Character & Word Counter

Q: Why does an emoji count as one character here but two elsewhere?

Because most counters report UTF-16 code units rather than characters, and an emoji occupies two of them. This tool counts graphemes — what a person actually perceives as one character — so a lock emoji counts as one, and a family emoji made of several joined people also counts as one despite being eight code units long. Both numbers are shown, because which one you need depends on the system you are feeding: a database column limit usually cares about bytes, while a human-facing character limit should care about graphemes.

Q: What is the difference between characters and bytes?

A character is a symbol; a byte is a unit of storage. In UTF-8, plain English letters take one byte each so the two numbers match, but an accented letter takes two, most CJK characters take three, and emoji take four. That gap is why a field advertised as holding 255 characters may reject your text at 200 — the limit is really 255 bytes. If you are sizing a database column or fitting into a protocol field, the byte count is the number that matters.

Q: How accurate is the token estimate?

It is a rough guide, not a precise figure. Language models split text into tokens using a learned vocabulary, and the ratio varies with the content: common English words are often one token each, while unusual words, code and non-Latin scripts break into many more. The estimate here uses roughly four characters per token, which is a reasonable average for ordinary English prose. Treat it as an order-of-magnitude check for whether text will fit in a context window, and use the model provider's own tokeniser when the exact count matters.

Q: How is reading time calculated?

From the word count, at 200 words per minute for silent reading and 130 for reading aloud. Those are conventional averages for adult readers of ordinary prose. Real speed varies a great deal with the material — dense technical writing is much slower, familiar light content much faster — so the figure is best treated as a rough label for an article rather than a precise measurement. It is rounded up to a minimum of one minute, since no piece of text takes zero time to read.

Q: Does it count words the same way as a word processor?

Close, but not always identical. This tool splits on whitespace, so a hyphenated compound counts as one word and a number counts as a word. Word processors apply their own rules and sometimes differ on hyphens, contractions and numerals, so a small discrepancy against a specific application is expected rather than an error. For the purposes this tool serves — checking against a limit, estimating length — the difference is immaterial.

Q: Is my text uploaded anywhere?

No. Every count is computed in your browser as you type, and there is no request to any server at any point. That matters for this tool in particular, because the text people want to measure is often a draft, a private document or something under embargo. You can paste an unpublished manuscript here and it stays entirely on your machine.

Count characters, words, sentences, paragraphs and UTF-8 bytes as you type, with reading time and a token estimate for AI context limits.

Your text

Characters

Words

Sentences

Bytes (UTF-8)

Detailed statistics

Counting happens entirely in your browser as you type. Your text is never uploaded, logged or stored.

What It Does

This counts what is actually in your text, live as you type. Characters, words, sentences, paragraphs and lines are the obvious part, and most counters stop there. The numbers that tend to be missing elsewhere are the ones that cause real problems: the UTF-8 byte count, which is what a database column or a protocol field is genuinely limited by, and a proper grapheme count, so an emoji registers as the one character a person sees rather than the two units a computer stores. Alongside those it estimates reading and speaking time, gives a rough token count for fitting text into an AI context window, and shows which words you have leaned on most. Everything updates as you type and nothing is sent anywhere.

When to Use It

You are writing a meta description or a social post with a hard character limit and need to know exactly where you stand as you edit.
A form or database field keeps rejecting text that looks short enough, and you suspect the limit is counted in bytes rather than characters.
You are checking whether a document will fit inside a language model's context window before pasting it in.
An article needs a reading-time label and you want the figure without installing a plugin.
You are editing prose and want to see which words you have repeated too often.

Worked Examples

The quick brown fox jumps over the lazy dog.

Plain ASCII, so the character and byte counts are identical — nine words, one sentence, 44 of each. This is the baseline case where every counter agrees.

Café naïve résumé 🔒

The interesting case. Four visible items but noticeably more bytes, because each accented letter takes two bytes in UTF-8 and the emoji takes four. This gap is exactly what breaks a field limit that was specified in characters but enforced in bytes.

First paragraph here.

Second paragraph follows. It has two sentences.

Shows the structural counts working: two paragraphs, three sentences, and a line count that differs from the paragraph count because of the blank line between them.

Features

Live counts updated as you type

Grapheme-accurate character counting for emoji and accents

UTF-8 byte count for database and protocol limits

Word

sentence

paragraph and line counts

Characters with and without spaces

Reading and speaking time estimates

Rough AI token estimate for context windows

Most frequent words in your text

Longest word and average word length

Entirely browser-based with no uploads

How to Use

1. Type or paste your text into the box — the counts update immediately. 2. The headline figures show characters, words, sentences and bytes. 3. Expand the detailed statistics for reading time, averages and the most frequent words. 4. Watch the byte count if you are working against a database or protocol limit. 5. Click Clear to empty the box and start again.

Common Mistakes

Assuming a character limit means characters. Database columns and many APIs count bytes, so text with accents, emoji or non-Latin script hits the limit sooner than its character count suggests.
Trusting a counter that reports code units. Most report string length, which counts an emoji as two and can count a single flag or family emoji as eight — a real problem for anything with a hard limit.
Treating an AI token estimate as exact. Tokenisation depends on the model's vocabulary, and code or non-English text produces far more tokens per character than ordinary prose.
Forgetting that trailing whitespace counts. A stray newline at the end of a paste is a real character, and it is enough to push text one over a limit for no visible reason.
Comparing word counts across tools and expecting them to match. Different applications treat hyphens, contractions and numerals differently, so a small discrepancy is normal.

Frequently Asked Questions

Why does an emoji count as one character here but two elsewhere?

What is the difference between characters and bytes?

How accurate is the token estimate?

How is reading time calculated?

Does it count words the same way as a word processor?

Is my text uploaded anywhere?

Related Tools

Case Converter