Parikh, KinjalKaufman, Danny M.Levin, David I.W.Jacobson, AlecMasia, BelenThies, Justus2026-04-172026-04-1720261467-8659https://diglib.eg.org/handle/10.1111/cgf70415https://doi.org/10.1111/cgf.70415Typography is an essential component of visual communication. Editing and animating text appearance for graphic designs, title sequences, commercials, and logos remain highly skilled tasks requiring detailed, hands-on efforts from professional artists. Automating these workflows requires balancing legibility and aesthetics while enabling creative expression. Variable fonts, parametric extensions to traditional fonts, offer new opportunities for automated typographic design and animation. These fonts provide parameters that allow smooth variation across a design space. However, artists must still manually tune these parameters. We introduce differentiable variable fonts by distilling the variable font specification into a compact mathematical formulation that connects font parameters to vector graphics representations of text. This enables gradient-based optimization of energies defined on vector graphics control points or rasterized images through differentiable SVG rasterization. We demonstrate applications including direct shape manipulation, overlap-aware modeling, physics-based text animation, and automated font-design optimization.CC-BY-4.0Parametric curve and surface modelsMedia arts10.1111/cgf.7041510 pages