Explore two surname parts with measured US frequency
A surname can change a fictional person's full-name rhythm in one word. This tool pairs an optional first name with two independently selected surname parts. Each surname part is in the United States Census 2010 surname table; their published ranks are used for the popularity bands. The generator does not manufacture a new surname and then label it as a Census record. A two-part surname may be appropriate for a character or a creative exercise, but it does not establish that two families, places or historical traditions are linked. Make that connection in your own context, if one is intended.
Family name 4 and Family name 5 each have their own beginning, ending, popularity and origin boxes. Set the first part's letters under Family name 4 and the second part's letters under Family name 5. The generator applies them to those exact words before combining them. For instance, a rare first surname part and a common second part require two different Census rank bands. Any removes a part's rank filter. If either part has no matching record, the tool returns no data for the whole request; it never substitutes a convenient but unverified surname.
Use sound menus as preferences, not etymology
Nature favours visible landscape-like fragments, Craft favours visible occupation-like fragments, and Place favours familiar location-like endings among real Census surnames. These are spelling cues. A name containing “wood” does not prove that a particular family came from a woodland; an occupation-looking word does not document an ancestor's trade. Crisp, Lyrical and Historic similarly change ordering by length or ending sound without assigning a verified age to the name. The character-impression menu is creative. Fictional species styling starts from recorded surname parts, then changes their interior sounds and adds an epithet. The styled result itself is not a Census record.
The surname table is a snapshot of people counted in 2010, not a surname-by-birth-year series. The birth-year field is unavailable here; an API request that supplies a year gets an explicit error. The country menu shows the number of exact Census surname spellings with a matching country-use record from the open Onomaverse release or Wikidata. Some choices have no indexed match and say so before you generate. Christian, Muslim, Jewish and Hindu selections use only explicit historical or etymological surname association descriptions from that release. They do not identify the faith of anyone with the surname. Atheist and Agnostic mean no preference; Jedi is fictional styling. Try a strong first-name and surname combination in dialogue, a formal address, a cast list and a possessive phrase. Compare the words as readers will encounter them, then research any real cultural or historical claim separately.
Sources: US Census 2010 surname table, Onomaverse surname-frequency v2026.10 (CC BY 4.0), Uplook Australian top-50 ABN-name analysis, and Wikidata language-of-use statements (CC0). Coverage is incomplete, and name-use evidence does not establish any individual's ancestry or religion.
Understand what a match proves
Given names on this page come from registered births in England and Wales between 1996 and 2025. The Office for National Statistics publishes a name, year, recorded Boy or Girl category, count and rank when the count is high enough to release. A missing entry is not evidence that nobody used a name: very small counts are suppressed, spellings differ, and a year outside the published period has no annual record here. The source records births, not the number of people alive today. If you select a year, it must fall within that annual series. The US Social Security Administration's 2020–2025 top-name table is a combined period and cannot honestly be treated as a single year.
Surnames on the last-name page come from the 2010 United States Census frequency table. That table counts people with a surname in a census, which is a different population and date from the birth-name records. Its rank bands describe how common a surname is in that US table, not how often it was given to babies in a particular year. Birth year is unavailable on the surname form because no surname-by-birth-year series is held. If a name part has a popularity slider, choosing Any removes that rank limit. Values one through five use the rank bands explained beside the slider; they are an editorial mapping over published ranks, not a number printed by the source site.
Origin boxes use explicit Wikidata language-of-use statements when the indexed name also appears in the relevant birth or surname table. Several languages can be attached to one name. Multiple checked boxes mean any checked context may match that one part, while the rules for separate name parts all apply. A language label is not a claim about the nationality, ancestry, religion or identity of anyone using that name. For given names, British context uses ONS births, American a separate SSA aggregate, Canadian Statistics Canada births, and French INSEE births; other country choices depend on sparse Wikidata links. For Last, open surname country-use records are intersected with the US Census surname list. All country labels mean observed name use, not citizenship; missing index coverage is stated on the form and an unmatched request reports no data.
Christian, Jewish, Muslim and Hindu selections intersect recorded given names with cited lists of biblical, Quranic and Hindu religious figures. A shared name may appear in more than one tradition. Last also has a small set of surnames whose published origin descriptions explicitly cite a religious naming tradition. This is a documented association for the spelling, never an estimate of the faith of someone with the name. Atheist and Agnostic mean no religious preference because there is no special secular name stock. Jedi is a fictional style cue, not a real-world religious dataset. On mixed given and family name forms these choices apply to given parts only. Source links and limits are recorded in the provenance ledger.
Make a careful shortlist
Try a single clear constraint first, inspect the resulting names, and add a second constraint only when the first still leaves useful choices. A request for a rare rank band, several letters, a narrow length and a particular documented context can legitimately have no published match. The tool does not silently drop a preference to fill a requested count. Start with one result to see exactly what changed, then ask for ten or more when you know the filters describe a real pool. The same seed and options produce the same order, which helps two people compare a shortlist. Save favourites in your browser if you want to return to them, and always check a final choice in the community and setting where it will be used.