Siva Kalyan Designing size-specific variants of a typeface is a challenging and laborious process. This is due mainly to the difficulty of achieving visual consistency between letterforms at different sizes. Specifically, an optically reduced letterform should convey the same impression as the full-sized letterform, without any apparent reduction in weight, and with minimal loss of detail. So far, the only way to achieve this has been to adjust each letterform by hand—a labour-intensive process that requires many rounds of proofing and revision.
Siva Kalyan will demonstrate a way to significantly reduce the effort involved, by combining machine learning with a computational model of human vision. The result is a piece of software that can produce, for any given font, a variant that is optically scaled for any desired size, with minimal need for further adjustment. He will demonstrate the use of this software to generate optically-scaled variants of Noto Serif, and show the generality of this approach by presenting a new optically scaled typeface for Tamil.