Random Emoji Generator

Press Generate and you get a random handful of emoji. Click any one of them and only that one lands on your clipboard, which is the part most generators skip. They hand you five faces and a single Copy button that takes the lot, so you end up picking the string apart in a text box before you can use any of it.

The other thing you get here is a straight answer about what you are drawing from. There are 646 emoji in the pool, sorted into nine groups, and every figure on this page was counted rather than guessed at. It is free, there is no sign up, and the draw happens in your browser.

1. Pick a category, or leave it on All to draw from the whole 646.
2. Set how many you want. The box takes 1 to 50. Type 999 and it settles on 50, type 0 or a stray letter and it settles on 1, so a slip of the finger cannot break it.
3. Choose a skin tone if you want one. It touches the people and the hands, nothing else.
4. Press Generate, then press it again for a fresh draw. Each press is independent of the last.
5. Click a single emoji to copy that one on its own. Copy all takes the whole set, joined with single spaces.

Nobody publishes this, so here it is. Counted by drawing the pool dry one category at a time, it comes to 646, and the nine groups add up to exactly that, which means no emoji is filed under two of them.

group	in the pool	share of an All draw
Animals and Nature	102	15.8%
Smileys	95	14.8%
Objects	82	12.7%
People and Hands	78	12.1%
Food and Drink	77	11.8%
Symbols	73	11.2%
Travel and Places	56	8.6%
Activities	53	8.3%
Flags	30	4.6%

Those shares are counted off 100,000 draws with the category left on All. They come out uneven because the draw is even over all 646 characters and the groups are not the same size, so an All draw of 50 averages 7.9 animals and 2.3 flags. If you want a balanced spread of themes, pick the themes yourself and press twice.

The 646 are hand picked out of Unicode rather than swept up wholesale, and that is deliberate. Unicode's own tables mark 1,219 individual characters as ones that come out in color by default. This pool holds 559 of them, plus 87 characters built from more than one piece, like the country flags. What got left out is mostly the newer and rarer end, which is exactly the end a five year old phone has never heard of. Our Emoji Picker keeps its own list of 637 and the two agree on 612, so even two tools on the same site, built to the same instinct, disagree about 59 characters.

A generator can call itself random forever without anybody checking, so I checked. Pressing Generate 5,000 times at 50 emoji a press is 250,000 draws, and even odds puts each of the 646 at 387 of them. The quietest character in the pool came up 322 times and the busiest came up 455, with a chi squared of 692.6 against 645 degrees of freedom, which is ordinary noise at that sample size. Nothing was missing, all 646 turned up.

That run answers a second question you cannot see from the outside. A character accidentally pasted into the list twice would show up around 774 times instead of 387, and the busiest one in a quarter of a million draws was 455, so nothing in the bag is in there twice.

Ask for 50 with No repeats switched off and you will almost certainly get a duplicate. In those same 5,000 sets, 4,328 carried at least one emoji twice, which is 86.6 percent, and the average set had 1.865 duplicate hits in it. Fifty picks out of 646 is the birthday problem, so a repeat is the arithmetic behaving rather than a bug, it is what independent draws do.

Switch No repeats on and every emoji in the set is different. Across 4,118 no repeat sets, spread over every category, I counted zero duplicates in any of them.

There is one place the bag empties. Flags holds 30, so a no repeats draw from Flags tops out at 30 whatever you type, and the tool prints a line on screen saying so instead of quietly padding the set. It is the only category that can do this, because the next smallest is Activities at 53 and the count box stops at 50.

The tone buttons act on the People and Hands group and nothing else, which is 78 characters out of 646. I set the tone to dark and drew all 78 of them, and every single one took it. That was worth checking rather than assuming, because four of them carry an invisible marker that forces the color version, and the marker has to come off before a tone will attach. The four are the writing hand, the index finger pointing up, the victory sign, and the hand with the fingers splayed.

Leave the category on All and a tone changes roughly one tile in eight. Measured across 20,000 tiles it was 11.9 percent. A grinning face and an avocado come out the same either way, since neither of them has skin to tone in the first place.

Now the honest limit. The pool holds no family groups, no couples, and none of the person plus job combinations. Those are several emoji stitched together, each piece able to carry its own tone, and one tone button cannot sensibly drive five people at once. So what you get is single people and single hands, in the default yellow plus five tones, and that is the whole of it.

Every setting goes into the share link, and so do the emoji currently on screen. Send it and the person opening it sees your draw rather than a fresh one, which is the point when you are running a game and everybody needs the same three characters in front of them.

It does make for a long address. Fifty emoji come out as 689 characters of query hanging off the end of the page URL, and even five flags on their own is 174. If you are pasting it somewhere with a length limit, draw fewer.

Opening somebody else's link will not overwrite the settings you had. I seeded a browser with one set of preferences, opened a link carrying different ones, and read the storage back afterwards: zero writes, everything intact. Your own settings save the moment you touch a control yourself, and never before that.

An emoji is a shared code, and every platform draws that code in its own house style. The grinning face you send from an iPhone arrives on Android as Google's version of the same character. Nothing changed in transit and the meaning almost always survives the trip.

The empty rectangle is the worse case, and it happens when the device at the other end is running an emoji set older than the character you sent. That is the reason the pool is curated instead of complete. 559 of the 646 are plain single characters that render in color with no help at all. Newer emoji tend to land at higher code numbers, and the two highest in the whole pool are the blueberries and the adhesive bandage, so if anything here ever comes out blank on a friend's aging phone, those two are the likeliest culprits.

Country flags sit in their own category of risk. A flag is not a picture at all, it is two letter shapes stitched together, and 24 of the 30 in the Flags group are built that way. Whether the receiving device turns those letters into a flag or just prints the letters is entirely up to that device, and it is the one thing on this page I had no way to test from here.