Restoring damaged artwork is usually a slow, delicate process that can take weeks, years, or even decades.
But a new method developed at MIT may completely change that. Using artificial intelligence and a special printed “mask,” researchers have found a way to restore paintings in just a few hours—without permanently altering the original work.
The technique was created by Alex Kachkine, a PhD student in mechanical engineering who also happens to be a passionate art restorer.
His method blends digital tools with physical materials, allowing conservators to apply a digital restoration directly onto an original painting using a thin, removable film.
For centuries, restoring a painting has meant identifying each tiny damaged area and carefully repainting it by hand, often mixing colors to perfectly match the original.
Some paintings have thousands of damaged spots, and each one needs attention. The process is painstaking, requiring expert judgment and a steady hand.
In recent years, digital tools have helped experts make virtual restorations—repaired versions of paintings that exist only on screens or in print. However, these digital restorations have never been easily transferred back onto the original artwork—until now.
Kachkine’s new method uses AI to scan a damaged painting and digitally reconstruct what it might have looked like when it was first created. Then, a computer program maps out exactly where repairs are needed and what colors should be used.
This information is printed as a two-layer mask on an ultra-thin polymer film. One layer contains full-color details, while the other prints a white base layer that makes colors appear correctly on the painting.
Once printed, the mask is carefully aligned and laid over the original painting using a thin layer of varnish. The mask can later be removed easily with conservation-safe materials, and its digital file can be saved so future restorers know exactly what changes were made.
In a recent demonstration, Kachkine restored a 15th-century oil painting that had thousands of damaged spots.
His method automatically detected 5,612 regions that needed repair and used over 57,000 different colors to fill them in—all in just 3.5 hours. By comparison, he recalled a similar restoration he once did by hand that took nine months of part-time work.
Kachkine emphasized that while the technology is powerful, it must be used responsibly. Ethical concerns remain, especially when it comes to preserving the artist’s original intent. That’s why he believes the system should always be used in partnership with trained conservators who understand the history and meaning behind the artwork.
He developed the idea after visiting several art galleries during a road trip to MIT in 2021. He realized how many artworks are hidden from view in storage simply because they’re too damaged to display and take too long to restore.
His hope is that this new method will give those forgotten pieces a second life—and make more art accessible to the public.
“We’re creating a framework that can be built upon,” Kachkine said. “If this method continues to grow, it could make art restoration faster, more precise, and more reversible than ever before.”
