The flickering images of Hollywood’s Golden Age, and the kinetic energy of classic martial arts films, are receiving a 21st-century makeover. Artificial intelligence is increasingly being deployed to restore decades-old films, enhancing image quality, repairing physical damage, and reconstructing degraded audio. This isn’t simply about making old movies look “new”; it’s a race against time to preserve cinematic history as the physical materials themselves deteriorate. The technology is transforming a painstaking, manual process into one that is scalable and, crucially, economically viable for studios and archives alike.
For decades, film restoration has been a meticulous, frame-by-frame undertaking. Experts painstakingly remove scratches, stabilize shaky footage, and correct color fading. But with global film archives facing accelerating decay – a consequence of the inherent instability of nitrate and acetate film stock – traditional methods are struggling to preserve pace. According to a report by the Pulitzer Center, vast collections of culturally significant content are at risk of becoming permanently inaccessible. AI offers a potential solution, automating many of these tasks and dramatically reducing the time and cost associated with restoration. This allows institutions to address larger backlogs and, importantly, unlock the commercial potential of dormant intellectual property.
China’s Push to Preserve Kung Fu Cinema
The urgency of preservation is particularly evident in China, where a large-scale initiative is underway to digitally restore 100 classic martial arts films. Unveiled at the Shanghai International Film Festival, the project, spearheaded by the China Film Foundation, aims to revitalize foundational works of Chinese cinema, including Jackie Chan’s “Police Story,” Bruce Lee’s “Fist of Fury” and “The Big Boss,” Jet Li’s “Once Upon a Time in China,” and Jackie Chan’s “Drunken Master.” Radii reports that the focus is on enhancing image and sound quality while carefully preserving the original aesthetic and narrative integrity of these films. The project explicitly prioritizes restoration of existing footage, rather than attempting to digitally recreate lost or altered performances.
The Line Between Restoration and Reconstruction
While AI-assisted restoration is largely welcomed, a more controversial application is emerging in the United States, raising complex questions about authenticity and artistic intent. Startup Fable (formerly Showrunner) announced plans to recreate missing footage from Orson Welles’ 1942 film, “The Magnificent Ambersons.” The original footage was infamously cut by the studio against Welles’ wishes and subsequently destroyed to free up vault space. Futurism detailed the ambitious project, which involves filming new scenes on physically recreated sets and then using AI to overlay digital recreations of the original cast’s faces and voices onto the new actors.
Fable CEO Edward Saatchi, whose father co-founded the advertising firm Saatchi & Saatchi, has described the project as a labor of love, driven by a desire to complete Welles’ original vision. The company has stated it has no intention of commercializing the restored version, as it does not own the rights to the film. However, the technical challenges are significant. TechCrunch reported that early tests revealed AI-generated errors, including a two-headed rendering of actor Joseph Cotten and a persistent “happiness problem,” where the AI consistently generated facial expressions that didn’t align with the film’s somber tone.
Authenticity at Risk: The Hallucination Problem
The “Ambersons” project highlights a broader concern surrounding AI’s role in film preservation: the potential to undermine authenticity. As Bloomberg Opinion points out, AI upgrades can inadvertently distort original cinematography, alter color grading, remove film grain, and modify motion in ways that deviate from the filmmakers’ original artistic intent. These aren’t merely theoretical risks. TechCrunch reported that Fable’s AI reconstruction already produced visual details inconsistent with surviving reference material, raising questions about historical accuracy when AI fills in gaps with plausible, but potentially incorrect, information.
The line between restoration and “hallucination” is often blurry. When AI upscales resolution or reconstructs damaged audio, it’s essentially making educated guesses about the original content. While these inferences may be statistically reasonable, they aren’t necessarily factually accurate. This means the output can appear authoritative while subtly diverging from what was originally filmed. For archival and historical material, this discrepancy has consequences that extend beyond aesthetics. It raises questions about the integrity of the historical record and the responsibility of those wielding these powerful new tools.
The debate isn’t about whether to employ AI in film preservation – the necessitate is too great, and the benefits too significant. Rather, it’s about establishing clear ethical guidelines and technical safeguards to ensure that these technologies are used responsibly, prioritizing preservation of artistic intent and historical accuracy over purely aesthetic enhancements. The future of cinematic history may depend on it.
Looking ahead, the focus will likely shift towards developing more sophisticated AI algorithms that can better distinguish between genuine restoration and potentially misleading reconstruction. The Society of Motion Picture and Television Engineers (SMPTE) is currently working on standards for AI-based restoration, aiming to provide a framework for responsible implementation. The next major update from SMPTE regarding these standards is expected in the fourth quarter of 2026.
What are your thoughts on the use of AI in film restoration? Share your comments below and let us grasp how you think this technology will shape the future of cinema.
Worth a look
