How emoji work

An emoji on screen looks like one thing. Inside the text it is almost never one codepoint. Modern emoji are sequences: a base character optionally followed by a presentation selector, optionally followed by a skin-tone modifier, optionally followed by a Zero Width Joiner that glues it to another emoji to form something composite. Flags are pairs of letters. The Spanish flag and the family of four are both built by the same machinery — a set of conventions agreed by font vendors and codified by the Unicode Technical Committee.

Emoji vs text presentation

Many emoji codepoints predate the emoji era. The HEAVY BLACK HEART ❤ at U+2764 was added in Unicode 1.1 in 1993, decades before phones rendered it as a colourful pillow. The standard distinguishes two presentations for these dual-purpose codepoints:

Text presentation

Black-on-paper, follows the surrounding font. Forced by appending U+FE0E (VARIATION SELECTOR-15).

Emoji presentation

Colourful, font-independent emoji glyph. Forced by appending U+FE0F (VARIATION SELECTOR-16).

U+2764           →  ❤   (default presentation, font-dependent)
U+2764 U+FE0E    →  ❤︎  (text — VS15)
U+2764 U+FE0F    →  ❤️  (emoji — VS16)

Each emoji codepoint has a default presentation listed in emoji-data.txt. The pure-emoji codepoints (the ones added specifically for emoji use, in the U+1F300–U+1F9FF range and elsewhere) have Emoji_Presentation = Yes and need no selector. The dual-use codepoints have Emoji_Presentation = No and need a VS16 if you want the colourful form to appear on systems that otherwise default to text.

Skin tone modifiers

The five skin tone modifiers were added in Unicode 8.0 in 2015. They are codepoints U+1F3FB through U+1F3FF, named after the Fitzpatrick scale (a dermatological classification of human skin types developed by Thomas B. Fitzpatrick in 1975). Applied to a human emoji, the modifier replaces the default yellow tone with the chosen skin colour.

Codepoint	Name	Fitzpatrick type	Glyph
U+1F3FB	EMOJI MODIFIER FITZPATRICK TYPE-1-2	I-II (light)	🏻
U+1F3FC	EMOJI MODIFIER FITZPATRICK TYPE-3	III	🏼
U+1F3FD	EMOJI MODIFIER FITZPATRICK TYPE-4	IV	🏽
U+1F3FE	EMOJI MODIFIER FITZPATRICK TYPE-5	V	🏾
U+1F3FF	EMOJI MODIFIER FITZPATRICK TYPE-6	VI (dark)	🏿

👋             U+1F44B  WAVING HAND          →  yellow default
👋🏽            U+1F44B U+1F3FD                 →  medium skin tone

UTF-8 bytes for 👋🏽:
  F0 9F 91 8B   F0 9F 8F BD     (8 bytes for 2 codepoints, 1 grapheme)

Not every human emoji accepts a modifier — the ones that do are listed in emoji-data.txt under the property Emoji_Modifier_Base. The fonts that ship with iOS, Android, Windows, and major Linux desktops all implement the modifier mechanism; a system that lacks the ligature will show the base hand and the colour swatch as two separate glyphs side by side.

Zero Width Joiner sequences

The Zero Width Joiner at U+200D is the most generative piece of emoji machinery. Originally introduced in Unicode 1.1 for Arabic and Indic ligatures, it was repurposed in 2010 for emoji: a sequence of emoji separated by ZWJs becomes a single composite emoji if and only if the font has a ligature for that exact sequence.

The family-of-four emoji is the canonical example:

👨‍👩‍👧‍👦   FAMILY: MAN, WOMAN, GIRL, BOY

Codepoint sequence:
  U+1F468  MAN
  U+200D   ZERO WIDTH JOINER
  U+1F469  WOMAN
  U+200D   ZERO WIDTH JOINER
  U+1F467  GIRL
  U+200D   ZERO WIDTH JOINER
  U+1F466  BOY

That is 7 codepoints.

UTF-8 bytes:
  F0 9F 91 A8  E2 80 8D  F0 9F 91 A9  E2 80 8D
  F0 9F 91 A7  E2 80 8D  F0 9F 91 A6
  — 25 bytes total. One grapheme cluster.

Other widely supported ZWJ sequences:

Glyph	Sequence	Codepoints
🏳️‍🌈	WHITE FLAG + VS16 + ZWJ + RAINBOW	U+1F3F3 U+FE0F U+200D U+1F308
🏴‍☠️	BLACK FLAG + ZWJ + SKULL AND CROSSBONES + VS16	U+1F3F4 U+200D U+2620 U+FE0F
👨‍💻	MAN + ZWJ + LAPTOP COMPUTER	U+1F468 U+200D U+1F4BB
👨‍👩‍👧	FAMILY: MAN, WOMAN, GIRL	U+1F468 U+200D U+1F469 U+200D U+1F467
👩🏽‍🚀	WOMAN + skin tone IV + ZWJ + ROCKET	U+1F469 U+1F3FD U+200D U+1F680
👨‍❤️‍💋‍👨	KISS: MAN, MAN	U+1F468 U+200D U+2764 U+FE0F U+200D U+1F48B U+200D U+1F468

On a font that supports the sequence, the rendered output is a single composite glyph. On a font that does not, you see the individual parts separated by small gaps where the ZWJs are invisible. The Unicode standard does not require fonts to support any particular ZWJ sequence — the list of recommended sequences is maintained in emoji-zwj-sequences.txt and grows with each release.

Flag emoji and Regional Indicators

Country flags use a different mechanism. Unicode does not assign a separate codepoint to each national flag (the political minefield was deliberately avoided). Instead, there are 26 Regional Indicator Symbol codepoints — U+1F1E6 through U+1F1FF — one for each Latin letter A through Z. A pair of Regional Indicators forms a flag if the resulting two-letter sequence corresponds to an ISO 3166-1 alpha-2 country code that the font has a flag for.

🇺🇸   UNITED STATES FLAG
     = Regional Indicator U  +  Regional Indicator S
     = U+1F1FA U+1F1F8

🇪🇸   SPAIN FLAG
     = Regional Indicator E  +  Regional Indicator S
     = U+1F1EA U+1F1F8

🇯🇵   JAPAN FLAG
     = Regional Indicator J  +  Regional Indicator P
     = U+1F1EF U+1F1F5

The Regional Indicators on their own render as styled letters (often as a letter inside a box). The pair-rendering is, again, a font ligature. Windows famously did not ship flag glyphs in Segoe UI Emoji for years — the underlying codepoints were there, but the font drew them as letter boxes, producing the now-familiar "ZW JP" rather than 🇯🇵. Windows 11 finally shipped flag glyphs in 2022.

For regional flags below national level — Scotland, England, Wales, Texas — Unicode introduced a separate mechanism in Unicode 10.0 (2017): the Tag Sequence. A black flag U+1F3F4 is followed by a sequence of tag characters from U+E0020–U+E007E spelling out the ISO 3166-2 subdivision code, terminated by U+E007F CANCEL TAG.

🏴󠁧󠁢󠁥󠁮󠁧󠁿   ENGLAND FLAG
     = U+1F3F4 U+E0067 U+E0062 U+E0065 U+E006E U+E0067 U+E007F
       (black flag + g + b + e + n + g + cancel tag)
       — that is "gb-eng", the ISO code for England.

Counting emoji

The simplest emoji is one codepoint. The compound family is seven. A kiss with two skin-tone modifiers — 👨🏽‍❤️‍💋‍👨🏿 — is ten codepoints and 35 UTF-8 bytes. A user sees one image; "👨🏽‍❤️‍💋‍👨🏿".length in JavaScript is 17 (because UTF-16 needs surrogate pairs for each SMP codepoint, and the ZWJs and VS16 are single units). The only correct way to count emoji-as-perceived is grapheme-cluster segmentation. See codepoint, character, glyph, grapheme for the four-way framing.

What goes wrong

Most emoji failures fall into a small number of categories:

• Missing ZWJ ligature. The font has the individual emoji but not the composite. The compound shows up as four people in a row instead of one family. The codepoints are correct; the font simply lacks the substitution rule.
• Missing variation selector handling. An old system displays a heart as a black outline because it does not honour U+FE0F. Adding the selector explicitly is harmless on modern systems and helpful on old ones.
• Truncation inside a sequence. A backend that truncates by codepoint can leave a dangling ZWJ at the end of a string, which then renders as a small box. Truncate by grapheme cluster instead.
• Surrogate splits. Code that operates on UTF-16 code units rather than codepoints can split a single SMP emoji into a high surrogate and a low surrogate, neither of which is valid on its own. Use codepoint iteration (for…of in JavaScript, iter() in Python).

For inspection, the character inspector shows every codepoint inside an emoji with its name and category; the UTF-8 encoder shows the byte counts. The Emoticons block page lists the original 80 smiley faces added in Unicode 6.0.