Here's how it works:
- Inside the guts of the script, the Risus: The Anything RPG rules are are used to generate phoneme frequencies for English. The actual text doesn't matter so much as that it was always the same. This model is hidden from the user.
- The user supplies a model corpus (in the Latin alphabet) that should produce very different phoneme frequencies. I particularly like the works of Clark Ashton Smith (in English or in translation).
- The script then maps the most common phonemes in Risus to those of the model text.
- The user enters text to be "translated". The script replaces phonemes using the map generated in the previous step.
Here is an example:
Model Text: Esperanto "translation" of The Witchcraft of Ulua.As long as the model text is the same, the "translations" will be consistent. As you can see above, the "-uary" suffix maps to "-uaro" and the "-ember" suffix maps to "-apan". The faux-conjugation of the verb "to map" also produces an interesting simulated grammar.
Text to Translate Output January February March April May June July August September November December Floluaro Maruaro Roj Ankid Ria Flula Flugo Eecurgh Tambrapan Betapan Damapan one two three four five six seven eight nine ten ola zi knua mien mita tij tatas iej bila kas spring summer autumn winter lil tullan eenuj fikan map mapping mapper mapped roj rognil rognan rognajn
Note that "march" and "map" both translate to "roj" due to a mismatch in the number of phonemes in the different model texts.
Anyway, this isn't a real language but it may be close enough for your roleplaying needs. Simply pick a different corpus for each language and you can always go back and generate more words as you need them.
UPDATE: Today's update adds support for many (but not all) accent marks that appear in English and other languages that use the Latin alphabet.