Skip to content

Generate a language

Starting words made from the sounds and syllables you pick, one for each meaning of a list, and the project they start.

Generate a language makes the starting words of a new project. You pick the sounds and the shape of the words, and Lemula gives a word to each meaning of a list. The project starts from those words, with an inventory of their sounds and a Latin orthography.

In the Create a project dialog, choose Generate a language, name the project and its first stage, and press Choose sounds. The generator opens in a window of six steps. Its tabs go to any step in any order, and its footer shows how many sounds there are and how many distinct words the settings allow. Closing the window takes you back to the dialog with the settings kept; Cancel there drops them.

Sounds

Start from one of three presets, or from sounds of your own. A preset fills every step: the sounds and how often each is used, the syllable templates, the word lengths and the stress. Any of them can be changed after.

Pick on the IPA chart opens the consonant and vowel charts, where a click adds a sound or takes it out. Each sound has a slider for how often it is used. Its share is counted among the consonants or among the vowels, since a consonant's place in a syllable only ever takes a consonant. Sample syllables show what the sounds make, and Other samples draws more.

Syllables

A word is built from three syllable templates: one for its first syllable, one for its middle syllables and one for its last. A template writes C for any consonant, V for a vowel, N for a nasal, L for a liquid and G for a glide. A letter in parentheses may be left out, so (C)V(N) makes a, ta, an and tan. A word of one syllable starts as a first syllable and ends as a last one.

Under each template are the consonants it allows before and after the vowel. A click bars a sound there, such as ŋ at the start of a word or voiced stops at its end, and a second click allows it again. Each template shows sample syllables and how many distinct syllables it makes.

Stress falls on the first, the penultimate or the last syllable of every word of two or more syllables, or is not marked. The first stage of the project takes it as its fixed stress, so borrowings are stressed the same way.

Word length

Each part of speech of the meaning list has a shortest and a longest length in syllables, and the length most of its words have. A length one syllable further from that one has a third as many words. The table shows how many meanings of each part of speech the list has, and how many distinct words the settings allow for it.

Meanings

The Short list holds 45 basic meanings, chosen from the concepts the Swadesh list and the Leipzig–Jakarta list share: twelve body parts, nouns of nature, common verbs and adjectives, pronouns, "this", "not" and "one". The Long list holds 1,416 meanings in 24 fields. It needs room for that many entries in a lexicon, which the Hobby plan has. Each word takes its meaning's gloss and part of speech as its features.

Some meanings can be made as compounds of others, as many languages make a tear from eye and water. Choose how many: the generator takes that many at random from the compounds whose parts are all in the list, and a seed decides which ones and which recipe each follows. A compound counts as a change, so compounds take at most a quarter of a plan's changes, 12 on the Free plan. The Short list can make 10.

Where the list needs more words of a part of speech than the settings allow, the step says so, and the project can't be created until the settings make room.

Lexicon

The Lexicon step shows every word with its form, its spelling and, for a compound, the words it is made from.

  • The seed decides the words: the same settings and seed always give the same words. Type a seed to go back to one.
  • Regenerate all draws a new seed and makes every word that isn't locked again.
  • The lock keeps a word as it is, whatever else changes.
  • The button at the end of a row makes that word again. It never comes back to a form it had.
  • Show lists all the words, the words of their own, or the compounds.

No two words sound the same, stress aside.

Changes

The Changes step gives the line some sound change before its second stage. It is optional: with no pattern picked, the project is made from the lexicon alone, and Create project works from the Lexicon step as well.

A sound change pattern is a few sound changes of one kind from the history of real languages, such as the lenition of Western Romance or Grimm's law. Its rules are written with features and classes, so it fits any sounds. Each card names the languages the pattern recalls and a real word before and after a change of its kind.

Pick one or more patterns. They run in the order you pick them, after the compounds, and each reads the words as the one before it left them. The number on a card is its place in that order; a second click takes the pattern out. Under the cards, each picked pattern lists its sound changes with how many words each one changes and a few of them before and after. A sound change that changes no word is left out.

The second stage takes the name you give it under Next stage, Old and the project's name until you change it. The step shows the words the patterns change, as the first stage has them and as the second stage will.

The sound changes count against the plan's changes together with the compounds, and the step shows how many are left. On the Free plan all eight patterns fit beside 12 compounds.

What the project starts with

Create project makes a project with one line, named after the project:

  • The first stage starts from a lexicon of the generated words, each with its gloss and part of speech.
  • An inventory of the sounds is attached to the first stage.
  • The first stage's word shape holds how often each sound is used, the three syllable templates with the sounds they bar, the word lengths and the seed.
  • A Latin orthography writes each sound as the common practice of Latin with digraphs does, ʃ as sh and ŋ as ng. Where two sounds would be written alike, the more used keeps the letters and the rarer takes a diaeresis: with ə used more than e, ə is written e and e is written ë.
  • Each compound is a compound change in the stretch to the second stage, before any sound change. Its main stress is on its first part, and it changes with the words it is made from.
  • The sound changes of the patterns picked follow the compounds in that stretch, in the order the patterns were picked, and the second stage takes the name given on the Changes step. Each is a sound change of one rule, which you edit like one you write.

Plan limits apply: the starting words are entries of the lexicon, and each compound and each sound change is a change.

Sources

The meaning lists are adapted from the sources below, each under the Creative Commons Attribution 4.0 licence. The Long list, its fields and its parts of speech come from the World Loanword Database (WOLD), and so do the compounds and how many languages make each. WOLD's list follows the chapters and numbering of the Intercontinental Dictionary Series. The Short list is chosen from the Swadesh and Leipzig–Jakarta lists, linked to WOLD's meanings through Concepticon.

  • Haspelmath, Martin & Tadmor, Uri (eds.) 2009. World Loanword Database. Leipzig: Max Planck Institute for Evolutionary Anthropology. wold.clld.org. CLDF dataset lexibank/wold v4.0, doi:10.5281/zenodo.5139859.
  • Key, Mary Ritchie & Comrie, Bernard (eds.) 2015. The Intercontinental Dictionary Series. Leipzig: Max Planck Institute for Evolutionary Anthropology. ids.clld.org.
  • List, Johann Mattis, Tjuka, Annika, Blum, Frederic, Kučerová, Alžběta, Ugarte, Carlos Barrientos, Rzymski, Christoph, Greenhill, Simon & Forkel, Robert (eds.) 2025. Concepticon 3.4.0. Leipzig: Max Planck Institute for Evolutionary Anthropology. concepticon.clld.org, doi:10.5281/zenodo.14923393.
  • Swadesh, Morris. 1955. Towards greater accuracy in lexicostatistic dating. International Journal of American Linguistics 21(2). 121–137.
  • Tadmor, Uri. 2009. Loanwords in the world's languages: Findings and results. In Haspelmath, Martin & Tadmor, Uri (eds.), Loanwords in the world's languages: A comparative handbook, 55–75. Berlin: De Gruyter Mouton.

The compound counts draw on 40 of WOLD's 41 vocabularies, each cited in full at wold.clld.org/contributions: Swahili (Thilo Schadeberg), Iraqw (Maarten Mous), Gawwada (Mauro Tosco), Hausa (Ari Awagana, H. Ekkehard Wolff, Doris Löhr), Kanuri (Doris Löhr, H. Ekkehard Wolff, Ari Awagana), Seychelles Creole (Susanne Michaelis, Marcel Rosalie, Katrin Muhme), Romanian (Kim Schulte), Selice Romani (Viktor Elšík), Lower Sorbian (Hauke Bartels), Old High German (Roland Schuhmann), Dutch (Nicoline van der Sijs), English (Anthony Grant), Kildin Saami (Michael Rießler), Bezhta (Bernard Comrie, Madzhid Khalilov), Archi (Marina Chumakina), Manange (Kristine Hildebrandt), Ket (Edward Vajda, Andrey Nefedov), Sakha (Brigitte Pakendorf, Innokentij Novgorodov), Oroqen (Fengxiang Li, Lindsay J. Whaley), Japanese (Christopher K. Schmidt), Mandarin Chinese (Thekla Wiebusch), Thai (Titima Suthiwan), Vietnamese (Mark Alves), White Hmong (Martha Ratliff), Ceq Wong (Nicole Kruspe), Indonesian (Uri Tadmor), Malagasy (Alexander Adelaar), Takia (Malcolm Ross), Hawaiian (ʻŌiwi Parker Jones), Gurindji (Patrick McConvell), Yaqui (Zarina Estrada Fernández), Zinacantán Tzotzil (Cecil H. Brown), Q'eqchi' (Søren Wichmann, Kerry Hull), Otomi (Dik Bakker, Ewald Hekking), Saramaccan (Jeff Good), Imbabura Quechua (Jorge A. Gómez Rendón), Kali'na (Odile Renault-Lescure), Hup (Patience Epps), Wichí (Alejandra Vidal, Verónica Nercesian), Mapudungun (Lucía Golluscio, Adriana Fraguas, Fresia Mellico).

Changes made

Lemula changed WOLD's list as follows:

  • Glosses are rewritten so each is unique and reads as a word's gloss: nouns and adjectives bare, verbs with "to", and a qualifier in parentheses for English homonyms, as in "spring (water)".
  • 44 meanings are dropped: 26 that name things of one region, the seven days of the week, six that would repeat an English gloss, three documents of a modern state, and two houses of worship of one religion.
  • Pronouns and question words are moved to the Function words field.
  • WOLD's function words are given parts of speech.
  • The compound recipes and their counts are derived from the vocabularies, from the words each contributor marks as compounds and the parts they gloss.