What Is an Emoji? Unicode, Sequences, and Why Yours Looks Different
Unicode decides the number and the name. Apple, Google, Microsoft and Samsung each draw their own picture. That one split explains almost every emoji surprise you have ever had.
Published
The single most useful thing to know about emoji is that they are not pictures. They are text. 😀 has a number and a name in the same standard that gives A and あ theirs, and the colourful thing you see is a font, drawn by whoever made your device. Almost every emoji surprise, from the wrong skin tone to the empty box, comes out of that one split.
The word is Japanese, and it has nothing to do with "emotion"
Emoji is a Japanese loanword written 絵文字: 絵 is picture, 文字 is character. The resemblance to "emotion" and "emoticon" is a coincidence, not an etymology.
Three words get used interchangeably and mean three different things. An emoticon is punctuation read sideways, like :-) and :D. A kaomoji is the Japanese upright kind, like ( ´▽｀), which needs no head tilt. An emoji is a coloured pictographic character defined by Unicode, like 😀. The first two are arrangements of characters you already had; the third is a character in its own right.
It started on Japanese phones in 1999
The usual starting point is 1999, when NTT DoCoMo launched i-mode with a set of 176 icons drawn on a 12×12 pixel grid so that people could put tone into messages that had room for very little else. Shigetaka Kurita designed them, and the Museum of Modern Art in New York acquired the original set in 2016.
That set was not the first. In 1997 the J-Phone SkyWalker DP-211SW, from the carrier that later became SoftBank, shipped about 90 monochrome pictographs. The phone sold poorly, so i-mode is the one everybody remembers.
For the next decade each Japanese carrier ran its own incompatible set. A smiling face sent from one network could arrive on another as a different picture, or as nothing at all. Getting emoji out of Japan meant first getting the world to agree on which slot holds which thing.
Unicode assigns the number and the name, never the artwork
Unicode 6.0, released in October 2010, brought 722 emoji into the standard, after Google and Apple proposed unifying the Japanese carrier sets. Since then an emoji is a character with a code point: 😀 is U+1F600, named GRINNING FACE.
And that is where the standard stops. It fixes the number and the name; it does not fix the drawing. Every platform ships its own font: Apple Color Emoji, Google's Noto Color Emoji, Microsoft's Segoe UI Emoji, Samsung's One UI set, and the open-source Twemoji used by a lot of the web. Same character, different picture, same message.
This is where the famous accidents come from. Apple redrew the pistol 🔫 as a green water pistol in iOS 10 in 2016 and the other vendors followed over the next couple of years, so for a while the same code point was a toy on one phone and a firearm on another. Google's cheeseburger 🍔 put the cheese under the patty until the internet noticed in 2017. The character never changed. The picture did.
Many emoji are more than one character
This is the part that catches people out. What looks like one emoji is often a sequence of code points stuck together:
- Skin tones: a base emoji followed by one of five modifiers (U+1F3FB to U+1F3FF), added in Unicode 8.0 in 2015.
- Flags: two regional indicator symbols spelling an ISO country code. 🇹🇼 is U+1F1F9 U+1F1FC, the letters T and W.
- ZWJ sequences: several emoji glued with the zero width joiner U+200D. 👩💻 is 👩 + ZWJ + 💻.
- Variation selectors: U+FE0F switches a character that would otherwise be plain black-and-white text into its emoji form. ❤️ is ❤ (U+2764) plus U+FE0F.
Two practical consequences. First, an emoji eats more of a character limit than it looks like it should: in JavaScript '👍'.length is 2 and '👨👩👧'.length is 8, and SMS and a lot of form validation count the same way. Second, when the receiving font does not know a whole sequence it falls back to the parts: a family arrives as three people in a row, 👨👩👧. That is designed behaviour, not a bug.
Red heart (with U+FE0F)
Two code points: U+2764 plus the variation selector U+FE0F that asks for the colour emoji form.
Family: man, woman, girl
Five code points: three people joined by two zero-width joiners (U+200D). A device that does not know the sequence shows the three people side by side instead.
Flag: Taiwan
Two regional indicator symbols, U+1F1F9 and U+1F1FC, i.e. the letters T and W. Every country or region flag with a two-letter code is built this way; England, Scotland and Wales use a tag sequence instead.
Thumbs up, medium skin tone
A base emoji plus a skin tone modifier (U+1F3FD). The five modifiers arrived in Unicode 8.0 in 2015.
Why you get an empty box or a question mark
An empty rectangle, often called tofu, or a box with a question mark in it, means your system knows there is a character here but the font on hand has no glyph for it. Nine times out of ten that is a version problem: the emoji is newer than the font on your device.
Unicode adds emoji roughly once a year, and a new set only reaches you when your operating system updates, which on older phones is never. That is also why the emoji list on this site stops at Emoji 15.1: anything newer is still a blank box on a lot of Android and Windows machines, and offering people something invisible to copy helps nobody.
Who decides that a new emoji exists
The Unicode Consortium's emoji subcommittee reviews proposals, and anyone may write one: the form and the criteria are public. They weigh evidence of demand, whether existing emoji already cover it, and whether it is too specific. Brands, specific people (living, dead or fictional) and deities are out by policy. Approved characters go into the next release, and each vendor then draws and ships them with a system update.
The names and search keywords are a separate dataset. Unicode's CLDR project maintains them per language, and that is where the English and Chinese names in this site's emoji list come from.
Can you just use emoji?
Two layers. The code points and names are a standard and free for anyone to type: putting 😀 in a headline, a product name or a button needs no permission. The artwork is not. Apple's emoji font is Apple's, and you cannot lift the images out of it for an app or an asset pack. If you need image files, use an openly licensed set: Twemoji is CC BY 4.0, Noto Color Emoji is under the SIL Open Font License.
Where to go next
If you came here to find an emoji rather than to read about them, the list below is searchable by name and keyword, and everything on it copies with one tap.