Random Name Generator: Full Names That Fit Their Culture

Which tab to open first

A cast for a manuscript, NPCs the players meet in ten minutes and placeholder users for a demo need different kinds of names, so the tabs are split the same way.

How I use a random name generator to cast a story

A random name generator is quick at producing candidates and no help in choosing between them. The choosing goes faster with the routine I follow for every new character:

1. Set origin and gender first. The origin decides the surname and the grammar of the full name. Choose Unisex only when the gender should stay open on the page.
2. Narrow the sound. If your lead is Katherine (three syllables, K), filter her sister to a different syllable count and another starting letter.
3. Generate about twenty. With five you settle; with fifty you skim. A meter above the results shows how many combinations your filters leave, and the ↻ button on a card swaps that one name without touching the others.
4. Star three to five. The shortlist shows letters, syllables, initials and monogram side by side, and can be copied or exported as TXT or an Excel-safe CSV.
5. Read each one aloud with the surname, in the line where it appears most: "Agent Papadaki?" A name that looks fine can still trip the tongue.
6. Check the initials. Adam Samuel Smith is lovely until it is embroidered on a bag. The shortlist flags initials such as ASS, PIG or SAD.

Two rules I never break for a main cast: no two leads share a first letter, and characters who share scenes differ in syllable count. Readers recognise names by their outline, so Mara and Maya in one chapter blur into one person. The full checklist, with nicknames and famous names to avoid, is in how to name a character.

Full names that are right for their culture

The most common fault in generated casts is grammar, not taste: a Russian woman called Anastasia Ivanov, a Polish Agnieszka Kowalski. My reference for these conventions is the W3C article Personal names around the world; the generator applies these rules automatically.

W3C's own examples: Boris Nikolayevich Yeltsin's wife was Naina Iosifovna Yeltsina, with both surname and patronymic in the feminine. In Mao Ze Dong the family name is Mao. María José Carreño Quiñones, daughter of Antonio Carreño Rodríguez and María Quiñones Marqués, carries the first surname of each parent, father's first. When a rule shapes a result, the card shows a short hint such as "family name first in Japanese".

Two limits are worth knowing. Icelandic names are patronymics rather than surnames (Björk Guðmundsdóttir is Guðmundur's daughter; her brother would be a Guðmundsson), and many people in parts of southern India, Malaysia and Indonesia have a given name only. The generator does not produce Icelandic names and assumes a family name exists. Patronymics, bin and binti and sign-up forms are covered in name order and surnames around the world.

Why a batch never shows the same name twice

A name pool is smaller than it looks. With 20 German male first names and 20 surnames, First + Last allows 20 × 20 = 400 full names. Drawn independently, with repeats allowed, the chance of a repeat follows the birthday problem:

From 24 names on a repeat is more likely than not. A middle name drawn from the same 20 names matches the first name 1 time in 20, so 64 % of twenty-name batches would hold a "Hans Hans" (1 − 0.95^20 = 0.642). The generator draws without replacement: every full name differs, and the middle name is never the first name.

A live meter above the results shows how many combinations your filters leave: F for first names only, F × S for First + Last, F × (F − 1) for First + Middle and F × (F − 1) × S with a surname, where F is first names left after filters and S surnames for the origin (two surnames: S × (S − 1) instead of S). With F = 20 and S = 20 that is 400 and 20 × 19 × 20 = 7,600. Up to 40,000 combinations the meter counts distinct full names, so it can sit a little under the formula: unisex First + Last across every origin is 566, not 570, because Morgan Jones, Morgan Williams, Morgan Thomas and Morgan Lewis appear in both the English and the Welsh list. Above 40,000 it shows the formula. Ask for 50 when only 20 exist and you get those 20 with the note "Only 20 combinations match these filters", not a list padded with repeats.

Placeholder people who cannot reach anyone

Demo data escapes: a staging server sends a real email, a novel prints a number and a reader dials it. The Test persona tab uses only addresses and numbers set aside for this; pick a country and the names, area codes and phone ranges follow it.

• Email: example.com, example.net and example.org only, reserved by RFC 2606, which also reserves .test, .example, .invalid and .localhost. Domains such as inbox.demo are not reserved, so anyone can own them.
• Phone: only the ranges regulators publish for drama and fiction, compiled on the fictitious telephone number list. In North America that is 555-0100 to 555-0199, not any 555 number. UK: London 020 7946 0xxx, mobiles 07700 900xxx; Paris: 01 99 00 xx xx. Where no range exists (Spain, Italy, Brazil, Japan) the phone stays empty.
• Birthday and age: ages run from 18 to 75, and the birth date is picked so the stated age is true today; "this year minus age" with a random month makes many people a year younger than claimed and breaks age-gate tests.
• Awkward names: Liam O'Brien gets liam.obrien, Jürgen Müller juergen.mueller; when two people in one batch would share an address, the second gets a two-digit number.
• Copying: one persona copies as a JSON object or a CSV row with 14 fields, and a whole batch exports as CSV or JSON.

Phone ranges by country and tricky test names are in safe fake names, emails and phone numbers.

The rules behind each name, and the cases I check them with

Each rule above is a small function tested against fixed cases before release, with the sources linked above; Polish and Greek feminine forms are standard grammar. Age is whole years completed today, and a birthday counts on the day itself. The name lists are hand-picked, not weighted by frequency, so a rare name turns up as often as a common one, and nothing carries a popularity rank, because without birth-registration data it would be invented. Syllables are an estimate from vowel groups: ia, io, iu and oa count as two (Liam, Noah), a silent final e does not count (Jane, Rose) and y is a consonant at the start before a vowel (Yuki). The initials check looks for any of sixteen short words, such as ASS, PIG, SAD or WTF, anywhere in the initials.