While all eyes in the world are focused on OpenAI, a project that has already been available for some time (since 2020) known as “Caligrapher AI” using a generative method built on a recurrent neural network, it is able to simulate human handwriting and adapt it to various styles according to needs.
Accessible through the Calligrapher.AI websitethe system is able to synthesize handwriting and edit it based on the style desired by users who have the possibility to choose 9 of them, as well as modify the speed cursors, the degree of readability and the width of the line.
But what is the peculiarity? As highlighted in the official documentation, unlike other similar tools designed to mimic handwriting, each sample generated by Calligrapher is unique and maintains the writing style, even if you change the parameters. Users are also given the option to export the result in SVG vector format.
Like any other neural network, the developers trained Calligrapher using a large dataset of handwriting samples. The database contains “shapes of handwritten English text captured on a blackboard”, with samples from 221 different people and more than 1,700 styles. The database includes 13,049 lines of text isolated and labeled in “online” and “offline” formats, for a total of 86,272 samples from a dictionary of 11,059 words.
The result is what you can see below, and which you can try directly on the Calligrapher website, completely free of charge.