TL;DR

The essentials

# Word diff: show which words changed between two files
$ wdiff old.txt new.txt
the [-quick-]{+slow+} brown fox
jumps over the [-lazy-]{+sleepy+} dog

Read one file from stdin (use '-' or omit the second arg)
$ echo 'the fast red fox' | wdiff - - <<< 'the slow red fox'
the [-fast-]{+slow+} red fox

Suppress common text (show only the changes)
$ wdiff -3 old.txt new.txt

Statistics to stderr (word counts + change breakdown)
$ wdiff -s old.txt new.txt
old.txt: 7 words  6 86% common  1 14% deleted  0 0% changed
new.txt: 7 words  6 86% common  1 14% inserted  0 0% changed

Re-format a unified diff at the word level (read from stdin)
$ git diff | wdiff -d

Colour, terminals, and CI

Force colour markers (e.g. for CI logs)
$ wdiff --color=always old.txt new.txt | less -R

Auto-pager (only when stdout is a tty)
$ wdiff -a old.txt new.txt

Quiet mode: exit 0/1 only, no output (for shell scripts)
$ wdiff -q old.txt new.txt && echo same || echo different

Advanced

Custom markers (e.g. ANSI bold instead of the default [- / {+ )
$ wdiff -w $'\033[1;31m' -x $'\033[0m' \
        -y $'\033[1;32m' -z $'\033[0m' \
        old.txt new.txt

Compare two files case-insensitively
$ wdiff -i OldFile.TXT newfile.txt

Show only the deleted words (e.g. for "what was removed?" reports)
$ wdiff -1 old.txt new.txt

When NOT to use wdiff

wdiff uses isspace() to detect word boundaries. On CJK text without spaces (Chinese / Japanese / Korean), the entire CJK line is treated as one "word" and the diff shows it as a single delete+add. For CJK-aware word diff, use dwdiff instead — it bundles libicu 78.3 for proper Unicode segmentation.