This working paper proposes a structural reading of selected Voynich folios that does not depend on linguistic decipherment. It argues that support, margins, vegetal forms, pigments, and text distribution can be read together as layered visual registers, producing pages that function less like simple illustrated text and more like organized visual scores. The aim is not to solve the manuscript, but to offer a reproducible descriptive framework that complements codicology, paleography, and image analysis.