Computing the Shapes: Parametric Fonts and a Hanzi AI

Designing typefaces for hanzi (Chinese characters) has long been a problem because of the script’s tremendous character set and the complicated construction of individual characters. Hence efforts to automate the designing of glyphs have been around for decades.

This presentation is about why, what, and how to generate hanzi glyphs automatically. The old parametric genre still holds its place, while the new deep learning approach quickly awes us with its magic power. Yehang Yin, Rui Zhang, and Zewen Chen will demonstrate how to modify existing typefaces parametrically, and how mixing up styles in AI results in new faces never seen before. By combining both approaches, they can even generate non-existing hanzi in quality bitmap form.

Besides its application in assisting with CJK type design, the team discovered that this technology has enormous potential to make traditional fine arts accessible to a wider audience. They hope to inspire type designers working with other scripts to see AI as an imaginative tool that can be helpful during the design process.