Pronunciation first · spelling estimates shown

Syllable Counter

Count syllables in a word, sentence, lyric, or poem — then see which answers come from pronunciation evidence and which are only spelling-based estimates.

Paste one word or a whole text. The counter reports the total, each line’s count, and a source-labelled result for every word. Retained CMUdict pronunciations supply exact source-derived counts where available; unfamiliar words use a conservative spelling estimate that is never presented as dictionary evidence.

Enter your text

Keep poem line breaks: each physical line gets its own count. Up to 20,000 characters, 250 lines, and 5,000 words.

Privacy: this feature analyzes the POST body in memory. It does not put your writing in a URL, save a draft in browser storage, or retain the submitted text.

Method, not magic

How this syllable counter reaches an answer

A syllable is organized around a spoken vowel sound. That makes pronunciation a better starting point than the number of written vowels: strengths has one spoken vowel nucleus, while adjacent written vowels can represent either one sound or more than one. The counter therefore follows a two-tier method and exposes the tier beside every result.

  1. 01

    Preserve lines and find words

    Runs of English letters are counted as words; straight and curly apostrophes may appear inside them. A hyphenated compound is counted as its component words. Numbers and symbols are not converted into guessed pronunciations. Original line breaks remain intact for verse.

  2. 02

    Count pronunciation nuclei when the source covers a word

    The compiled data release retains CMUdict pronunciations in ARPABET. Each vowel phoneme carries a stress digit: 0 for unstressed, 1 for primary stress, or 2 for secondary stress. Counting those marked vowel phonemes produces the source-derived syllable count. The displayed stress string lets you audit that derivation.

  3. 03

    Use a small fallback, and call it an estimate

    If there is no pronunciation record, the fallback counts adjacent groups of a, e, i, o, u, y, treats an initial y as a consonant, and removes a likely silent final e unless the word has a consonant + le ending. It never returns fewer than one. The evidence cell shows the groups and adjustment used.

Using the count for poems, lyrics, and prose

Haiku and syllabic verse

Keep each intended verse line on its own physical line and compare the line column with your target pattern, such as 5–7–5. Then inspect every tilde. A line can hit the numerical target while still depend on an uncertain name or a pronunciation that differs in your accent.

Song lyrics and spoken performance

Singers routinely compress or expand words, and speech can elide an unstressed vowel. Treat the report as a neutral first pass. Read the line aloud at performance tempo; the performed count is the one that governs the phrase.

Headlines, speeches, and classroom work

The total and average help compare drafts, while the per-word table makes a useful audit trail. They do not by themselves measure reading difficulty: sentence structure, familiarity, meaning, and audience matter too.

A quick manual check when a count matters

  1. Filter the report to spelling estimates and say each highlighted word naturally.
  2. Check whether the source pronunciation matches your accent and intended meaning.
  3. For verse or lyrics, read the complete line aloud; context may trigger elision.
  4. Record your deliberate reading if collaborators need to reproduce the count.

Limits and interpretation

One primary pronunciation

A word may have alternate pronunciations, and a spelling may represent different words. This report uses the primary retained variant without sentence-level meaning detection.

Accent and dialect

CMUdict primarily documents North American English. Vowel mergers, non-rhotic speech, regional forms, and individual delivery can produce a different count.

No printed syllable breaks

ARPABET provides a phoneme sequence and stress, not trustworthy boundaries in the original spelling. The tool does not invent hyphenation points.

English-letter tokenization

Accented letters, non-Latin scripts, numerals, emoji, and abbreviations are not expanded into speech. Hyphens split a compound into its written parts.

Pronunciation results reflect data version v1/2026.07-lexical.6. Source coverage and primary variants can change between releases. See data sources and licensing for the build's provenance.

Frequently Asked Questions

How does the syllable counter work?
It counts stress-marked vowel sounds in the primary retained CMUdict pronunciation when one is available. Unknown words use a small spelling heuristic and are visibly marked as estimates.
Can I count the syllables in a sentence or poem?
Yes. Paste one word, a sentence, lyrics, or a poem. The report gives a total, a count for every physical line, and evidence for each distinct word.
Why can two people count a word differently?
Accent, emphasis, elision, and alternate pronunciations can change the spoken count. The tool reports one primary North American English pronunciation where the source covers the word; it does not declare every other reading wrong.
Does the counter split words into syllables?
No. A pronunciation supplies vowel sounds and stress, but not a reliable mapping from each sound back to printed-letter breakpoints. Dictionary hyphenation and spoken syllable boundaries are related but are not the same job.
Is my poem or text saved?
No. Text is sent in a POST body, analyzed in memory, returned to your browser, and not stored by this feature. It is never placed in the page URL or browser local storage.