Word Counter
Words, characters, sentences and reading time, counted live as you type — with SMS and X limits and a case converter. Your text stays in your browser.
Counted in your browser. Nothing you type here is sent anywhere.
- Words
- 51
- Characters
- 292
- Sentences
- 3
- Reading time
- 13 sec
Details
Time to get through it
- Reading13 sec
at 238 words a minute
- Speaking24 sec
at 130 words a minute
Vocabulary variety
82%42 different words out of 51. Short texts score high; long ones repeat more.
- Longest word
- everything
- Paragraphs
- 2
- Lines
- 3
- Words per sentence
- 17
- Average word
- 4.5 chars
- Without spaces
- 240
- Unique words
- 42
- UTF-16 length
- 292
- JavaScript .length
- UTF-8 bytes
- 296
- as a file stores it
Change case
Rewrites the text above. Undo puts it back exactly as it was.
Common limits
- X post
292 / 280 · 12 over
Weighted as X counts it: emoji and CJK characters count double, every link counts 23.
- Meta description
292 / 155 · 137 over
Google cuts snippets by pixel width, roughly 155 characters on desktop. Beyond it, the end is replaced by an ellipsis.
- Page title
292 / 60 · 232 over
Search results show about 60 characters of a title before truncating it.
- SMSUCS-25 parts
5 messages joined together, 43 units left in the last one. These characters are outside the GSM-7 alphabet, so the whole message switches to UCS-2 — 70 per message, 67 per part: — ’
Top keywords
The most repeated words, leaving out common English words like “the” and “and”, numbers and single letters.
- type2 × · 3.9%
- case2 × · 3.9%
- paste1 × · 2%
- text1 × · 2%
- everything1 × · 2%
- around1 × · 2%
- updates1 × · 2%
- go1 × · 2%
- words1 × · 2%
- characters1 × · 2%
Table of contents
How to use the word counter
- 1
Paste or type your text
Everything updates as you type: words, characters, sentences, paragraphs, reading time and the limits panel. There is no Count button to press.
- 2
Read the numbers you need
The four headline figures sit under the box. Details below give characters without spaces, UTF-8 bytes, unique words and the longest word; Common limits shows how the text fares as an X post, a meta description, a page title and a text message.
- 3
Change the case if you need to
Pick Title Case, Sentence case, snake_case or any of the others to rewrite the text, and Undo to put it back. Switch to "Copy the result" to copy a converted version without touching what you wrote.
What counts as a word
Counting words sounds like counting spaces, and that is how many counters do it. It goes wrong in two directions. Text with spaced dashes, bullet symbols or stray punctuation gets extra “words” that are not words at all, and languages written without spaces get almost none: a whole paragraph of Japanese comes out as one.
This counter uses Unicode’s word-boundary rules, the same ones your browser uses when you double-click to select a word, through its built-in Intl.Segmenter. On top of those, the rules are short enough to list in full:
- A word contains at least one letter or digit. A dash, an ampersand or an emoji on its own is not one.
- Apostrophes stay inside a word: don’t, it’s and rock’n’roll are one word each, with a straight or a curly apostrophe.
- Hyphenated compounds are one word: e-mail, co-op, mother-in-law. A dash with spaces round it, or an en or em dash, separates.
- Numbers are one word with their separators: 3.14 and 1,000.
- Chinese, Japanese and Thai are split using the dictionaries your browser ships with, so 日本語の文章です is four words and สวัสดีครับ is two.
- A hashtag or a handle counts its word part: #launch is one word.
Those choices put it close to Microsoft Word on ordinary prose. Where they differ, it is usually because Word counts anything between two spaces — a lone dash included — and Google Docs splits hyphenated words. If a submission system gives you a hard word limit, check with its own counter as well: the rule that matters is the one the person enforcing the limit is using.
Sentences are found the same way, with Unicode’s sentence rules, which already know that the full stop in 3.5 or e.g. does not end one. Titles before a name do trip them up, so Mr., Mrs., Dr., Prof. and a handful of others are taught not to end a sentence. One rule is worth knowing: a full stop followed by a lower-case letter does not end a sentence either, so text typed without capitals counts as fewer sentences than it has. Paragraphs are blocks of text separated by a blank line.
Characters, code units and bytes
A “character” is three different numbers depending on who is counting, and the difference only shows up once there is an emoji or an accent in the text. Take Hi 👋🏽:
- 4 characters as you see them. The waving hand with its skin tone is one symbol on screen, so it counts as one here.
- 7 UTF-16 units, which is what JavaScript’s
.length, many web forms and the UCS-2 text messages use. The hand and the skin tone are two units each. - 11 UTF-8 bytes, which is what a file on disk takes up and what a MySQL
VARCHARor a URL actually stores.
Text messages are the case that costs money. An SMS carries 160 characters only while every character is in the GSM-7 alphabet — plain letters, digits, common punctuation and a few accented letters like é and ü. A single character outside it, such as a curly apostrophe pasted from a word processor or any emoji, switches the whole message to UCS-2, and the limit falls to 70. Longer messages are split into parts of 153 or 67, because each part gives up some room to the header that stitches them back together, and € [ ] { } ^ ~ | and \ cost two places each. The limits panel shows the encoding, the parts and exactly which characters forced UCS-2, so you can swap them out.
X counts differently again: most Latin, Greek and Cyrillic characters count one, Chinese, Japanese and Korean count two, every emoji counts two, and a link always counts 23 whatever its length. The X figure here follows those rules for links starting with http or www; X also shortens bare domains, which this does not detect.
The case converter
Prose cases change the letters and leave every space, line break and mark of punctuation where it was. Title Case capitalises each word but keeps a, an, the, and, but, or, of, in, on, at, to, for and by lower case in the middle of a title — never at the start or after a colon — so “the lord of the rings” becomes “The Lord of the Rings”. Sentence case capitalises the first word of each sentence and line and restores “I”. Both leave acronyms and brand names like NASA and iPhone alone, unless the whole text was in capitals, in which case there is no way to tell an acronym from shouting.
Code cases rebuild each line as an identifier: first name becomes firstName, FirstName, first_name, first-name or FIRST_NAME. They also read existing identifiers, so parseXMLHttpRequest converts cleanly to parse_xml_http_request. Paste a column of field names and each line converts on its own.
Counting from a script
The same counting code answers at /api/v1/text-stats. Send the text as a JSON body — never in the URL, where it would end up in logs — and add "case": "title" or any other case to get it converted too. Text sent to the API does travel to our server to be counted; it is not stored or logged. Text typed into this page never does.
# Count a text, and convert it to Title Case while you are at it
curl -X POST https://zeal.tools/api/v1/text-stats \
-H 'content-type: application/json' \
-d '{"text": "The lord of the rings. A long read.", "case": "title"}'
# → { "words": 8, "characters": 35, "sentences": 2, "readingSeconds": 2,
# "sms": { "encoding": "GSM-7", "segments": 1, ... },
# "converted": "The Lord of the Rings. A Long Read.", ... } MCP clients can call the same core: add https://zeal.tools/mcp to your client and use the count_words tool.
Frequently asked questions
Because there is no single definition of a word, and each tool picks one. Microsoft Word counts anything between spaces, so a dash with a space either side adds a word, and a URL is one word. Google Docs splits at more punctuation, so "e-mail" can become two. This counter follows Unicode's word-boundary rules: a word must contain a letter or a digit, contractions and hyphenated compounds are one word, and punctuation or an emoji on its own is not a word at all. On ordinary prose the three usually agree to within a percent or two; the gap shows up in text full of dashes, symbols, links or code.
Yes, which most word counters do not. Those languages are written without spaces between words, so a space-counting tool reports a whole sentence as one word. This one uses the dictionaries built into your browser to find word boundaries, so 日本語の文章です counts as four words. Some style guides count Chinese and Japanese by character instead — the character count is right beside it for that.
No. The page counts it with your browser's own text-segmentation engine, on your machine. If you leave "Remember on this device" ticked, the text is kept in this browser's local storage so it is still here next time; untick it and the stored copy is deleted. It is never put in the address bar, so sharing the page link never shares your text. The separate API, for scripts, does receive the text you send it — that is what an API is — and does not store it.
Words divided by 238 words a minute, the average silent-reading speed for non-fiction found by a 2019 review of 190 studies. Fiction reads a little faster, technical writing slower, and your own speed may be well off the average. Speaking time uses 130 words a minute, a comfortable pace for a talk or a sermon; conversation runs nearer 150 to 160. Both figures assume English; Chinese and Japanese word counts are not comparable with English ones, so the times are rough there.
"Characters" counts what you see — so 👍🏽 is one character, even though it is built from two code points. "Without spaces" removes spaces, tabs and line breaks, which some essay and form limits ask for. The Details panel adds UTF-16 length, which is what JavaScript and many web forms count (👍🏽 is four), and UTF-8 bytes, which is what a file or most databases actually store (👍🏽 is eight).
How often a word appears as a share of all the words — keyword density. Common English words like "the", "and" and "is" are left out, along with numbers and single letters, so the list shows what the text is about. Search engines stopped rewarding a target density long ago; use the list to spot a word you have leaned on too often.
Yes: POST your text to /api/v1/text-stats as JSON, with an optional "case" to convert it, and you get every number on this page back. It is also an MCP tool, count_words, for AI agents. No key is needed; there is a light rate limit of 120 requests a minute.
The Zeal Promise
Non-negotiable, on every tool we ship.
- No sign-up — ever
- No watermarks on any output
- No ads near the download button
- Works fully on mobile
- Privacy-first: processing stays in your browser
- Open source (MIT) — audit it, contribute, self-host
- Zero runtime dependencies in tool logic