Your text

Dictionary

Loading the dictionary: `nspell` (MIT, ~41 KB) plus the `dictionary-en` en_US Hunspell data (MIT AND BSD, ~562 KB), imported on this page only.

Result

Start typing above — the results update as you write.

What this checker does and cannot do

Names, jargon, brand names and code-like words are reported as unknown, because they are not in a general English dictionary. Racira, Kubernetes, Nikhil and localhost are all "misspellings" to this tool — add them under Words to accept rather than changing them.

The dictionary is the en_US one, so British spellings are reported: colour is unknown and suggests color, organise suggests organize. That is the dictionary talking, not a judgement — the Consistent Spelling Checker is the tool for reviewing US and UK spellings as a style choice.

Suggestions are produced by walking character edits and keeping the candidates that exist in the dictionary — there is no hand-written correction table anywhere in this tool. The scanner skips 6 kinds of code-like span (URLs, email addresses, Markdown links, file names, snake_case), words shorter than 2 letters, and whichever of the three checkboxes above are ticked.

It checks spelling only. It has no grammar, no punctuation and no sense of meaning: "form" and "from" are both correct words to it.

This spell checker names the dictionary behind it — the en_US Hunspell word list from the dictionary-en package, read by nspell and loaded in your browser on this page alone — so its misspellings and its suggestions can be checked rather than trusted.

What is Spell Checker?

A spell checker is only as good as the word list behind it, so this one names its own. The words come from the dictionary-en package, a Hunspell en_US dictionary derived from SCOWL, and the lookups are answered by nspell, which walks character edits to propose corrections. Both are open-source (MIT, and MIT AND BSD, about 41 KB and 562 KB unpacked), and both are imported lazily, so the dictionary text is fetched only by a visitor who opens this page. No word list was typed from memory into this tool.

Type or paste a draft and the scanner splits it into word-shaped tokens, decides which ones are worth looking up, and asks the dictionary about the rest. Six kinds of code-like span are skipped before any lookup: web addresses, email addresses, Markdown links, file names with a known extension, snake_case and SCREAMING_SNAKE identifiers, and anything inside backticks. Tokens shorter than two letters are ignored, and three checkboxes control the rest — ALL-CAPS acronyms, words containing digits, and camel-case brand shapes are skipped by default. The code-like spans are always skipped, because a URL is not a spelling and https is not a misspelling.

Suggestions are generated, never hand-written. For each unknown word nspell walks single-character edits — deletions, transpositions, replacements and insertions — and keeps only the candidates that exist in the dictionary, which is why receive appears for recieve and separate for seperate. Up to six are shown, and suggestions are computed for the first 40 distinct unknown words, so a draft full of odd tokens cannot turn a millisecond into a stall. The remaining words are listed with a note that no suggestions were computed rather than being left to look mysterious.

The dictionary is the American English one, and the page says so because the results show it. colour, organise and centre are reported as unknown and offered color, organize and center. That is the dictionary talking, not a judgement about correctness: both spellings are English, and the tool would be lying if it pretended every reader shares one convention. If your document is British, add the handful of forms you rely on to the Words to accept field, or review the whole draft with the Consistent Spelling Checker, which treats US and UK forms as a style choice instead of an error.

Names, jargon, brand names and code-like words are flagged as unknown, because a general dictionary has no way to hold them. Racira, Kubernetes, Nikhil and localhost are all misspellings to this tool. That is not a quirk to hide: it is the honest consequence of using a real dictionary, and the Words to accept box exists for exactly the proper nouns your work depends on. The summary separates words that were looked up from tokens that were skipped, so you can always see the size of the part the dictionary actually judged.

What it cannot do is just as clear. It checks spelling and nothing else — no grammar, no punctuation, no sense of meaning — so form and from are both correct words to it, and a sentence can be spelled perfectly while saying the wrong thing. When a report is long, work through the most frequent words first, since a wrong spelling repeated is usually one habit rather than many mistakes. Nothing is uploaded: the dictionary and your text stay in the browser, and closing the tab removes both.

How to use Spell Checker

  1. Paste your draft; the dictionary loads once for the page and lookups begin immediately after it is ready.
  2. Read the unknown words table from the top — the most frequent spelling is usually the one worth fixing first.
  3. Add the names, products and jargon you use to the Words to accept field so the real errors stand out.
  4. Apply the suggestions you agree with, remembering they come from edit distance against a dictionary, not from meaning.
  5. Download the report and keep it beside the draft while you make the changes.

When to use Spell Checker vs related tools

Use this spell check as the first mechanical pass over a draft, then hand the results to tools that answer different questions. When the question is which variant to use rather than whether a word exists, the Consistent Spelling Checker compares US and UK forms across the whole document, and the Homophone Confusion Detector covers the pairs a dictionary cannot judge at all, like their and there.

If a clean spell check leaves you wondering whether the prose itself works, the Grammar Checker runs its published list of mechanical sentence checks next.

Privacy & Security

This tool runs entirely in your browser — no data ever leaves your device. There is no server round-trip, no upload, no logging, and no account required. Your input is processed locally using client-side JavaScript and is never stored, transmitted, or accessible to anyone else. When you close the tab, everything disappears.

Frequently asked questions about Spell Checker