How to Make a Word Cloud: Visualise Open-Ended Survey Responses for Free and Read Them Correctly
A training survey came back with 100 open-ended comments. You don't have time to read them all, but you still need to report what people said. That's where a word cloud helps: the more often a word appears, the bigger it is drawn. Made carelessly, though, it fills up with filler words, splits “group work” into two words, and counts “not fun” as “fun”. This guide covers how word clouds count words, step-by-step instructions for survey comments, how to remove and merge words, how font size is calculated, which shapes and colours work best, and how to read the result without drawing the wrong conclusion, all with numbers we actually counted.
The short answer: count, clean up, then lay out
A word cloud counts the words in a piece of text and draws the most frequent ones in the biggest letters. Making one has three stages: (1) split the text into words and count them, (2) remove words you don't need and merge different spellings of the same word, and (3) size each word by its count and arrange it inside a shape. Stage 3 decides how it looks, but stages 1 and 2 decide whether it's any good.
| Common question | Answer |
|---|---|
| Can I make one for free? | Yes. A tool that runs in the browser needs no install and no sign-up |
| Does it work for open-ended survey answers? | Yes. Paste one response per line and count each word once per response to get the number of people |
| What does the word size mean? | Only the count. Position, colour and direction mean nothing |
| Can I draw conclusions from a word cloud alone? | No. Negatives and context are lost, so check the frequency table and the original answers |
| How do I put it in a report or slide? | Save a PNG with a transparent background and add the frequency table (counts and percentages) |
| Does it work for languages without spaces? | Only if the tool can split words properly. Japanese and Chinese need special handling |
Below, we use a 12-response training survey as an example and walk through the steps with real counts. The counts come from this site's word cloud generator with its default settings.
What a word cloud can and can't tell you
The strength of a word cloud is that it summarises a lot of text at a glance. Reading 100 comments carefully takes half an hour or more; a word cloud shows in seconds that “time”, “presentation” and “team” were the main topics. It suits meetings and reports where you want to convey the overall picture rather than precise numbers.
| It shows | It doesn't show |
|---|---|
| Which words were written most often | Whether a word was meant positively or negatively (was there too little “time” or just enough?) |
| The themes people talked about most | How words connect (what exactly was “fun”?) |
| How words rise and fall compared with a previous round (with the same settings) | Exact differences in counts (sizes are hard to compare by eye) |
| The phrases respondents tend to use | Rare but important comments (a word that appears once is drawn tiny) |
In other words, a word cloud tells you what people wrote about, not how they rated it. Keep that limit in mind and you can use it in reports without misleading anyone. We go into reading it properly in “Reading a word cloud correctly” below.
How words are split and counted in each language
In English, words are separated by spaces, so splitting is easy. The problem is the words that appear in every text: if “the”, “and”, “was” and “really” aren't removed, they become the biggest words and the cloud tells you nothing. Languages written without spaces, such as Japanese and Chinese, have the extra problem of deciding where one word ends and the next begins.
| Language | How the tool splits words | Points to know |
|---|---|---|
| English | Splits on spaces and punctuation with the browser's word segmenter | Common words (the, and, was…) are removed. “workshop's” is counted as “workshop”. Upper and lower case are combined |
| Chinese | The browser's word segmenter (Intl.Segmenter) splits the text into words | Common words such as 的 and 了 are removed. Words of one character aren't counted by default |
| Korean | Splits on spaces, then strips particles (은, 는, 이, 가…) | “학교에서” is counted as “학교”. Predicates such as “했다” aren't counted |
| Japanese | Treats runs of kanji, katakana or Latin letters as nouns | Verbs aren't counted; adjectives are counted in their dictionary form |
Here is how the English sample is counted. From “The group work was fun. Our team worked together all the way to the presentation.” the tool counts “group”, “work”, “fun”, “team”, “worked”, “together” and “presentation”; words such as “the”, “was”, “our”, “all” and “way” are dropped as common words. Notice two side effects: “group work” is split into two words, and “work” and “worked” (like “presentation” and “presentations”) are counted separately because the tool doesn't reduce words to their base form. Both can be fixed in the next stage.
| Input | Words counted | Note |
|---|---|---|
| The workshop was great. Great speakers and great food. | great ×3, workshop, speakers, food | Case is combined, so “Great” and “great” are one word |
| I loved the workshop's hands-on sessions. | loved, workshop, hands, sessions | “workshop's” becomes “workshop”; “hands-on” is split and “on” is dropped |
| Zoom / zoom / ZOOM | Zoom ×3 | Shown in the most frequent spelling |
| 2025 survey | survey | Number-only words are removed by default |
Step by step: a word cloud from open-ended survey responses
The steps are the same whether your answers come from Google Forms, Microsoft Forms, SurveyMonkey or a spreadsheet you typed up from paper. The key is to paste one response per line.
- 1
Copy the column of open-ended answers
Select the comment column in Excel or Google Sheets and copy it. You can also copy from the form's results view. Leave out the header cell (“Comments”, “Any other feedback?” and so on).
- 2
Paste it into “Enter your words”
Paste it into the text box of the word cloud generator and the cloud appears straight away. The language is detected automatically, and a line such as “12 lines · 62 words counted · 40 unique” shows how much was counted.
- 3
Turn on “Count once per line (response)”
Each word is then counted only once per person, however many times they wrote it, so the size means “how many people wrote this word”. That is much easier to explain in a survey report.
- 4
Go down the frequency table and exclude what you don't need
Press “Exclude” in the table and the word is added to the words to exclude and disappears from the cloud. Words that name the topic itself, such as “training” or “survey”, are big simply because everyone writes them; removing them makes the content easier to see.
- 5
Merge split words and spelling variants
Put phrases that get split, like “group work” or “customer service”, in “Count as one word”, and variants like “presentation=presentations” in “Count as the same word”.
- 6
Choose a shape and palette, then save
Under “Design”, choose the shape, aspect ratio, word direction and palette, and press “Shuffle layout” until you like the arrangement. Save as PNG or SVG, and keep the table with “Copy table” or “Save CSV”.
As an example, we counted the English version of a 12-response training survey (“The group work was fun. Our team worked together all the way to the presentation.”, “The presentation time was short and I wanted time for questions.” and so on). With the default settings it found 62 words, 40 of them unique. The top of the list looked like this.
| Rank | Word | Count (default) | Count (once per response) |
|---|---|---|---|
| 1 | fun | 4 | 4 |
| 1 | time | 4 | 3 |
| 3 | ideas · presentation · team · work | 3 each | 3 each |
| 7 | clear · explanation · group · hard · longer · read · slides · together | 2 each | 2 each |
| 15 | atmosphere · break · examples · handouts · new and 21 more | 1 each | 1 each |
“time” appears 4 times, but two of those are from the same person (“The presentation time was short and I wanted time for questions.”). Counted once per response it drops to 3, so you can correctly say “3 people mentioned time”. If you want to talk about numbers of people, always turn this switch on.
The clean-up that makes or breaks it: exclusions, phrases and variants
If your word cloud feels vague, the cause is almost always too little clean-up. Adjust it with these three boxes; separate entries with line breaks or commas.
| Box | When to use it | Example | Result |
|---|---|---|---|
| Words to exclude | The topic word or meaningless words are too big | training, survey, think | Removed from the cloud and the table |
| Count as one word | A phrase gets split into separate words | group work, customer service, hands-on | Counted as the whole phrase instead of “group” and “work” |
| Count as the same word | Plurals, verb forms, spelling variants | presentation=presentations | Both forms counted together as “presentation” |
In the sample survey, adding “group work” as one word and “presentation=presentations” as the same word changes the top of the table (counting once per response) to: fun 4, presentation 4, ideas 3, team 3, time 3, and “group work” appears as a word of its own. “Count as the same word” displays the first form you write, so you can also use it to merge verb forms: “enjoy=enjoyed=enjoyable” is counted as “enjoy”.
- Upper and lower case: by default “Zoom”, “zoom” and “ZOOM” count as one word, shown in the most common spelling
- Minimum word length: if there are too many very short words, choose “2+ characters” (on “Auto”, English is counted from 1 character and Chinese and Korean from 2)
- Number-only words: removed by default. In “2025 survey”, only “survey” is counted
- Number of words: the default is the top 100. With few responses, 30–50 words is easier to read; even with many, about 150 is the practical limit
- Only words that appear N+ times: if words that appear once make it cluttered, raise this to 2 or 3
How font size is calculated: linear, square root and logarithmic
Word size is calculated from the count. If size is simply proportional to count, the top word becomes huge and the rest become too small to read. Many tools soften the difference with a square root or a logarithm. This site's tool lets you choose between three options under “Count to size”.
We calculated how big each word would be when the most frequent word appears 100 times and the least frequent once, taking the biggest word as 100% (with “Size contrast” on Normal, the smallest word is 22% of the biggest).
| Count | Proportional (bigger differences) | Square root (normal) | Logarithmic (gentle) |
|---|---|---|---|
| 100 | 100% | 100% | 100% |
| 50 | 61% | 75% | 88% |
| 25 | 41% | 57% | 77% |
| 10 | 29% | 41% | 61% |
| 5 | 25% | 33% | 49% |
| 2 | 23% | 26% | 34% |
| 1 | 22% | 22% | 22% |
- Proportional: when you want to show how dominant the top words are. Suits data where a few words are overwhelmingly frequent
- Square root: a good balance between contrast and legibility for most data. Choose this if unsure
- Logarithmic: when counts vary hugely and you still want mid-ranked words to be readable. Differences become less visible
“Size contrast” sets how big the smallest word is relative to the biggest: Small is 42%, Normal 22% and Large 11%. To make a heart or star shape stand out clearly, use 60 or more words and set the contrast to Small so the outline fills with small words.
Design tips: shape, aspect ratio, direction and colour
Words are placed from biggest to smallest, each one searching outwards from the centre in a spiral for free space. That's why the most frequent word sits in the middle and smaller words fill the edges. On top of that, choose a shape and aspect ratio for where the image will be used.
| Where it's used | Aspect ratio (image size) | Suggested shape |
|---|---|---|
| Slides (PowerPoint, Google Slides) | Landscape 16:9 (1600×900 px) | Circle, rectangle, cloud |
| Blog header image, link preview on X | Landscape 1.91:1 (1600×838 px) | Rectangle, cloud |
| Instagram post | Square 1:1 (1400×1400 px) | Circle, heart, star |
| Stories, tall posters | Portrait 9:16 (900×1600 px) | Rectangle, diamond |
| Printed report | Landscape 4:3 (1600×1200 px) or portrait 3:4 (1200×1600 px) | Circle, rectangle |
- Word direction: “Horizontal only” is the most readable. “Mix horizontal and vertical” (35% vertical by default) looks livelier; English words are rotated 90°, while Chinese, Japanese and Korean words can be set in true vertical writing
- Palette: for reports use “Blues (for reports)” or “Monochrome”; for events and farewell cards “Colourful”, “Pastel” or “Cherry blossom”. With colouring set to “By rank”, colour also signals the top words
- Font: sans-serif is the most legible, serif feels formal, rounded feels friendly. Keep bold on so small words stay readable
- Spacing: “Tight” is dense and punchy; choose “Loose” if you want small words to be readable
- Your own shape: load a silhouette image (a dark shape on a white or transparent background) such as a logo, an animal or a map, and the words are arranged inside it
Reading a word cloud correctly: negatives, context and minority views
The most common mistake is reading “big word = something people liked”. Only the word counts are used, so the meaning of each sentence is gone. Negatives in particular simply disappear, because “not”, “no” and “wasn't” are among the common words that are removed.
| Response | Words counted | What happens |
|---|---|---|
| The session was fun | session, fun | As expected |
| The session was not fun | session, fun | The negative disappears and it counts the same as the positive |
| It wasn't helpful | helpful | Counts as “helpful” |
| Not clear at all | clear | Counts as “clear” |
| The explanation was unclear | explanation, unclear | The negative survives because it's part of the word |
Even if “fun” is big, some of those mentions may be “not fun”. For any word that matters, search the original answers (Ctrl+F in Excel) and check the context. To count negatives separately, add phrases such as “not fun” or “not clear” to “Count as one word” so they appear in the cloud as words of their own.
- Big words are topics, not ratings: “time” may be big because people complained there wasn't enough
- Read the words that appear only once too: rare but serious comments, about safety or harassment for example, end up tiny
- Don't rush to conclusions with few responses: with 12 answers, “time” appearing 4 times may be chance. Below about 20–30 responses, reading them all is more reliable
- Use the same settings when comparing rounds: different exclusions, word counts or sizing make sizes incomparable. Compare the counts in the table
Exporting for reports and slides
| What you want | Setting and how to save |
|---|---|
| Lay it over a slide background | Set the background to “Transparent (PNG)” and save as PNG |
| Print it or enlarge it without blur | Save as “High-res PNG (2×)” (3200×1800 px for 16:9) |
| Edit the text later in Illustrator or similar | Save as SVG (text is kept as text) |
| Paste it straight into a chat or slide | Use “Copy image” and paste |
| Put the counts in Excel | “Copy table” and paste as tab-separated data, or “Save CSV” (UTF-8 with BOM, so Excel opens it without garbled text) |
SVG keeps the words as text, so if the computer that opens it doesn't have the same font, a different font is used and words may overlap. If you just want to paste it in, PNG is the safe choice. Why CSV files sometimes look garbled in Excel is explained in how to fix garbled CSV files in Excel.
Add a line or two under the image, for example: “Created from the free-text answers of the post-training survey (48 responses). Each word counted once per response; topic words (training, survey) excluded. Word size shows the number of people who wrote the word.” With the data range, the counting method and the excluded words stated, readers can interpret it correctly.
How to ask open-ended questions that make readable word clouds
How readable a word cloud is gets decided largely before you collect the answers. If you only ask “Any comments?”, praise and complaints get mixed together, and the same word, “time” for example, ends up meaning both. Splitting the question makes each word cloud's meaning clear.
| Question | As a word cloud |
|---|---|
| Any other comments? | Praise and complaints mixed; you can't tell what the big words mean |
| What did you like? | Big words are directly “what people liked” |
| What could be improved? | Big words are directly “what to fix” |
| Describe this event in one word | Short answers with many adjectives; also great as a keepsake |
- Ask for good points and improvements in separate questions, and show the two word clouds side by side
- “In one word” or “in three words” questions produce consistent vocabulary and suit word clouds well
- Export form responses to Excel or Google Sheets and copy each question's column separately
- Required questions attract lots of “none” and “N/A”. Make them optional, or add those words to the words to exclude
Other uses beyond surveys
- Class reflections: type up students' feedback cards and show the class's voice at the start of the next lesson
- Meeting and interview notes: see which words dominated minutes or transcripts (for writing the minutes themselves, see how to write meeting minutes)
- Reviews: paste product or shop reviews and find the words customers use most (“service”, “friendly”, “price”)
- Checking your own writing: see whether you overuse certain words in an essay or report
- Gifts and cards: collect messages for a graduation, farewell or birthday and turn them into a heart-shaped or custom-shaped word cloud
- Social media and blogs: paste your own posts and find the themes you write about most
- Chinese, Japanese and Korean text: the language is detected automatically, and Korean particles are stripped (“학교에서” counts as “학교”)
If you also need to check the length of the text itself (an essay word limit, or the average length of survey comments), how to count words and characters explains the different ways of counting so you can report both together.
Common problems and fixes
| Problem | Cause | Fix |
|---|---|---|
| “the”, “and”, “really” are the biggest words | Common words weren't removed | Turn on “Remove common words” and exclude any that remain |
| Only the topic word is huge | A word everyone writes (“training”, “product”) is included | Add it to the words to exclude and note that under the image |
| “group work” shows up as “group” and “work” | The phrase is split at the space | Add it to “Count as one word” |
| “presentation” and “presentations” appear separately | Plurals aren't merged automatically | Add “presentation=presentations” to “Count as the same word” |
| One long answer dominates | One person wrote the same word many times | Turn on “Count once per line (response)” |
| Small words are unreadable | Proportional sizing makes the contrast too big | Choose square root or logarithmic and set the size contrast to Small |
| The heart or star shape isn't visible | Too few words to fill the outline | Use 60+ words and set the size contrast to Small |
| Words overlap when the SVG is opened | The font isn't installed on that computer | Save as PNG, or choose the font again in the program that opens the SVG |
FAQ
- What is a word cloud?
- It's an image that shows the words in a text, with more frequent words drawn bigger. It's used to show at a glance which words are common in text that would take a long time to read one by one, such as open-ended survey answers, feedback or reviews. Size shows the count; position, colour and direction have no meaning.
- Can I make a word cloud for free?
- Yes. This site's word cloud generator is free and needs no installation or sign-up. Paste your text and it counts the words; then choose a shape, palette and font and save as PNG or SVG. The text is processed only in your browser and never sent to a server.
- How do I make a word cloud from open-ended survey responses?
- Copy the comment column from Excel or your form results and paste it with one response per line. Turn on “Count once per line (response)” so each word's size equals the number of people who wrote it. Excluding words that name the topic itself, such as “training” or “survey”, makes the content easier to see.
- Why is my word cloud full of words like “the” and “really”?
- Because common words weren't removed. Words that appear in every text dominate unless they are filtered out. Turn on “Remove common words” and press “Exclude” in the frequency table for any that remain.
- How do I exclude specific words?
- Add them to “Words to exclude”, separated by line breaks or commas, or press “Exclude” next to a word in the frequency table. If it's for a report, note under the image which words you excluded to avoid misunderstandings.
- Can I merge plurals and spelling variants?
- Yes. In “Count as the same word”, write the forms on one line joined with “=”, such as “presentation=presentations” or “color=colour”. They're counted together and shown in the first form. You can also merge verb forms, such as “enjoy=enjoyed=enjoyable”.
- Does a big word in a word cloud mean people liked it?
- No. A big word only means it was mentioned often. “Not fun” can be counted as “fun”, so search the original answers for any word that matters and check the context. Adding the frequency table and a few quoted answers makes the message clear.
- Can I make a heart-shaped word cloud?
- Yes. Choose Rectangle, Circle, Heart, Star, Cloud or Diamond as the shape, or load a silhouette image to use any shape you like. For the outline to show clearly, use 60 or more words and set the size contrast to Small.
- How should I save a word cloud for PowerPoint?
- Set the aspect ratio to “Landscape 16:9 (slides)” and the background to “Transparent (PNG)”, then save as PNG so it sits on top of your slide background. For a large screen, choose “High-res PNG (2×)”. You can also use “Copy image” and paste it directly.
Tools in this article
Related articles
How to Write Meeting Minutes: A Template and How to Finish 90% During the Meeting
Minutes take a long time because you write them after the meeting. Reconstructing what people said from memory eats more time than the meeting itself. This guide covers the three elements every set of minutes needs, a template you can use as-is, a routine for finishing almost everything during the meeting, and how to review the minutes before sending them out.
Read article →How to Have Text Read Aloud: Free Text-to-Speech in Your Browser
You want to hear a draft to check it, listen to notes while memorising, or use a synthetic voice on a stream. For all of these, the speech synthesis built into your browser is enough. This guide explains how to have text read aloud with no installation, and gives guidelines for speed and voice settings.
Read article →How to Use the Pomodoro Technique: Tips for Staying Focused for 25 Minutes and Tracking Your Sessions
Work for 25 minutes, rest for 5. The rule is simple, yet many people find it falls apart once they try it. This guide covers the basic steps of the Pomodoro Technique, how to handle the interruptions that usually break it, and how to record your sessions as a picture of how you spend your day.
Read article →