Type revival has always been part archaeology, part invention. A designer works from scanned specimen sheets, worn metal type, or a handful of surviving proofs, and has to infer the letters that were never cut.
Machine learning is now doing some of that inference — trained on a family's existing glyphs, models can propose plausible extensions into weights, scripts, or characters the original foundry never got around to.
The results still need a trained eye. But the bottleneck is shifting from drawing every glyph by hand to editing and curating what a model proposes, which is a very different skill.
#type design
#revivals
#AI