Skip to content
TextArray
100% local

Typo generator

Sprinkle realistic keyboard typos over clean text — adjustable rate, QWERTY or QWERTZ.

Input
Output
Related tools
Find and replaceBasic tools
Text similaritynewBasic tools
Shuffle textText manipulation
Leet speak translatorStylish text
Sort wordsnewText manipulation
Swap wordsnewText manipulation
Typo generator
Software that handles human input eventually meets human typing — and human typing is messy. This tool manufactures that mess on demand: paste clean text, set an error rate, and it comes back with the four mistakes real fingers make. A neighboring key hit instead of the right one (tge for the), two letters transposed (teh), a letter doubled (thee) or dropped (th). The substitutions follow an actual keyboard adjacency map — QWERTY or QWERTZ — so a t becomes r, y, f or g, never a distant x, which is exactly the error distribution fuzzy-matching algorithms are tuned for.
Each typo kind has its own toggle, so you can produce pure transposition data for testing a Damerau-Levenshtein implementation, or substitution-only strings of unchanged length. The rate runs from a subtle 1% of letters to a barely readable 20%, and the tally reports how many typos actually landed. Case is preserved through substitutions, and everything that is not a plain letter — digits, punctuation, accented characters, emoji — passes through untouched.
Typical uses: test fixtures for spellcheckers and "did you mean" features, fuzzy search evaluation sets, training data augmentation for NLP models, demonstrating why exact-match deduplication fails on human-entered names, or checking that your form validation tolerates realistic input. The randomness comes from the browser's crypto source, and every run is different by design.
Everything runs locally in your browser — your text never leaves your device.
FAQ
Which typo kinds are simulated?
The four classics of typing-error research: substitution by a physically adjacent key, transposition of two neighboring letters, doubling and omission. Each has its own toggle, so you can isolate one error class for targeted test data.
Why does the keyboard layout matter?
Substitutions replace a letter with a physical neighbor, and QWERTY and QWERTZ disagree about where y and z live. Pick the layout your users actually type on, and the fake typos match the real ones your logs will show.
Why are accented letters left alone?
The adjacency map covers the 26 base keys, where the neighbor model is meaningful. Accented characters are typed via dead keys or long-press, and their real-world typos are different in kind — substituting them by base-key neighbors would produce unrealistic data.
Is the number of typos exact?
No — each letter independently has a rate-sized chance of a typo, so the count fluctuates around the expected value, run to run. That randomness is deliberate: uniform, evenly spaced errors are exactly what real typing does not look like.
Is my text uploaded anywhere?
No. The typos are generated entirely in your browser and your text never leaves your device.