Design · 11 min read

Word cloud design: choosing shapes, fonts and colour that stay readable

Why most word clouds turn to mush, and the specific choices — silhouette, typeface, palette, density — that keep one legible at poster size and on a phone.

Most word clouds fail the same way. Too many words, all fighting, in a typeface that was never meant to be read at nine pixels, coloured by a palette that puts mid-grey text on a mid-grey background. The words are correct and the image is unreadable.

Everything below is about the gap between those two things. It applies to any generator; the settings named are the ones in WordCloud Studio.

Density is a trade, not a quality setting

Density controls how tightly words pack. It is tempting to treat higher as better — a dense cloud looks impressive and expensive — but every percentage point of extra coverage is bought by shrinking the smallest words.

The honest way to choose is to decide the viewing distance first:

Where it will be seenAim for
Slide, viewed from the back of a roomLow density, 20–40 words, large minimum size
Report or web page, viewed at arm’s lengthMedium density, 60–120 words
Poster or print, viewed up closeHigh density, 200+ words, small minimum size is fine

A dense cloud that will be projected is not ambitious, it is just illegible. The words below the fold are decoration and everyone in the room knows it.

Raise the minimum font size before you lower the density

If the cloud is too busy, the instinct is to reduce the word count. The better first move is to raise the minimum font size and let the packer drop whatever no longer fits. That keeps the biggest, most meaningful words at full size and removes the ones that were never readable, which is exactly the trade you wanted.

Typeface: pick for the small sizes, not the big ones

In a word cloud the largest word will look fine in almost any typeface. The design is decided by the smallest ones, which is the opposite of how people usually choose a font.

What actually matters at small sizes:

  • Weight. Heavy, condensed faces pack far more tightly because their letters have less internal white space, and they hold up better when scaled down. This is why the classic dense word clouds all look like they use Impact, and why they do.
  • Even stroke width. Faces with strong thick-thin contrast — most display serifs, anything script — lose their thin strokes entirely below about twelve pixels. The word becomes a dotted smudge.
  • Open counters. The enclosed spaces in a, e, o and g fill in first when text gets small. Faces with generous counters stay legible several sizes lower.

A practical rule: choose one condensed sans for density, or one clean geometric sans for clarity, and use it for the whole cloud. Mixing typefaces inside a single cloud reads as an accident, not a choice, because the size differences already carry meaning and a second variable muddies it.

The exception is a deliberate two-tier design, where the three or four largest words use a display face and everything else uses a single neutral one. That reads as intentional. Five faces do not.

Colour: contrast first, palette second

The failure mode here is choosing a palette that looks lovely as a row of swatches and produces unreadable words when half of them land on the background. Two habits fix nearly all of it.

Keep every colour on one side of the background

On a dark background, every word colour should be light. On a light background, every one should be dark. A palette that mixes both will always have words that vanish, and which words vanish is decided by the packing algorithm rather than by you.

Vary hue, not lightness

Once all your colours are on the correct side, get variety from hue rather than from brightness. Six colours at roughly the same lightness look like a designed set and read evenly. The same six spread across a lightness range look like some words matter more than others, which conflicts with the sizing that already encodes importance.

If you want colour to carry meaning — categories, sentiment, source — set it per word rather than letting the tool assign randomly, and put a one-line key next to the image. Random colour is decoration and readers correctly ignore it. Assigned colour is data and they will read it, so it has to be right.

Shape: the silhouette has to survive being filled

Pouring words into a heart, a star, a map or an uploaded logo is the feature people come for, and it is where clouds most often fall apart. The rule is that a shape reads only if its outline survives being made entirely out of words.

Shapes that work:

  • Big, closed, convex forms. Hearts, circles, clouds, simple animal silhouettes, bold single letters. The eye completes the edge even when words interrupt it.
  • Solid logos. A logo with heavy strokes and no fine detail engraves into a cloud beautifully.

Shapes that do not:

  • Thin lines and outlines. A shape made of strokes thinner than about two words will read as a scattering of text, not a shape.
  • Fine internal detail. Faces, filigree, small text inside the logo. The detail is smaller than the smallest word and disappears.
  • Anything with important negative space. If the shape only makes sense because of a gap inside it, check that the gap survives at your density.

If you are uploading your own image, feed the tool the highest-contrast version you have. Dark or opaque areas become the shape and light areas become empty space, so a flat silhouette in solid black on white gives a far cleaner result than a photo or a gradient.

Test the shape at thumbnail size

Shrink the finished cloud until it is about the size of a postage stamp and look at it. If you can still tell what the shape is, it works. If it turns into a blob, no amount of colour or typography will rescue it, and the fix is a bolder silhouette rather than more words.

Export at the size you will actually use

Word clouds are laid out mathematically rather than as pixels, which means the same design can be rendered at any resolution without re-flowing. Take advantage of it:

  • Slides and web: HD or QHD PNG is plenty.
  • Print: 4K for A4, 8K or 16K for a poster.
  • Anything that might be resized later: SVG. It stays sharp at any size and the text remains real text, so it can be edited in a vector editor afterwards.

A transparent background is worth exporting alongside the flat one. It costs nothing and it saves the argument when someone drops the image onto a coloured slide.

The short version

  1. Choose density from viewing distance, not from taste.
  2. Raise the minimum font size before cutting words.
  3. Pick a typeface that survives at nine pixels: heavy, even stroke, open counters.
  4. One typeface, or a deliberate two-tier pairing. Never five.
  5. Keep every word colour on one side of the background.
  6. Vary hue, not lightness, unless colour means something.
  7. Use bold, closed silhouettes; test them at thumbnail size.
  8. Export SVG when the final size is not yet decided.