Everyone shares a secret word except one person, and the table takes turns describing it until somebody's bluff falls apart. This page sets that up with a single phone. Pick three to twelve players, deal the round, and each seat taps «Show my card», reads it, taps «Got it» and passes the phone on. The liar can be told the category or nothing at all, and from seven players you can run two liars. The deck holds one hundred words across food, animals, places, jobs and objects — written separately for Korean, English, Japanese and Spanish so the words are ones that language's players actually know. No paper, no install, nothing sent anywhere.
The first speaker is always a civilian. A liar opening the round has nothing to work from and is caught almost every time — at a real table nobody makes the liar start either.
Two liars unlock at seven players. Below that, two liars leave so few civilians that the question stops being «who is lying» and becomes «who is not».
The word and the liar exist only on this screen. Nothing is sent to a server and reloading clears everything.
Each language has its own deck rather than a translation, because a translated word is often one the players do not know, or one that gives itself away in a single line.
Everyone gets the same word except one or two players, who are the liars. The table goes round describing the word one line at a time, and at the end everyone votes for whoever they think is faking.
The liar knows they are the liar but not the word. Their job is therefore two jobs at once: reverse-engineer the word from what others say, and contribute a line that sounds like knowledge. Doing both takes time, and that hesitation is the main tell.
Civilians have the mirror problem. They must sound like they know the word without handing the liar anything to copy. That is why civilians end up frustrated with each other, and that frustration is most of the fun.
The only technical problem this game has is whether a card stays hidden while the phone moves. So the card takes two presses: one to open it, one to cover it again. The next seat is not announced until it is covered, which removes the usual accident of handing the phone over while a card is still open.
Liar cards are visually distinct. Confusion about whose card was on screen ruins the round outright, and a different background is recognisable even out of the corner of an eye.
When everyone has looked, the page names the first speaker, chosen from the civilians — a liar speaking first has nothing to go on and gives themselves away immediately.
First, everyone must know the word. A player who has never heard of it looks exactly like a liar, and the game turns into a quiz about vocabulary rather than a game about bluffing.
Second, the word must have several angles — shape, use, place, memory. If there is only one thing to say about it, the third and fourth players have nothing left and start repeating each other.
Third, it must survive the first line. That rules out proper nouns: the Eiffel Tower is over in one sentence. «Giraffe» is nearly as bad — one word about the neck and it is done — while «hedgehog» takes several lines to circle.
The Korean, English, Japanese and Spanish decks were each written separately, because translating a word usually breaks one of the three rules above. Korean gimbap loses half an English table on the first rule; Spanish churros, translated into Korean, is «a Spanish pastry» and the round is over on the third.
So each deck substitutes words of the same character instead. Every deck holds one hundred words: twenty each across food, animals, places, jobs and objects, matched in size so that picking a category does not change the difficulty.
This is also why the tool is served in only four languages. Opened in a language whose deck has not been written, it would deal words the table does not know, and the game would not work at all.
The most common addition is a last chance: a liar who has been voted out still wins if they can name the word. It stops civilians from being too explicit and makes every line a judgement call.
The second is two laps instead of one. A single lap gives so little information that the vote is close to a guess; on the second lap people must reuse what they already said, and the liar's contradictions surface. With six players or fewer, two laps is the better default.
The third is running the same word twice with a different liar. Watching who repeats a phrase from the previous round is revealing, and it works well in groups who play together often.
Nothing is saved. The word, the liar assignment and your settings live on the screen only, and leaving or reloading the page clears them. Nothing is sent to a server.
One consequence: pressing Reveal by accident ends that round for good. Dealing again produces a new word and a new liar, and there is no way to restore the previous one.
Three is the minimum, but four to eight plays best. With three there are only two civilians, so one bad vote ends it; past nine a full round takes long enough that the liar can hide at the back. Add a second liar when the group is large.
For a first game, yes. Told nothing, the liar is usually caught on their first line and the round is over in thirty seconds. Once the group has the hang of it, switching to «nothing at all» makes rounds far tighter.
One sentence, roughly one noun's worth. Too specific and the liar simply echoes you; too abstract and the civilians start suspecting each other. «I go there most weekends» is about the right altitude.
Tables differ, but two rules are common. First, the liar wins by not being voted out. Second, a liar who is voted out still wins if they can name the word. Adding the second rule stops civilians from being too explicit and makes the whole round sharper.
No. It runs in the browser with no account, one phone is enough, and adding the page to your home screen makes it work offline.