Font replacement based on visual similarity

A system, comprising:

• a machine learning system;
• and
• a document preparation module configured to:
• produce a visual appearance for multiple local fonts, each of the respective visual appearances comprising a font image including multiple glyphs;
• input the font images to the machine learning system;
• receive, from the machine learning system, a font descriptor and a version value of a font similarity model of the machine learning system for each of the font images, the font similarity model one of a plurality of accessible font similarity models;
• associate the font descriptors and version values with respective fonts of the multiple local fonts;
• and
• store the font descriptors and version values associated with the respective fonts of the multiple local fonts.

A computing device comprising:

• one or more processors;
• and
• a computer-readable storage medium having stored thereon multiple instructions that cause the one or more processors to perform acts comprising:
• inputting an image including multiple glyphs rendered using a font such that the image represents a visual appearance of the font to a font visual similarity model trained with machine learning, the font visual similarity model one of a plurality of accessible font visual similarity models;
• computing a font descriptor corresponding to the font responsive to the inputting, the font descriptor including font features derived from the visual appearance of the font by the font visual similarity model;
• associating the font descriptor with a version value indicative of a version of the font visual similarity model;
• determining a similar font descriptor by comparing the font descriptor to multiple font descriptors associated with the version value that respectively correspond to multiple available fonts;
• ascertaining from among the multiple available fonts a similar font that corresponds to the similar font descriptor;
• and
• controlling presentation of a document using the similar font.