A Computer Can Replicate Individual Handwriting With Impressive Accuracy


No matter how illegible your personal handwriting scrawl may be, a computer at University College London’s Visual Computing lab can probably read it. What’s more, it could mimic it. My Text in Your Handwriting, an algorithm created by university researchers, can recreate idiosyncratic handwriting patterns with a quick analysis, Smithsonian magazine reports.

Using an example of the handwriting, the algorithm learns the features of handwritten text—like spacing, line thickness, and pressure—to reproduce versions of the source’s writing. Handwriting is quite variable, both between individuals and within a person’s own handwritten texts (think of how different your signature can be depending on the day). The algorithm takes this into account, looking at each letter as a glyph, or pictograph, and then applying the patterns it learned to make standard letters into the unique strokes of an individual writer.

The lab tested the algorithm on the handwriting of famous historical figures such as Arthur Conan Doyle, Abraham Lincoln, and Frida Kahlo. It even reproduced a BBC reporter’s illegible scrawl on demand, with fair results, according to the man himself.

While writing by hand is rarely necessary these days (except in the states that mandate cursive proficiency in school) the technology could be used to add personal touches to flower deliveries or even just to get that handwritten feel on a letter without getting a hand cramp. Sorry, counterfeiters, but the BBC reports that it’s easy enough to verify under a microscope what text has been written by a computer, so it won’t be a great forging tool.

Their code is on GitHub if you want to check it out.

[h/t Smithsonian]

Know of something you think we should cover? Email us at tips@mentalfloss.com.