The First Herculaneum Scroll We Can Read Without Opening

For most books, reading begins with access to a surface. Open the codex. Unroll the scroll. Expose the page. The Herculaneum papyri have made that ordinary sequence unusually dangerous for more than two centuries.

The rolls were carbonized by the eruption of Vesuvius in 79 CE. Their survival is miraculous in the least decorative sense of the word: many are black, compressed, brittle objects whose layers can be damaged simply by trying to separate them. Earlier attempts at physical unrolling recovered texts, but also left a history of torn papyrus and lost material.

In June, researchers reported that PHerc. 1667 had been completely virtually unwrapped and read without physically opening the roll. The Vesuvius Challenge describes it as the first Herculaneum scroll to be virtually unwrapped and read end to end.

The popular shorthand is that AI read an ancient scroll. That leaves out most of the interesting machinery.

Researchers used high-resolution phase-contrast micro-CT at the European Synchrotron Radiation Facility to map the interior of the roll. From those scans, computational methods had to follow crushed papyrus layers through three dimensions, recover their geometry, turn warped surfaces into something a scholar could inspect, and identify ink. Papyrologists then still had to read and judge the result. Machine learning belongs inside that process, alongside imaging physics, geometry, software, conservation, and very traditional scholarship.

The page on screen was never lying flat inside the scroll

A facsimile normally begins with a visible page. The photograph may be excellent or poor, but the underlying surface was there to be photographed.

PHerc. 1667 presents a different object to the reader. The flat image is assembled from the geometry of papyrus that remains wound around itself. What appears as a page on screen corresponds to a real writing surface, but nobody opened the roll and found that page waiting in the form we now see it.

The practical advantage is that scholars can check the reconstruction against the scan. If a letter or line is doubtful, they can return to the corresponding place in the three-dimensional data and ask whether the reconstructed surface and the visible ink support that reading. A transcription alone cannot show that.

Another Herculaneum papyrus, PHerc. Paris 4, shows how imaging is improving as well. In that scan, ink can be seen directly in the tomographic volume, giving researchers another way to check what the reconstructed surface shows.

The digital image is therefore not a substitute for the scroll. It is a way of inspecting parts of the scroll that cannot safely be exposed by hand.

Herculaneum adds a new kind of textual transmission

Ancient texts have always reached later readers through transformations. Oral poetry becomes writing. Rolls are recopied into codices. Medieval manuscripts produce printed editions. Editors regularize damaged passages. Translators recast Greek or Latin into another language. Digitization adds searchable text and images to that chain.

Those transformations are familiar enough that readers often stop seeing them. A clean modern page can hide the damaged manuscript, disputed reading, editorial decision, or typographical convention behind it.

Virtual unwrapping is unusual because the physical state of the source is the thing being left alone. The scroll does not have to be mechanically converted into an open object before its writing can enter the rest of the textual tradition.

That has large consequences at Herculaneum. The villa's papyri constitute the only large-scale library known to survive from classical antiquity, and many rolls remain unopened. If the method used on PHerc. 1667 can be repeated reliably, the surviving corpus can grow without making physical unrolling the price of admission.

The same research has already produced title and author evidence for another roll, PHerc. 139, identifying it as Philodemus' On Gods, Book 8. That does not mean scanners are about to pour out a shelf of famous lost masterpieces. The work is still difficult, expensive, and dependent on the condition of individual rolls. “Scalable” in a research paper should not be mistaken for “automatic.”

Still, an old conservation dilemma has changed. A closed carbonized roll is no longer necessarily an unreadable one.

Sources

• Giorgio Angelotti et al., Complete virtual unwrapping and reading of a rolled Herculaneum papyrus (2026)
• Vesuvius Challenge
• European Synchrotron Radiation Facility, “The day the Herculaneum scrolls began speaking again”