DanfioSearch tools/

Type a tool name, a pinyin spelling or a synonym - both languages are searched.

Word & Character Counter

Global tools

Paste or type a piece of text and this page counts it while you type: words, characters, sentences and paragraphs, a reading time, a speaking time, the words you use most, and a report you can copy or download. It works on an article, an email, a chapter, a script or a caption, and the numbers move as you edit - the same function that produced the counts already in the page you loaded runs again in your browser on every keystroke, with no request of any kind.

Word & Character Counter - Danfio

https://danfio.com/tools/word-counter/ - printed from this page

Everything happens in this tab as you type. There is no upload button on this page because there is no upload: the text never leaves the browser, and the share link only carries your settings, never the text itself.

The single setting that moves a word count the most. A word processor splits `state-of-the-art` into four; a reader would call it one thing.

Whether `2026`, `40%` and `3.14` are words. A translator counts them and a word processor counts them; a word count for a voice-over usually does not.

200 is the usual figure for silent reading of ordinary prose. A technical paper is slower, a list of names faster, and the reading time below is only as good as this number.

130 is an unhurried presentation pace, which is why the same script takes about half as long again to say aloud as to read.

How many words to list, commonest first. The table cannot be longer than the number of distinct words in the text.

Leave it blank for "count-report.txt" and "count-word-list.txt".

Results update as you type. Nothing leaves your device.

Words

72

370 characters, 7 sentences - about 22 sec to read and 33 sec to say aloud.

Words
72
Characters
370 code points
Sentences
7
Paragraphs
2
Read time
22 sec 1 min read badge
Speak time
33 sec

Words and sentences

Unique words
54 case-insensitive
Sentences
7
Average sentence
10.3 words
Average word
3.9 characters
Longest word
character 9 characters
Lines
3 1 blank
Tokens with a hyphen
1 the switch above decides these
Tokens that are numbers
3 the switch below the first one decides these

Characters

Code points
370 what most counters call characters
Grapheme clusters
370 what a person points at: an emoji is one
Without spaces
300
UTF-16 code units
371 what text.length says - an emoji costs 2
UTF-8 bytes
373 what the text weighs on disk
Letters
271
Digits
11
Punctuation
17
Emoji
1 characters, not words

The same text, counted the other way

The ruleWordsRead at your speed
Compounds split, numbers counted - this is the one above7222 sec
Compounds split, numbers skipped6921 sec
Compounds kept whole, numbers counted6921 sec
Compounds kept whole, numbers skipped6620 sec
A word processor's count is usually closest to the first row, which is why that row is the default: it splits compounds and counts numbers.
None of the four is wrong. Each answers a different question, and the two switches on the form decide which question this page asks.

The differences are small on this text and large on others: a page of hyphenated medical terms can move by a tenth, which is more than enough to explain a discrepancy with another counter.

The whole count, as a report

Word counter report
-------------------

Text: The word counter is a small tool with a long argument behind it. Is …

Rules in force
  Hyphenated compounds.......... split into their parts (well-known = 2)
  Standalone numbers............ counted as words
  Reading speed................. 200 words per minute
  Speaking speed................ 130 words per minute

Counts
  Words......................... 72
  Unique words.................. 54
  Sentences..................... 7
  Paragraphs.................... 2
  Lines......................... 3
  Blank lines................... 1
  Characters (code points)...... 370
  Characters (graphemes)........ 370
  Characters without spaces..... 300
  UTF-16 code units............. 371
  UTF-8 bytes................... 373
  Letters....................... 271
  Digits........................ 11
  Punctuation................... 17
  Emoji......................... 1
  Chinese/Japanese characters... 0
  Average word length........... 3.9 characters
  Average sentence length....... 10.3 words
  Longest word.................. character (9 characters)
  Reading time.................. 22 sec (a badge would say 1 min read)
  Speaking time................. 33 sec
  Tokens holding a hyphen....... 1
  Tokens that are numbers....... 3

Most used words
    1. is                       5  6.9%
    2. the                      4  5.6%
    3. a                        4  5.6%
    4. word                     3  4.2%
    5. of                       3  4.2%
    6. one                      3  4.2%
    7. or                       2  2.8%
    8. he                       2  2.8%
    9. counter                  1  1.4%
   10. small                    1  1.4%

Nothing was uploaded: the text never left the page that counted it.

Copy it, or download it. The bytes in the box are the bytes in the file, and the first download button hands you exactly this.

The words used most

WordTimesShare of all words
is56.9%
the45.6%
a45.6%
word34.2%
of34.2%
one34.2%
or22.8%
he22.8%
counter11.4%
small11.4%
The 10 commonest of 54 distinct words.
Case is folded, so `The` and `the` are one entry, and the punctuation around a word is not part of it.

Useful for finding the word you have used eleven times without noticing, and for checking that the phrase a page is meant to rank for is at least present.

What the numbers do not say

  • Hyphenated compounds are split here, so `state-of-the-art` counts as four words. Change the switch above and the count drops - which is the whole reason two counters can look at one page and disagree. This text holds 1 token with a hyphen in it.
  • Standalone numbers count as words, so `2026` and `40%` are words. This text holds 3 tokens made of digits and punctuation alone.
  • A sentence ends at a full stop, a question mark, an exclamation mark, an ellipsis or the CJK stop `。`, with four exceptions: a stop between two digits is a decimal point, a known abbreviation (`Mr.`, `Dr.`, `Fig.`) does not end one, a lone capital in front of another (`J. R. R. Tolkien`) is an initial, and a stop followed by a lower-case word continues the sentence (`index.html for`). The rule that catches the last of those also means `the U.S. It was` reads as one sentence, which is the price of not needing a dictionary.
  • Chinese and Japanese are written without spaces between words, so a whitespace count calls a whole paragraph one word. Each Han, Hiragana or Katakana character is counted as a word here instead, which is what a word processor does. Korean is deliberately not treated that way: Hangul is written with spaces, and each syllable is not a word.
  • A paragraph is separated by a blank line, not by a line break, because that is how Markdown and Word read it. A poem or a code listing pasted in therefore counts as one paragraph, and a line count is printed beside it for the cases where that is the number you wanted.
  • A word is a run of text holding at least one letter or digit, so `don’t`, `3.14` and `some/path/notes.txt` are one word each and a dash on its own is no word at all. Punctuation is not part of a word in the list or the frequency table, while interior punctuation stays, which is why `stop,` and `stop.` are one entry instead of two. Emoji are counted as characters and not as words.
  • Reading and speaking times are arithmetic at the two speeds on the form (200 and 130 words per minute), so they are estimates at your settings rather than measurements. Pasted from a style guide: a blog badge rounds the same arithmetic up to whole minutes, and the report prints both numbers so a badge saying "3 min read" is not a contradiction.
  • Four character counts are printed because "a character" has four defensible answers: code points, grapheme clusters, UTF-16 code units and UTF-8 bytes. A form that says "maximum 100 characters" almost always means UTF-16 code units, where one emoji costs two, and a meta description limit is usually counted in code points.
Counted in this tab by the same function that produced the numbers already in the page you loaded, so nothing was sent anywhere, and there is no endpoint behind this page that could accept a document even if something tried to send one. The text never leaves this tab either: a link or a copy of the result carries your settings, never what you pasted. Paste a draft under embargo if you like.

Two counters can look at one page and disagree, and the reason is always the same: each one is answering a different question. Is `state-of-the-art` one word or four? Is `2026` a word? Is this emoji 👍 one character or several? Does `Mr.` end a sentence? Both of the decisions that move a real number are switches on the form, the table under the result shows all four counts at once, and every other convention this page follows is written out beside the numbers rather than left for you to discover by disagreeing with it.

About this word & character counter

Counting a page sounds like one number and is really four, because two counters can look at the same text and disagree without either being wrong. This page counts words, characters, sentences and paragraphs while you type, adds a reading and speaking time, and prints the words you use most in a table you can copy or download.

The method is on the form: is a hyphenated compound one word or several, and is a bare number a word at all? Those two switches move a real count, and the page also prints all four defensible character totals - code points, grapheme clusters, UTF-16 units and UTF-8 bytes - because a form that says a hundred characters rarely means the same hundred the next box counts.

There is no account, nothing is uploaded, and the count runs on your own machine on every keystroke. Treat the numbers as conventions this page states rather than universal truths: where a rule is arguable it is written beside the figure, so a disagreement with another counter becomes something you can check instead of something you have to trust.

How this word & character counter works

What counts as a word

A word is a run of text holding at least one letter or a digit, taken apart at the spaces. That single rule is why `don’t`, `3.14` and `some/path/notes.txt` are one word each - nobody would call them two - and why a dash on its own is not a word at all. A hyphenated compound is either one word or several depending on the switch you choose, and that setting alone can move a count by a few per cent on ordinary prose and by a tenth on a page full of hyphenated technical terms.

Chinese and Japanese are written without spaces between words, so the same rule that works for English would call a whole paragraph one word. Each Han, Hiragana and Katakana character is counted as a word instead, which is the convention a word processor follows. Korean is deliberately excluded: Hangul is written with spaces, and its syllables are not words. Emoji are counted as characters, never as words, and the frequency table reads a word without the punctuation around it, so `stop,` and `stop.` are one entry rather than two.

What counts as a sentence, and as a paragraph

A sentence ends at a full stop, a question mark, an exclamation mark, an ellipsis or the CJK stop `。`. Four things are deliberately not sentence ends: a stop between two digits, which is a decimal point; a known abbreviation followed by a capital, as in `Mr. Smith` or `Fig. 3`; a lone capital in front of another, as in `J. R. R. Tolkien`; and a stop followed by a lower-case word, which is how `see index.html for` stays in one piece. The rule that catches the last of those also means `the U.S. It was` reads as a single sentence, which is the price of a rule that needs no dictionary and can be printed here in full.

A paragraph is separated by a blank line, not by a line break, because that is how Markdown and Microsoft Word read it. A poem or a code listing pasted into the box therefore counts as one paragraph - so a line count and a blank-line count are printed beside it, for the many cases where the line is the number you actually wanted.

Characters, and why four numbers are printed

A character has four defensible answers and the page prints all four. Code points are what most counters call characters. Grapheme clusters are what a person points at, where a skin-tone emoji or a flag is one thing rather than several - measured with the browser’s own `Intl.Segmenter`, and on a browser that has no segmenter the code-point count is used instead and labeled as such. UTF-16 code units are what `text.length` says in JavaScript, where one emoji costs two, and UTF-8 bytes are what the text weighs in an email, a database column or on disk.

The fourth number matters more often than people expect. A form that says "maximum 100 characters" almost always means UTF-16 code units; a meta description limit is usually counted in code points; a database column limit is bytes. Pasting one sentence into three boxes and getting three totals is not a bug in any of them, and the point of printing all four here is that the difference should be visible rather than mysterious.

Reading time is arithmetic, not a measurement

Reading and speaking times are the word count divided by the two speeds on the form: 200 words per minute for silent reading of ordinary prose and 130 for an unhurried presentation, both of which are defaults you can change rather than facts about you. A technical paper is slower, a list of names is faster, and a figure computed at one speed can be off by forty per cent at another - which is why the speeds are on the form, quoted in the report, and never presented as a stopwatch result.

The report also prints the same time rounded up to whole minutes, because that is the number a blog’s "3 min read" badge shows, and a visitor comparing the two should find them agreeing rather than fighting. Both numbers are printed, so nothing is hidden by the rounding.

Worked examples

Each example below was run through the calculator on this page when the site was built, so the numbers match what you see when you enter the same inputs.

A paragraph of ordinary prose

The text to count: The quick brown fox jumps over the lazy dog. It was, on the whole, a well-known sentence. · A hyphenated compound: Counts as its parts (state-of-the-art = 4) · A standalone number: Counts as a word (2026 is one) · Reading speed: 200 words per minute · Speaking speed: 130 words per minute

Words

18

89 characters, 2 sentences - about 5 sec to read and 8 sec to say aloud.

Words: 18 · Characters: 89 · Sentences: 2 · Paragraphs: 1

Nothing unusual here except the compound: it is split, so `well-known` contributes two words and the count is the one a word processor would give.

The same paragraph, counted the other way

The text to count: The quick brown fox jumps over the lazy dog. It was, on the whole, a well-known sentence. · A hyphenated compound: Counts as one word (state-of-the-art = 1) · A standalone number: Not counted (2026 is nothing) · Reading speed: 200 words per minute · Speaking speed: 130 words per minute

Words

17

89 characters, 2 sentences - about 5 sec to read and 8 sec to say aloud.

Words: 17 · Characters: 89 · Sentences: 2 · Paragraphs: 1

One word fewer, from a single hyphen. Two counters can differ by exactly this much and neither of them is broken.

A line written in Chinese

The text to count: 今天天气很好。我们去公园散步。 · A hyphenated compound: Counts as its parts (state-of-the-art = 4) · A standalone number: Counts as a word (2026 is one) · Reading speed: 200 words per minute · Speaking speed: 130 words per minute

Words

13

15 characters, 2 sentences - about 4 sec to read and 6 sec to say aloud.

Words: 13 · Characters: 15 · Sentences: 2 · Paragraphs: 1

There is no space anywhere in this text, so a whitespace count would call it one word. Each character is counted instead, and the two `。` stops make two sentences.

Numbers, with the number switch off

The text to count: The report says 40% growth in 2026 and 3.14 months of runway. · A hyphenated compound: Counts as its parts (state-of-the-art = 4) · A standalone number: Not counted (2026 is nothing) · Reading speed: 200 words per minute · Speaking speed: 130 words per minute

Words

9

61 characters, 1 sentence - about 3 sec to read and 4 sec to say aloud.

Words: 9 · Characters: 61 · Sentences: 1 · Paragraphs: 1

The three tokens made of digits and punctuation alone drop out of the count - the setting a voice-over script wants, where `40%` is read aloud as "forty per cent".

Frequently asked questions

Why does this give a different number from Microsoft Word or Google Docs?

Usually because of a hyphenated compound or a number, so try the other setting of the two switches first. Beyond those, every counter has to decide what a word is: some split `state-of-the-art` and some do not, some count `2026` and some do not, and some strip the punctuation attached to a word while others keep it. This page moves both of the decisions that change a real number onto the form, prints the other counts beside the result, and lists every remaining convention in the notes - so a difference you cannot explain should at least be a difference you can locate.

Is this a character counter for social posts and meta descriptions?

Yes, and it prints four of them deliberately. A platform that says 280 characters usually means UTF-16 code units, where an emoji costs two; a meta description limit is usually counted in code points. The counts without spaces and in UTF-8 bytes are printed too, because a database column and an email are measured in bytes and a form rarely says which of the four it meant.

How is the reading time calculated?

Word count divided by the speed on the form - 200 words per minute by default for reading, 130 for speaking - with the exact time printed beside the same figure rounded up to whole minutes, which is what most "min read" badges show. It is arithmetic at your settings rather than a measurement of anybody’s reading speed: change the speed and both numbers change accordingly.

Does it count emoji as words?

No. An emoji is counted as a character - one grapheme cluster, however many code points it is made of - and the word count leaves it alone. The rule behind that is the same one that stops a dash or an ellipsis from being counted: a token holding no letter and no digit is not a word. The emoji count is printed separately, so a text full of them still reports what it contains.

Can I use it as a keyword density tool?

The frequency table is the density part of that: it lists the most used words with the number of times each appears and its share of all the words, with case folded so `The` and `the` are one entry and the punctuation around a word left out. It stops at single words on purpose. A phrase-level density figure needs decisions about what counts as a phrase and about stemming (`count`, `counts` and `counting`), and folding those together without doing the linguistic work would produce a number that looks authoritative and is not - so the table shows what it can defend.

Is my text sent anywhere?

No. There is no upload endpoint behind this page, no request made while you type, and nothing stored: the counting function runs in the tab, which is also why the tool keeps working with the network switched off. Paste a confidential draft, a client brief or an unreleased announcement if you like - the only copy of it is the one you pasted.

What happens with a very long text?

It still works. Counting is linear in the length of the text, so a book-length paste is counted in a fraction of a second in the browser and the word list download simply grows with it. The two parts that get slower are the frequency table, which sorts every distinct word, and the report, which prints it all - so on something enormous the counts stay right and the tables just get long.

Assumptions and sources

  • Words are split on whitespace with hyphens and numbers handled by the page's own switches; Chinese and Japanese are counted per character, and Korean is excluded because Hangul uses spaces.
  • Reading and speaking speeds of 200 and 130 words a minute are conventions you can change, and the page prints the exact time and the rounded-up version.
  • Nothing is uploaded; the text is counted in your browser and stays out of the share link.

Last reviewed 2026-09-17. Runs locally on your device; keep backups; no financial advice applies. Read the full disclaimer.

Where a figure on this page comes from an official agency — HUD, IRS, FICO or a state body — the site keeps a dated snapshot of the source page in its sources library, and a scheduled job re-checks that snapshot, so the number can be traced back to the document it came from.

Related tools

Open Graph Preview & SitemapDeveloper ToolsRead a page the way a crawler does: the search result, the share card and a 20-point metadata checklist, plus sitemap.xml and robots.txt builders. No upload.Preview social links →Regex Tester & ReplacerDeveloper ToolsTest a JavaScript regular expression against your own text: every match with its groups and positions, what each flag does, and how it runs in Python or PCRE.Test a regex →Minifier & BeautifierDeveloper ToolsMinify or beautify CSS and HTML in your browser: comments, whitespace and two safe shortenings, with an honest list of what it will not touch. No upload.Minify or beautify code →Base64 Encoder & DecoderDeveloper ToolsEncode text to base64 or decode it back, in your browser, with the standard and URL-safe alphabets, UTF-8 handled properly, percent-encoding and data URLs too.Encode or decode Base64 →Acronym GeneratorText & WritingTurn a phrase into an acronym of first letters, with a letter-by-letter table of where each initial came from. Deterministic, private, honest about trademarks.Make an acronym →Business Name GeneratorText & WritingTurn a few keywords into business name ideas, usernames and hashtags. Six naming shapes, seeded so the list reproduces, and no false claims about availability.Generate business names →