Create a word cloud from a text passage
Paste your source into the text field, choose the settings, and press Generate. The tool counts words and builds an image from the most frequent candidates. You can exclude common words, choose a maximum of 20, 40, 60 or 100 words, select one of three color palettes, and use a white or transparent background. Download becomes available after words have been drawn.
Start with a complete passage that answers a clear question. For a classroom discussion, that might be the answers to one prompt. For an editorial review, it might be the body of one article. Mixing unrelated sources produces a combined count, not separate clouds or a comparison between groups. Keep the source focused so the image has a meaning you can explain.
The Example button replaces the input with a short sample. It is useful for checking the controls before using your own material. Clear removes the current text and result. Changing the input or any setting also clears the old result, so press Generate again before downloading. This prevents a previous image from appearing to represent newly edited text.
During layout, you can cancel without deleting the source text. Cancellation does not produce a finished image; generate again when you are ready. Wait for a result or an error message before deciding what to change next.
Prepare text that gives a meaningful count
Use the actual passage rather than a web address or the name of a document. This page counts the characters you paste; it does not fetch the contents of a URL or import a document file. If you copy from an article, remove navigation labels, repeated headers and unrelated footer text before generating. Otherwise those words become part of the same count.
Repeated material can change the emphasis dramatically. A survey question pasted before every answer may become more prominent than the answers themselves. Remove repeated instructions when the aim is to visualize responses, but keep them when they are genuinely part of the material being studied. Record that preparation alongside the image so another person knows what the source included.
The maximum source length is 50,000 characters, and individual words must not exceed 80 characters. Some symbols use more than one character in the underlying text. An unusually long unbroken token can therefore stop generation even when the entire passage is short. Check pasted tracking strings or text joined without spaces if you see the word-length warning.
Do not remove duplicates automatically when repetition is the thing you want to measure. If the same response genuinely occurs several times, each occurrence contributes to frequency. If duplicated lines are accidental copying errors, the duplicate line remover can help prepare the source. Decide which situation applies before changing the input; cleaning a dataset and changing its evidence are not always the same thing.
How words are grouped and filtered
Case, accents and punctuation
The counter groups equivalent ways of writing accented letters and converts text to lowercase using the page language. Different capitalization therefore joins the same word, while ordinary accents are retained. A precomposed accented letter and its equivalent letter-plus-accent sequence are grouped. An accented word and an unaccented spelling remain separate; the tool does not silently correct one into the other.
Words are recognized from letters and combining marks, with internal straight or curly apostrophes supported. Curly apostrophes are treated like straight ones for counting. Numbers and emoji are not counted as words. Spaces, punctuation and hyphens separate words rather than becoming part of them. A phrase containing several words is not kept together as a single weighted label.
Plural forms, related verb forms and synonyms are not merged automatically. A word cloud is therefore a view of word occurrences, not an interpretation that knows which expressions mean the same thing. If you standardize spellings before generating, keep a note of the changes. Use the find and replace tool for deliberate corrections, not for silently rewriting evidence to make a preferred term larger.
Common-word exclusion follows the page language
Exclude common words is enabled initially. The English page uses an English list, while the Spanish and Portuguese pages use their corresponding lists. This is a fixed filter, not automatic language detection or a complete linguistic dictionary. A mixed-language passage may retain common words from the other language even when the checkbox is selected.
Turn the filter off when small grammatical words matter to your task, such as comparing sentence patterns. Leave it on when those words obscure the topic terms you want to inspect. Either choice changes the candidates and may change their relative visual emphasis. When comparing two sources, use the same language page and filtering choice so you are not comparing different counting rules.
Three examples with predictable frequencies
A small repeated vocabulary
Enter the following text on the English page with common-word exclusion enabled:
the and art art color color color design
The surviving counts are color: 3, art: 2 and design: 1. The words the and and are removed by the English filter. With such a small vocabulary, any of the available maximum-word settings can include all three candidates. The image arrangement can vary between generations, but moving a word to a new position does not change its count.
Use this sample to compare palettes and backgrounds. Generate after each change. The frequency table should stay the same because these settings affect presentation, not counting.
Capitalization does not remove accents
Enter CAFÉ café cafe with common-word exclusion disabled. The result contains café twice and cafe once. Uppercase CAFÉ is grouped with lowercase café; the unaccented spelling remains distinct. If an accented letter was copied as a letter followed by its separate accent, that equivalent spelling joins the accented group as well.
This helps with text copied from different sources, but does not repair every spelling variation. Review the table before assuming similar entries are duplicates: a real spelling difference may explain why they remain separate.
An apostrophe stays inside a word
Enter don't don’t DON'T with exclusion disabled. All three occurrences are counted as don't. Case is folded and the curly apostrophe is standardized, while the internal apostrophe stays within the word. By contrast, a phrase separated by spaces is several words, and a hyphen separates its surrounding parts.
For an empty-result check, try only 123 and punctuation. There are no letter-based words to display. A passage containing only filtered common words can produce the same outcome for a different reason. Add meaningful text or switch off common-word exclusion when that matches the task, rather than repeatedly generating from unchanged input.
Choose word count, colors and background
The maximum-word menu limits the candidate list after counting and filtering. It does not mean a word must appear 20 or 60 times. Candidates are sorted by descending frequency; equal counts are ordered alphabetically using the page language. If your passage has fewer distinct retained words than the chosen maximum, the tool does not invent additional words to fill the space.
Choose 20 for a more selective view, then try 40 or 60 when the smaller set leaves out useful detail. The 100-word setting includes more candidates, but it does not guarantee that every candidate will fit. Long words take more horizontal space, and several large words compete for room. More candidates can make the image denser without necessarily making it more informative.
Ocean, Forest and Berry are preset palettes. Colors help separate the displayed words; they do not encode sentiment, respondent groups or another measure in the source. The tool keeps words horizontal and uses a fixed font rather than offering custom fonts, rotation or shaped masks. Choose a different palette if the result needs more contrast against its intended surroundings.
White adds a white background to the exported image. Transparent leaves the space around the words transparent, which is useful when placing the PNG on a slide or document background. The preview area itself is white, so transparency may not look different there. Inspect the downloaded file against the actual destination background, especially when a darker background reduces the contrast of some word colors.
Read the image together with its frequency table
More frequent words are generally assigned larger type, but the image is not a proportional bar chart. The sizing uses a curved scale, and long words can be reduced to fit within the available width. You should not infer that a word appearing twice as wide occurred twice as often. Check the frequency table when the exact numerical difference matters.
The status below the image reports displayed words against selected candidates. If a result says 48/60, it means 48 of the 60 candidates were drawn. It does not mean the original text had only 60 words, or that 12 occurrences were deleted from the source. The layout may fail to place some candidates, and words reaching the image boundary may be omitted from the drawing.
The frequency table lists the selected candidates, including candidates that did not appear in the final image. It is not a complete export of every word in the source when the maximum-word setting cuts the list. Use it to inspect counts, spelling and the words considered for this particular cloud. A term absent from the picture may still be present in that table.
A large word shows repetition, not whether the passage approves of it. Negation, quotations and context can change meaning without changing a word's count. Read the original sentences before drawing conclusions about an audience or author. When presenting the image, add a short description of the source and filter choices rather than treating the cloud as a complete summary of the argument.
Download a PNG and finish the presentation
The download is a PNG image measuring 1536 by 1152 pixels, in a 4:3 shape. Its filename is word-cloud.png. It is a raster image, not editable text, an SVG or a spreadsheet. Put it into your document at a suitable size and inspect the small words at that size. Enlarging it substantially cannot add detail that was not in the exported pixels.
Keep the original passage separately if you expect to revise the cloud. Editing words inside the downloaded image is not part of this tool. To change the wording or filtering, return to the input, adjust it, generate another cloud and download the new result. A regenerated layout may differ even with the same frequencies, so save the specific image you intend to use.
If generation fails, try the built-in example to distinguish a source problem from a drawing problem. Check the character and word limits, lower the candidate count, and generate again. If downloading fails, retry after a completed result is visible. Do not treat a blank preview or a cancelled layout as a finished export.
For an accessible presentation, include a plain-language explanation or the important counts alongside the image. The PNG alone does not preserve the page's readable frequency table. A short caption identifying the source, the leading terms and the filtering choice gives readers a way to understand the result without relying entirely on visual size or color.