C++23 header-only libraries compiled to WebAssembly with Emscripten. Everything runs in your browser — no server required. Try personality traits and Big Five scores.
Generates a complete fictional persona with persona-generator, which composes eight libraries through entity-generator fields: country, city, birthday, body, name, personality, religion, hobbies, favourite genres and occupation in one call. Each field is conditioned on the ones it depends on (the country sets the city, the name culture, the body and the job; the age and sex set the rest), and the persona's seed replays all of it. This is fictional sample data, not a description or prediction of a real person. GitHub →
Runs the persona, address, phone, email, device, vehicle and weather libraries for one country with one call seed. The persona sets the country, name, age and city; the rest follow from it (the email is built from the persona's name and birth year, the weather is the persona's city today). It is the demos below composed in the page, not a pipeline: the vehicle-generator and weather-generator are not part of persona-generator yet. A part the country has no data for is marked, not hidden. Fictional sample data only.
Who a persona is: names, birth, body, mind and livelihood.
Generate real-world first names and surnames from culture/gender-specific lists. GitHub →
Culture filters all name parts. Gender filters first names, not surnames. Repeated alternatives are possible.
Generate synthetic birthdays using the library's population data and a reference date. GitHub →
Defaults to today's local date. Each draw generates a birthday and its completed-year age as of this date; changing the date does not age an existing profile. Population-model life expectancy is not an individual forecast.
Generate synthetic physical characteristics — height, weight, eye/hair color and blood type — from the library's population model, not measurements of real people. GitHub →
Compare male and female inputs with the same country and seed. “Any” country is resolved once for both. This demo leaves the optional age input unset, so no age effects apply; differences are synthetic draws, not medical evidence.
Generate fictional personality traits and resolved Big Five scores. Scores are centered on zero before trait effects, not percentages or a personality assessment. GitHub →
Use 1–16 digits (0–9, a–f), without 0x. Replay with the output seed and the same trait count.
Each score starts with a normal draw (mean 0, standard deviation 1) and adds the selected traits' catalogue loadings. Base values also influence trait selection; exclusive groups, conflicts and synergies constrain the draw. These are unbounded model values, not measured traits or percentiles.
Generate fictional religion, hobbies and favourite genres using country, age and sex inputs. Compare age inputs while holding the country, sex input and seed fixed. Outputs are synthetic, not predictions about individuals or groups. GitHub →
Comparison uses ages 8, 30 and 72, regardless of the single-draw age. Unknown sex applies no sex factor. “Any” country is resolved once; religion is country-weighted, and age factors/limits apply to hobbies and genres. Results can coincide.
Generate a fictional employment status, education level, job and gross annual income from country, sex and age inputs, using ILO labour force and earnings statistics. Compare the four education levels while holding the country, sex and seed fixed. Outputs are synthetic, not predictions about individuals or groups. GitHub →
Comparison fixes age 35 and an employed status, and varies only the education input. “Any” country and sex are resolved once. Under 15 a person is always a student with no job. Status and education come from measured rates (139 and 185 surveyed countries; others use a regional mean); the job catalogue is curated. Results can coincide.
Where they live, from countries down to street addresses, the weather there and the car they drive.
Sample countries with geography, politics, culture and telecom metadata. Compare population-weighted and uniform selection within the same region filter. See a country's locales, time zones and public holidays. GitHub →
Single draw is population-weighted. Batches draw with replacement: repeated countries count again. Each pair uses the same random seed. Uniform gives every eligible catalogue row equal probability; weighted uses max(population, 1). Observed shares are samples, not exact population proportions.
Each locale formats today's date, 15:04, 1234567.891 and 1,234.50 of its currency with its own CLDR patterns. Zone offsets and local times are for now, from the tzdata rules built into the library, not the browser's. Holidays are national days from rules fitted to python-holidays. Generate picks the drawn country here.
Sample cities from the embedded GeoNames lite catalogue. Compare population-weighted and uniform selection within the same country filter. GitHub →
Single draw is population-weighted. Batches draw with replacement: repeated cities count again. Each pair uses the same random seed. Uniform gives every eligible catalogue row equal probability; weighted uses max(population, 1). Observed shares fluctuate and cover the lite catalogue, not every city or resident.
A year of daily weather for any point on land, drawn around the CRU TS 1991-2020 normals of the nearest cell: wet and dry spells, warm and cold runs, rain, snow, thunder, cloud, wind and the sun. Click the map to choose a place, then pick a day on the chart for its 24 hours. GitHub →
Each year is one seeded series: a day depends only on the seed, the place and the date, so the day's hours below are that day of the year above. The dashed line is the cell's normal mean temperature. Points more than 100 km from land have no weather. The library can also draw 1991-2020 as it was, from a 23 MB table this page does not load.
Generate fictional postal addresses with real street names (from OpenStreetMap, weighted by how many addresses each street has), in each country's own layout and postcode format. The street data is a data pack: the page loads the shared tables at start and a country's streets only when you ask for that country, so nothing you do not use is downloaded. Outputs are synthetic. GitHub →
“Any” picks a country uniformly in this demo, so its streets can be fetched first; the library itself draws countries by population. Compare scripts draws one call seed and shows the address in the country's own layout and in the Latin one; only 14 countries have a Latin layout, and a street keeps its local name when the data has no Latin one.
Generate national numbers and documents that fit a person and a date: each number built from the birth date, sex, place and names where its country's format encodes them, and a passport or identity card with its machine-readable zone. Fields only, as a table; the library makes no image of a document. GitHub →
Test data. Without the safe option a number is structurally valid and may match a real person's; nothing is looked up. Never use these to identify anyone, to open an account or to pass an identity check. The safe option means something different per scheme: see the limits.
The person is an adult drawn for the country, with a name, and “Any” country is drawn by population among those with names. A country may have no national number at all, or none for that person yet. A country without an identity card gives no card.
Generate bank accounts and payment cards that fit a person, a country and a date: an IBAN or national account with its bank, branch and check digits right, and a card in a network sold there. Fields only, as a table; the library makes no image of a card, and no magnetic stripe, chip or PIN data. GitHub →
Test data. Without the safe option an account is structurally valid and may match a real one, and a card number from a network's range may match a real card; nothing is looked up. Never use these on a live payment system or to pass a payment check: trying card numbers on a live system is card testing, which is fraud. For many accounts the safe option promises nothing: see the limits.
The person is an adult drawn for the country, with a name, and “Any” country is drawn by population among those with names. A number in a network's range cannot be asked for with the safe option on. A test number has the sandbox's own expiry and security code.
Generate a household's home in a country: the dwelling type by area and household size, tenure by household type, age and dwelling type, rooms, bedrooms, floor area and year built, and, given an income, the monthly cost, whose share of the income falls as the income rises. Sources are census and survey tables (EU-SILC, US ACS and AHS, national censuses), the UN housing compendium and DHS; where a country has none, a similar country's figures fill in, and the card says which. GitHub →
The earner is one employed person of the reference person's age in the country, drawn by occupation-generator; a household's other earners are not counted here (the persona card counts them all). No home is built after the date. A household whose type and size the country's table does not hold is drawn again from the type, then the size.
Generate a company that fits its country: the industry, size and head count, legal form and founding date follow the country's business statistics, and the name, registration numbers (register, tax, VAT, LEI) and domain follow from them. Names are in the country's business-name languages and scripts. The safe option gives numbers no company holds and .example domains; realistic numbers pass their check digits and may belong to a real company. Nothing is looked up, and a name can coincide with a real firm's.
GitHub →
A forced industry, size or legal form that no company of the country can have is reported as an error. The legal forms offered are the chosen country's.
Generate a car or motorcycle someone owns in a country at a date, with parts that agree: a make sold there and a model built in its model year, a fuel as common as it was among that year's new vehicles, a trim, gearbox and drive (a motorcycle's final drive is a chain, belt, shaft or hub) the model was offered with, a colour of its era, an odometer that fits its age, a licence plate in the format the country issued on the registration date, and a VIN with the maker's prefix, check digit and model-year code. The safe option gives plates no authority issues and VINs whose maker prefix NHTSA has no record of. Outputs are synthetic; nothing is looked up in a register. GitHub →
“Any” country is drawn by the vehicles in use of the chosen kind, and “Any” kind by the country's mix of cars and motorcycles. A body belongs to a kind, so picking a kind limits the body list. The plate regions are the country's: US states, Canadian provinces and Australian states pick that jurisdiction's plates, and German districts, Swiss cantons or Japanese offices the code the plate carries; without a country, one that has the region is drawn. A motorcycle has no doors, and petrol mopeds of 50 cc and under are not drawn. The date (today by default, 2000 to 2050) sets the fleet's age, the fuels and the plate format of each registration. Compare countries draws one call seed in Germany, the United Kingdom, Spain, the United States and Japan.
How you reach them: handles, phone numbers, email, devices and accounts.
Generate gamer-style nickname alternatives — with an optional name input or from wordlists. GitHub →
Blank name uses wordlists. With a name, the library uses it in one in four draws on average; other draws still use wordlists. No forced style or guaranteed name derivation. Repeats are possible.
Generate fictional phone numbers that are valid in a country's numbering plan, drawn from libphonenumber's mobile and fixed-line ranges and written the way the country writes them. The safe option uses the numbers regulators reserve for fiction (for example 555-01xx in the US and Canada) where a country has such a range; elsewhere it avoids the allocated ranges. Outputs are synthetic. GitHub →
“Any” country is drawn by population. Compare safe draws one call seed and shows the number with and without the safe option; the type may differ when the country has a reserved range for only one type. “Reserved for fiction” is Yes only for numbers from a regulator's range.
Generate fictional email addresses in a country's style from a name, nickname and birth year, with the providers people use there. Leave the names empty for a handle. Compare the same name and seed across eight countries. The safe option uses the reserved example domains. Outputs are synthetic. GitHub →
Names are folded to ASCII (Ø is o) and family-name particles are skipped (da Silva is silva). A style is only used when the inputs can fill it: no birth year, no year in the address. “Any” country is drawn by population. Comparison uses the same call seed in US, DE, BR, JP, CN, RU, FR and EG.
Generate the device someone uses to reach a service, with parts that agree: a model on sale at the date, an OS version it runs, a browser that runs on that OS at a version released by then, the user agent and client hints that exact combination sends, its screen, and an IP address registered in the country. The safe option gives documentation addresses only (RFC 5737, RFC 3849). Outputs are synthetic. GitHub →
“Any” country is drawn by internet users. The date defaults to today and picks the models, OS and browser versions out by then; a date after the data's newest release is seen as that release's date. The locale only sets the Accept-Language header. Compare dates sees one call seed in June 2021 to 2024 and today: the country, platform and address stay, the model and versions move. A registered address belongs to the country's blocks and may be in use.
Generate fictional user accounts for apps and test data: a person with a username, email, phone number and postal address that agree on the country, plus a password, sign-up and last-login times, a platform, a locale and a time zone, and, if asked, the device of the last sign-in (from device-generator) with its user agent and IP address. The safe option uses reserved example domains, where a country has one a phone number from the range set aside for fiction, and a documentation IP address for the device. Outputs are synthetic; the passwords are test data, never real credentials. GitHub →
The 100 countries listed are the ones every part has data for. “Any” picks one uniformly in this demo, so its streets can be fetched first; the library itself draws by population. Ages are as of today. The device is the one of the last login (the sign-up for an account that never logged in), on the account's platform and locale; asking for it changes nothing else. Compare safe draws one call seed and shows the same person with and without the safe option.
Families and friends, and the engine that composes everything else.
Generate the people around a person: parents and grandparents, siblings and half-siblings (twins included), partners, children and grandchildren, with ages, deaths and surnames that follow the country's naming convention. The numbers come from UN fertility, marriage and household statistics. GitHub →
The person is the one the inputs describe; what is left out is drawn. Ages are as of 1 July 2023, the date of the UN estimates. The same seed and inputs give the same family. Relatives who have died are listed with the age they died at. Stepparents (a parent's new partner, the children they had with the parent and their own children, the stepsiblings) and the household (who the person lives with) are off unless chosen.
Composable entity generation with typed components. This demo generates an RPG character sheet. GitHub →