Haiku mode: the 5-7-5 check
Turn on Haiku mode and the first three non-empty lines are checked against the classic 5-7-5 pattern, with a per-line ✓ or the actual count when a line misses. Two honest notes for haiku writers: the 5-7-5 rule counts English syllables, which is already an approximation of the Japanese on the form actually counts (Japanese sound units run shorter than English syllables — many translators treat around 10–14 English syllables total as closer in spirit); and contemporary English-language haiku frequently abandons strict 5-7-5 entirely. If your goal is the school-assignment or puzzle version of the form, the checker enforces it exactly; if your goal is haiku as an art, treat 17 as a ceiling, not a target.
How the counting works — and where it can’t be perfect
English spelling doesn’t encode syllables, so every automatic counter — this one included — is a heuristic. The core method: find vowel groups (a run of a, e, i, o, u, y counts as one nucleus), then adjust for the patterns spelling uses systematically: a final silent e drops (make = 1), a consonant + le ending adds one back (ta·ble = 2), -es counts only after hissing sounds (boxes = 2, makes = 1), -ed counts only after t or d (wanted = 2, walked = 1), and vowel-pair hiatus like i·o in violin splits into two nuclei. That resolves the vast majority of words. What it can’t resolve is where spelling underdetermines speech: fire and hour are one syllable or two depending on dialect, every is two or three depending on speed, and names follow no rules at all. That’s why uncertain patterns are flagged rather than silently guessed, and why every word is correctable.
When a word shows the wrong count — fix it
Click the word. A small − / + control appears; adjust to the count you actually pronounce, and the correction applies to every occurrence of that word in your text and in the line and total counts. The segmentation display keeps showing the heuristic split (segmentation and count are separate questions), but the numbers everywhere use your override. Overrides live only in your tab for this session — nothing is stored or sent.
Spanish syllable counting
Spanish is a genuinely different — and easier — problem, because Spanish spelling encodes syllables almost perfectly. Vowels divide into strong (a, e, o) and weak (i, u): two strong vowels together are separate syllables (po·e·ta), a weak + strong pair forms one diphthong syllable (tie·rra), and a written accent on a weak vowel breaks the diphthong (dí·a = 2). Switch to Spanish mode and these rules are applied directly — counts for Spanish text are near-exact rather than heuristic, which is why the low-confidence flagging mostly disappears in that mode. (Poets scanning Spanish verse meter should note verse counting adds conventions — synalepha across word boundaries, final-stress adjustments — that word-level counting deliberately doesn’t apply.)
Frequently asked questions
Why does the counter disagree with my dictionary?
Dictionaries record one reference pronunciation; the heuristic reconstructs syllables from spelling. Words where spelling and speech diverge — chocolate (2 or 3), different (2 or 3), camera (2 or 3) — genuinely have multiple correct answers by speaker. When it matters (a syllable-counted form), click the word and set the count you’d actually say.
What do the dots in the words mean?
They mark the approximate syllable boundaries the heuristic found — shown so you can see why a word got its count, not just the number. Boundary placement follows the usual pattern of a single consonant starting the next syllable (ta·ble, ca·me·ra) and consonant pairs splitting between syllables (win·ter).
Does it handle hyphenated words and numbers?
Hyphenated words count as their parts (twenty-one = 3). Digits aren’t expanded to words — “100” counts as zero syllables — so write numbers out (one hundred) in verse you’re counting.
Is my text uploaded anywhere?
No — counting runs in your browser tab with no server or network request, and the page works offline once loaded.