Guides · Published September 6, 2026
Everyone who reads fantasy can hear the difference between an elven name and a dwarven one. Very few people can say what the difference actually is. So we measured it. NamingRealm's generators run on hand-curated corpora — around 2,700 names across thirteen character-naming traditions — and those corpora are, in effect, a record of how the genre conventionally names each race. This article is what falls out when you count.
A note on what the numbers represent: these are statistics about our corpora, which capture the genre's conventions as practised across D&D, Tolkien-descended fantasy, and modern worldbuilding. They are not canon from any single source. That is what makes them useful — they describe the shared instinct, not one author's rulebook.
The Headline Table
Length is measured in letters, syllables are approximated by vowel groups, and vowel density is the share of all letters that are vowels (counting y). The "signature" column lists the two-letter clusters most over-represented in that tradition compared with all others — the sounds that are, statistically, its accent.
| Tradition | Avg length | Syllables | Vowel density | Favourite endings | Signature clusters |
|---|---|---|---|---|---|
| Draconic | 9.2 | 3.1 | 35% | -th, -on, -ax | ax, xo, vy |
| Arcane / Wizard | 8.9 | 3.2 | 38% | -en, -el, -la | my, yl, yt |
| Elven | 8.7 | 3.1 | 41% | -en, -or, -is, -el | ae, iw, dw |
| Vampiric | 8.4 | 3.0 | 37% | -ix, -ra | ct, tu, nv |
| Tiefling | 7.9 | 2.6 | 34% | -ix, -ra, -la | lp, py, hv |
| Dragonborn | 7.2 | 2.3 | 32% | -iv, -ar, -th | ek, bh, kh, kk |
| Medieval | 7.1 | 2.6 | 42% | -an, -in, -ic | ch, ul, ic |
| Halfling | 6.9 | 2.4 | 39% | -de, -er, -in | ck, wi, ic |
| Gnomish | 6.5 | 2.3 | 36% | -le (51%), -ix | dl, xl, ob |
| Orcish | 6.5 | 2.1 | 33% | -ak, -sh, -ra | zg, uk, kn |
| Goblin | 6.4 | 1.8 | 29% | -le (65%), -ix | zz, xt, iz, zl |
| Dwarven | 6.4 | 2.0 | 32% | -ak (38%), -na, -in | md, ln, mr |
| Norse / Viking | 6.0 | 1.9 | 34% | -ld, -ar, -rd | lj, bj, lf, jo |
Finding 1: Length Is Status
The length column is almost a social ranking. The three longest traditions — Draconic, Arcane, and Elven — are the genre's ancient, learned, and immortal. The three shortest — Norse, Dwarven, and Goblin — are its warriors, craftsmen, and vermin. A dragon's name is a title that takes a full breath; a viking's name is something you can shout over a shield wall. Fantasy authors have converged on this without anyone legislating it: syllable count signals age, learning, and rank.
For worldbuilders the lesson is direct. If you want a character to read as ancient or exalted, give them three syllables or more. If you want them to read as practical, grounded, or dangerous, cut to two.
Finding 2: The Melodic/Harsh Divide Is One Number
When readers describe elven names as "beautiful" and orcish names as "brutal," they are responding to vowel density. Elven and Medieval names are over 40% vowels; Goblin names are under 30%. That gap of a dozen percentage points is the entire perceived difference between song and snarl. Everything in between tracks the same way: Halfling and Arcane sit high and sound warm or scholarly; Dwarven, Dragonborn, and Orcish sit low and sound hard.
This is the single most useful dial in naming. Take any name, swap one vowel for a consonant cluster, and it moves down the scale toward the martial races. Add a vowel and it drifts back toward the courts.
Finding 3: Endings Are Tribal Badges
Some traditions are astonishingly consistent about how a name ends. Two-thirds of Goblin names in our corpus end in -le; so do half of Gnomish names. Thirty-eight percent of Dwarven names end in -ak. Vampiric names end in -ix or -ra seven times out of ten. By contrast, Medieval, Halfling, and Norse names spread their endings across dozens of options with no single one above 10%.
That contrast is meaningful. The invented races have a badge ending — a suffix that instantly marks membership — because fantasy authors need readers to identify a race from a name alone. The realistic traditions don't, because they are built from actual historical naming stock, and real cultures are messier. If you are inventing a culture from scratch, give it a badge ending. It is the cheapest possible way to make names feel like they belong together.
Finding 4: Consonant Clusters Are Accents
The signature-cluster column is where each tradition's accent lives. Some are obvious once you see them: Norse names lean on bj and lj (Bjorn, Ljot); Dragonborn on aspirated bh and kh (Bharash, Kharzul); Goblin on the buzz of zz and xt. Others are subtler — Elven names over-use ae and dw, Vampiric names lean on ct and nv, and Draconic names have a taste for ax and xo that no other tradition shares.
A cluster that appears ten or twenty times more often in one tradition than the rest is, for practical purposes, a phonetic passport. Drop one into a name and it declares a homeland. This is also why mixed names read as mixed: a name with an Elven ae and a Dwarven -ak is legitimately telling you the character has two heritages, whether the author intended that or not.
Finding 5: First Letters Are Less Diagnostic Than You'd Think
We expected initials to matter more. They do for a few traditions — 21% of Vampiric names start with V, 24% of Goblin names with S, 19% of Orcish names with G — but most traditions spread their first letters fairly evenly. The ear apparently weights the end of a name far more than the beginning, which matches what linguists find about word recognition generally. If you only have budget to make one part of a name distinctive, make it the ending.
How to Use This
Treat the table as a tuning chart. Pick the row that matches the feel you want, then check your candidate name against it: is it the right length, the right vowel density, does it carry a badge ending or a signature cluster? Our Fantasy Name Analyzer does that check automatically — type any name and it scores it against all thirteen traditions using the same corpora these numbers came from. And if a name falls short, the matching generator will produce a hundred that don't.
Method
Statistics were computed over the character-name corpora that power NamingRealm's generators (2,705 names across 13 traditions at time of writing). Names were lower-cased and stripped of non-letters. Syllables are estimated as the number of contiguous vowel groups. "Signature clusters" are letter bigrams appearing at least six times in a tradition, ranked by the ratio of their frequency in that tradition to their frequency across all corpora. The full corpus is available for download as CSV under CC BY 4.0 if you want to check the numbers or run your own analysis. For how the underlying Markov models work, see How Our Name Generators Actually Work.