Language detect
Guess a text's language from its trigram statistics, ranked. Short or mixed-language text stays ambiguous, so this shows candidates, not one verdict. Runs on your device.
Language detect scores a pasted text against trigram statistics for many languages, ranking every candidate rather than picking one verdict. Below a minimum length, 10 characters by default, it refuses to guess and reports und instead. A text that mixes languages, or runs only a few words long, stays genuinely ambiguous, so the top-ranked candidate is not always right.
Questions
How does it guess the language?
By trigram statistics over ISO 639-3 codes, using the franc-min library; it is a statistical comparison against known language patterns, not a neural or machine-learning model, which is why short or mixed-language text produces genuinely ambiguous results.
What does "und" mean in the result?
Undetermined: either the text is shorter than the minLength option, 10 characters by default, or no candidate language scored a confident enough match against the trigram statistics.
How many candidates does it show?
Up to ten, ranked by score, rather than a single best guess, since the underlying statistics rarely produce one clearly dominant answer for anything but longer, single-language text.
What if the detected language code has no name in the list?
The bare ISO 639-3 code is shown instead of a guessed name; the tool has a table of common language names and does not invent one for a code outside it.
Can I lower the minimum length to detect short text?
Yes, through minLength, though it must be a positive whole number; setting it lower does not make short-text detection more accurate, only more willing to attempt it.