Word Counter – Words, Characters & Reading Time

Whitespace mode counts each non-empty token separated by whitespace. Punctuation stays inside a token: one,two and well-known each count as one. Numbers, URLs, hashtags, emoji and punctuation-only tokens can also count as tokens. This mode does not split an unspaced Chinese or Thai passage into individual words.

Where Intl.Segmenter is available, Unicode mode counts the browser’s word-like segments. It can split punctuation-separated text and some unspaced scripts differently from whitespace mode. Segmentation can vary by browser and language data; it is not guaranteed to match a word processor. Whitespace mode remains available.

The character total uses grapheme clusters when supported, so many emoji sequences and combined characters count as one cluster. A separate total removes whitespace before counting. UTF-16 units and Unicode code points are shown separately. Browsers without grapheme segmentation use a clearly labelled code-point fallback. These metrics are different measures, not interchangeable character limits.

Sentences are estimates from browser sentence segmentation, or a punctuation-based fallback that includes trailing text. Segments need letters or numbers to count. Paragraphs are non-empty blocks separated by blank lines; non-empty lines are counted separately. CRLF and CR line endings are normalized for paragraph and line analysis.

Analysis follows the chosen word mode, normalizes tokens to NFC and lowercase, and strips leading and trailing characters other than Unicode letters, marks, numbers or underscores. Internal punctuation remains in whitespace mode. Empty normalized tokens are excluded. Up to 20 unique tokens and 10 repeated tokens are shown; full totals are separate. No accent folding or language-specific case equivalence is promised.

No. Every transformation creates separate output. Duplicate removal compares whitespace tokens and rejoins survivors with single spaces, so layout is flattened. Exact comparison is the default; optional normalized comparison follows the unique-token rules. Replace Input is available only while the output matches the current source revision and fits the input limit.

Time is the current word or token count divided by your chosen words-per-minute rate, rounded up to a whole second. Defaults are 225 wpm for reading and 150 wpm for speaking. These are assumptions you can edit, not measurements of your personal speed.

Counting, transformations and file decoding run in the browser. This tool does not send editor text to a processing service. Draft saving is off by default. Enable device saving to store a draft in this browser; turning it off or deleting the saved draft removes its stored copies. Theme preference is saved separately. Other scripts on the same origin may access browser storage. Export anything important before clearing browser data.

Input supports up to 200,000 UTF-16 units. Files must be UTF-8 .txt or .md, no larger than 1 MB, and their decoded text must fit the input limit. Unsupported or oversized imports are rejected without replacing existing text. Direct oversized typing or pasting remains visible but pauses analysis and draft saving until corrected. No input is silently truncated. Output supports up to 600,000 units because case conversion can expand text.

A restored device draft becomes the initial history state. Pending edits are committed before Undo, allowing Redo to restore them. History keeps up to 30 states within a two-million-unit text budget. Programmatic replacements, imports and clipboard paste also preserve the previous valid input when it remains in history.

Tools use different tokenization, punctuation, Unicode and document-format rules. This page analyzes literal text in its editor, including pasted markup or Markdown syntax. It does not parse Word/PDF documents, hidden text, footnotes or document structure. Use the counting rules shown here when comparing results.

The earlier output stays available for copying or downloading and is marked as belonging to an earlier input revision. Replace Input is disabled until you create fresh output. Reports also label earlier output so it is not presented as a transformation of the current input.