WordCloud Cheat Sheet

01Setup & Importonce per project

$ pip install wordcloud★
Pulls in numpy, pillow & matplotlib.
from wordcloud import WordCloud★
Everything lives on this one class.
from wordcloud import STOPWORDS, ImageColorGenerator
Built-in stopword set & image-based colouring.
import matplotlib.pyplot as plt
To display the result inline.
from PIL import Image; import numpy as np
Needed only when using a mask image.

02Three Ways to Buildthe core

WordCloud().generate(text)★
From a raw string — counts words for you.
WordCloud().generate_from_frequencies(freq)★
From a {word: count} dict you already have.
wc = WordCloud(width=800, height=400, background_color="white")★
Options go in the constructor, not the method.
wc.generate(text)
Returns the WordCloud object (chainable).
# build once → recolor / export many times
Layout is the expensive step; reuse it.

03Input & Frequencieswhat goes in

.generate(text)★
Tokenize + count + lay out, all at once.
.generate_from_frequencies({"data": 40, "code": 25})
Skip tokenizing — you supply the counts.
.process_text(text)
Just the tokenizer → a frequency dict.
.fit_words(freq)
Alias of generate_from_frequencies.
from collections import Counter
Counter(words) makes a ready freq dict.
df["col"].value_counts().to_dict()
Pandas counts → feed straight in.

04Canvas & Backgroundthe frame

05Font Sizingthe core mapping

06Layout & Orientationwhere words go

prefer_horizontal=0.9★
Fraction of words tried horizontally first.
prefer_horizontal=1.0
Keep every word horizontal.
random_state=42★
Fix the layout so it's reproducible.
repeat=True
Repeat words until the canvas fills.
# placement = spiral pack, largest first
Each word avoids the ones already placed.

07Text Processingtokenizing

08Stopwordswhat to drop

from wordcloud import STOPWORDS★
The built-in set (used when stopwords=None).
stopwords=STOPWORDS.union({"said", "one"})★
Extend the defaults with your own.
stopwords=set(["a", "the", "of"])
Replace the list entirely.
generate_from_frequencies(..)note
Ignores stopwords, collocations & regexp.

09Colour & Colormapthe palette

10Custom Colour Functionsfull control

color_func=lambda *a, **k: "steelblue"★
One solid colour for every word.
color_func=my_fn
Overrides colormap entirely.
# signature Vega gives your function:
word, font_size, position, orientation, font_path, random_state
from wordcloud import get_single_color_func
One hue, varied brightness per word.
ImageColorGenerator(img_array)★
Take each word's colour from a source image.
wc.recolor(color_func=ImageColorGenerator(img))
The classic "colour from the mask" trick.

11Masks & Shapesdraw in a shape

12Output & Exportget it out

13Display with matplotlibshow it

plt.imshow(wc, interpolation="bilinear")★
bilinear smooths the rendered words.
plt.axis("off")★
Hide the axes & ticks.
plt.figure(figsize=(10, 5))
Set the display size before imshow.
plt.savefig("out.png", dpi=300, bbox_inches="tight")
Save the figure (with any contour/title).
plt.show()
Render it.

14Command Lineno code needed

$ wordcloud_cli --text in.txt --imagefile out.png★
The whole thing from a shell.
$ wordcloud_cli --help
Flags mirror the constructor params.
$ pdftotext doc.pdf - | wordcloud_cli --imagefile wc.png
Pipe a PDF's text straight in.
$ ... --mask shape.png --contour_width 1
Masks work from the CLI too.

15Recipes & Patternscopy & adapt

column: .generate(" ".join(df["text"]))
Join a text column into one string.
counts: .generate_from_frequencies(counter)
A Counter or value_counts dict.
transparent: mode="RGBA", background_color=None
PNG that drops onto any background.
single hue: color_func=lambda *a,**k: "white"
On a dark background → clean monochrome.
shaped + coloured: mask + ImageColorGenerator
Words in a shape, coloured by that image.
reproducible: random_state=42
Same layout & colours every run.

16Introspect & Depsunder the hood

wordcloud cheat sheet