

Piano, guitar and drums each sound independently with multi-touch or several keys pressed together. The Otamatone, like the real instrument, plays only one note at a time — press several spots and the most recent one wins.

A toy-meets-instrument from the Japanese company Maywa Denki. You set the pitch by pressing the stem attached to its head-shaped body, and squeeze the body to open and close its mouth as it sounds. Here, dragging along the neck glides smoothly between notes, just like the real thing.

A 54-card deck — 52 playing cards plus a black joker and a color joker — is shuffled face down and players take turns flipping one card at a time. Whoever flips the color joker is picked and the round ends. Whether the pick is a forfeit or a prize, it works whenever you need to single out one person, such as who buys coffee or who goes first. The One Card attack and block rules mean someone sometimes has to draw several cards in a row, and that is exactly when the color joker is most likely to turn up.

An ace, a two or the black joker attacks the next player. The player under attack first draws one card. If that card can pass the attack on, its value is added and it goes to the next player; if it is a 3 matching the suit of the last attack card, the attack is gone. Otherwise the card just drawn counts as one and you draw the remaining cards one at a time. For example, if you are hit by an ace (3 cards) and draw a 7, you draw 2 more.

A black joker attack (7 cards) can only be passed on with the ace of spades. That adds 5, so 12 cards go to the next player, who can block with the 3 of spades or pass them on again with another ace or the 2 of spades. The black joker has no suit, so a 3 cannot block it. The black joker can also be used to pass on another attack, but only an ace or two of a black suit (spades or clubs).

Every card in the middle does nothing. Only the last card of the penalty takes effect, just like a normal turn: a K means draw again, a Q reverses the order and an ace attacks the next player. The color joker is the one exception — if it shows up partway through a penalty, that player is picked on the spot.

All 54 cards are shuffled at once with the browser's cryptographic random numbers (crypto.getRandomValues). The order never appears on screen — only the number of cards left — and there is no seed number that would let anyone work out the order in advance. Cards and names are used only in this browser and never sent to a server. Names and the pick description are saved in this browser only, so they are ready next time.

From 2 to 8 players. The first turn is drawn at random every round and play then follows the order the names were entered. Seats that draw earlier are slightly more likely to be picked in this game, so choosing the starting seat at random gives everyone the same chance. With 2 players a Q's reverse means nothing, so play simply passes to the other player, while a J skips the other player so the same person draws again. There is always exactly one color joker in the deck, so someone is always picked before all 54 cards run out.