Word cloud generator
Turn any text into a word cloud, with each word sized by how often it appears. Everything runs in your browser — download as PNG or SVG when you are happy with it.
Comma separated. Useful for words you already know dominate the text.
Paste some text to build your word cloud.
| # | Word | Count | Share |
|---|
How a word cloud is built
Three steps turn a block of prose into a picture. First the text is tokenised — split into words at spaces and punctuation, with apostrophes kept so "don't" survives as one word. Then each distinct word is counted, after discarding the filler the count would otherwise be swamped by: stop words, numbers, anything below the minimum length, and whatever you have added to the ignore list. Finally the surviving words are sorted by count, the most frequent kept, and each assigned a font size scaled from its count.
The size mapping is less obvious than it looks. Scaling font size in direct proportion to count makes the largest word overwhelm everything when one term dominates, so this tool scales by the square root of the count instead. That makes the area a word occupies roughly proportional to its frequency, which is closer to how the eye actually compares the words on screen.
Placement is the last problem, and the interesting one. Each word is positioned by starting at the centre of the canvas and walking outwards along a spiral, testing at every step whether its bounding box overlaps anything already placed, and settling in the first gap that fits. Because the biggest words go first, they claim the middle and the smaller ones fill in around them — which is what produces the characteristic dense, roughly oval shape. When a word cannot find a gap anywhere on the spiral it is skipped, and the count beneath the cloud tells you how many were dropped.
The history of word clouds
The idea of sizing words by frequency to make a picture predates the web. In 1976 the social psychologist Stanley Milgram ran an experiment asking residents to name Paris landmarks, then drew a map of the city with each name set in a size proportional to how many people had mentioned it — a weighted word visualisation in all but name. The web version arrived with social bookmarking: Flickr's tag cloud in 2004, followed quickly by Delicious and Technorati, made the "tag cloud" one of the defining visual motifs of Web 2.0, and for a few years no site was complete without one.
What people picture today, though, is Wordle — not the 2021 guessing game, but the 2008 visualisation tool written by Jonathan Feinberg at IBM Research. Feinberg's version abandoned the alphabetical, line-wrapped layout of tag clouds in favour of tightly interlocked words at varied angles and colours, and it was attractive enough that the aesthetic escaped data visualisation entirely into classrooms, conference slides and newspaper graphics. The spiral placement approach this tool uses descends from that lineage, by way of Jason Davies' open-source d3-cloud implementation.
What word clouds are good and bad at
Word clouds are excellent at one thing: giving a viewer an immediate, pre-verbal sense of what a body of text is about. For a quick read on a set of survey responses, the themes in customer reviews, the vocabulary of a draft, or the subject of a document you have not read, nothing conveys the gist faster. They are also genuinely good at surfacing surprises — a term you did not expect to be prominent is obvious at a glance in a way it never is in a column of numbers.
They are correspondingly poor at anything precise, and it is worth knowing why before you put one in front of an audience. Frequency is not importance: a word repeated because of a quirk of phrasing looks exactly like a word repeated because it matters. There is no sentiment or context, so a cloud of complaints and a cloud of compliments about the same product look much the same. Word forms fragment, splitting "manage", "manages", "managing" and "management" into four smaller entries that understate a single theme. Long words take more space than short ones at identical font sizes, so they read as more prominent than they are. And comparing two clouds is close to meaningless, because the layout is partly random and the size scale is relative to each cloud's own maximum. Data journalists have been making this criticism for well over a decade, and it is fair.
The practical conclusion is to treat the cloud as an opening illustration rather than evidence. When you need to state something exact — this term appeared 47 times, that one 12 — read it off the frequency table under the cloud, or use the word frequency counter, which gives every word with counts and percentages.
Settings that change what you see
Four controls do most of the work. Stop words should normally stay on; with them off, "the" and "and" will dominate any English text and tell you nothing. Max words trades detail for legibility — 25 gives a bold, readable headline graphic, while 200 gives texture at the cost of a crowd of unreadable small words. Minimum word length is the quickest way to clear out residue like "it" and "is" that slipped past the stop list. And the ignore list is the one people reach for last and should reach for first: in a set of reviews for one product, the product's own name is usually the biggest word and the least informative, and removing it lets everything else become visible.
For appearance, the heavy display font produces the familiar poster look, mixed rotation packs words more tightly at some cost to readability, and a transparent background is what you want when the cloud is going onto a coloured slide rather than a white page.
Frequently asked questions
How do I make a word cloud from text?
Paste or type your text into the box and the cloud draws itself — there is no generate button to press. The tool splits the text into words, counts how often each appears, discards common filler words, and sizes the rest so the most frequent are the largest. From there you can set the word limit, pick a palette and font, choose how much rotation you want, and press Reshuffle to try a different arrangement of the same words. When it looks right, download it as a PNG for slides and documents or an SVG for print.
Why are some words missing from my word cloud?
Four filters can remove a word, and one layout limit can too. Stop words such as "the" and "and" are discarded by default; words shorter than the minimum length are skipped; only the top N most frequent words are drawn, so rarer words fall outside the limit you set; and anything in the custom ignore list is excluded. Beyond those, a word can be dropped because the canvas simply ran out of space for it — the status line under the cloud reports how many were skipped for that reason, and raising the word limit, shortening the text or choosing a narrower font usually makes room.
What are stop words and should I remove them?
Stop words are the structural words that every English text is full of — the, and, of, to, is, that, it. In ordinary prose "the" alone accounts for roughly 7% of all words, so leaving stop words in produces a cloud whose largest entries are words that say nothing whatsoever about the subject. Removing them is the right default for essentially every practical use. The exception is linguistic or stylistic analysis, where the proportion of function words is the thing being measured — authorship attribution studies, for instance, rely heavily on exactly those words.
Can I download the word cloud for a presentation or print?
Yes, in two formats chosen for different jobs. PNG exports at twice the size shown on screen, which is ample for slides, documents, blog posts and social images, and the transparent background option means it can sit on a coloured slide without a white rectangle around it. SVG is a vector file, so it scales to any size — a poster, a banner, a printed report — with no blurring at all, and the words stay editable text when opened in Illustrator, Figma or Inkscape, so a designer can restyle or recolour it afterwards.
Does a bigger word mean a more important word?
It means a more frequent word, which is a different claim. A word cloud has no notion of meaning, context or sentiment, so it cannot distinguish praise from complaint, and a word that recurs because of a habit of phrasing looks identical to one that recurs because it matters. Related forms also fragment — "manage", "manages" and "management" become three separate smaller words rather than one theme — and because longer words occupy more area at the same font size, they read as more prominent than their count justifies. Use the cloud for a fast impression of subject matter, and the frequency table beneath it whenever you need to say something precise.
Is my text uploaded anywhere?
No. The word counting, the layout calculation and the drawing all happen inside your browser, using JavaScript and an HTML canvas, and no part of your text is ever sent to a server. The only copy kept anywhere is in your own browser's local storage, which is what lets your draft survive a page reload; clearing the text box or clearing your browser data removes it. That makes the tool safe for material you could not paste into a hosted service — interview transcripts, raw survey responses, internal documents or unpublished writing.