Paste any text — an article, a meeting transcript, a speech, your own journal — and see what it is really about. Common filler words are stripped out, the survivors are sized by how often they appear, and the cloud is packed automatically so nothing overlaps. Export it as an image or as a vector file. Everything happens on your device; the text never leaves it.
Text is split on anything that is not a letter, lower-cased, filtered by length and by the stop-word list, then counted. Sizes are scaled by the square root of the count rather than the raw count: with raw counts a single word repeated a hundred times swamps everything else and every other word becomes unreadable. Square-root scaling keeps the shape of the distribution while leaving the smaller words legible.
Placement walks each word outwards along a spiral from the centre — the largest word first — and drops it at the first spot where its box does not touch a word already placed. That is why re-arranging gives a different but equally valid cloud: the spiral start angle changes, the packing stays valid.
A word cloud shows frequency, not importance, and it has no idea what a sentence means — treat it as a quick way to spot the vocabulary of a document, not as analysis. The PNG is exported at twice the displayed size; the SVG is true vector text you can edit or scale in a design tool. Your text is processed in the browser only: no request is made and nothing is stored.