Scientists have discovered the first words in a charred scroll sealed in volcanic ash.

Artificial intelligence and experts in ancient texts have teamed up to decipher scrolls that were turned to charcoal by the eruption of Mount Vesuvius in 79 CE. These unique artifacts were found in the 1750s among the ruins of a luxurious villa in Herculaneum, an ancient Roman city destroyed along with Pompeii.
The scrolls from the so-called Villa of the Papyri could have given humanity new knowledge about the philosophical and literary works of the ancient world. However, their preservation has proven to be both a blessing and a curse: covered in volcanic ash and charred by high temperatures, they have survived, but have become fragile and almost illegible. Any attempt to unroll them by hand results in destruction.
For more than 250 years, scientists have tried to find ways to read these texts. In 2023, several tech companies launched the Vesuvius Challenge , offering cash rewards for successful attempts to decipher the scrolls using machine learning, computer vision, and 3D reconstruction techniques.
This week, the organisers announced a breakthrough: researchers managed to obtain the first images of the inside of one of the three scrolls stored in the Bodleian Library at Oxford University. According to the co-founder of the challenge, Brent Seales, this image contains more recoverable text than has ever been obtained from scanning the Herculaneum scrolls.
The scanning was carried out using the powerful Diamond Light Source , an X-ray source operating at the particle accelerator in Harwell (near Oxford). Then, artificial intelligence helped the researchers detect traces of ink, improve their visibility, and virtually unroll the scroll using segmentation technology.
However, AI is far from perfect: only individual words were made out of the text. One of the first to be recognised was the ancient Greek word for “disgust”. The project participants are confident that further improvement of the methods will allow them to decipher more fragments.
The scientists are calling on researchers of ancient texts to join the work to speed up the process of interpreting the text. Peter Toth, curator of the Greek collection at the Bodleian Library, noted that the research is just beginning. “We need better images, but the team is confident that we can further improve the clarity and readability of the text,” he told AP.
One of the goals of the project is to make the technology available to other institutions so that the remaining two Oxford scrolls do not have to be transported to the research center. In addition, the National Library of Naples holds about a thousand more similar scrolls that may also hide unknown works of the ancient world.

Artificial intelligence and experts in ancient texts have teamed up to decipher scrolls that were turned to charcoal by the eruption of Mount Vesuvius in 79 CE. These unique artifacts were found in the 1750s among the ruins of a luxurious villa in Herculaneum, an ancient Roman city destroyed along with Pompeii.
The scrolls from the so-called Villa of the Papyri could have given humanity new knowledge about the philosophical and literary works of the ancient world. However, their preservation has proven to be both a blessing and a curse: covered in volcanic ash and charred by high temperatures, they have survived, but have become fragile and almost illegible. Any attempt to unroll them by hand results in destruction.
For more than 250 years, scientists have tried to find ways to read these texts. In 2023, several tech companies launched the Vesuvius Challenge , offering cash rewards for successful attempts to decipher the scrolls using machine learning, computer vision, and 3D reconstruction techniques.
This week, the organisers announced a breakthrough: researchers managed to obtain the first images of the inside of one of the three scrolls stored in the Bodleian Library at Oxford University. According to the co-founder of the challenge, Brent Seales, this image contains more recoverable text than has ever been obtained from scanning the Herculaneum scrolls.
The scanning was carried out using the powerful Diamond Light Source , an X-ray source operating at the particle accelerator in Harwell (near Oxford). Then, artificial intelligence helped the researchers detect traces of ink, improve their visibility, and virtually unroll the scroll using segmentation technology.
However, AI is far from perfect: only individual words were made out of the text. One of the first to be recognised was the ancient Greek word for “disgust”. The project participants are confident that further improvement of the methods will allow them to decipher more fragments.
The scientists are calling on researchers of ancient texts to join the work to speed up the process of interpreting the text. Peter Toth, curator of the Greek collection at the Bodleian Library, noted that the research is just beginning. “We need better images, but the team is confident that we can further improve the clarity and readability of the text,” he told AP.
One of the goals of the project is to make the technology available to other institutions so that the remaining two Oxford scrolls do not have to be transported to the research center. In addition, the National Library of Naples holds about a thousand more similar scrolls that may also hide unknown works of the ancient world.