What AI Restoration Can Do Today
For most of film history, restoring a damaged or deteriorating movie meant one thing: patience. Dozens of trained technicians, working frame by frame, manually painting out scratches, reducing flicker, and stabilizing footage by hand. A single feature film could take months and cost hundreds of thousands of dollars. Only the most commercially important titles ever got restored at all.
That era is ending. Artificial intelligence has fundamentally changed what film restoration means — who can afford it, how long it takes, and what quality is achievable. This is the story of how AI is rewriting the rules of one of cinema’s most demanding crafts.
The Old Way: Why Manual Restoration Was Broken
To understand the revolution, it helps to understand what came before.
A standard 35mm film frame is roughly 18×24mm. A two-hour feature contains approximately 172,800 frames. Each frame can carry dozens of defects — dust particles, vertical scratches, chemical stains, flicker from inconsistent exposure, grain patterns, and the progressive fading of dyes that began the moment the film was processed.
Manual restoration meant a digital artist would examine each frame on a monitor, identify the defects, and paint them out using retouching software — frame by frame, defect by defect. Even with skilled artists working efficiently, a single reel of film might take weeks to restore properly.
The economics were brutal. Only major studios, national archives with government funding, or exceptionally valuable films could justify the cost. The vast majority of the world’s film heritage — including millions of hours of home movies, newsreels, regional cinema, and culturally important documentaries — was simply left to deteriorate.
What AI Changes
Modern AI-based film restoration works differently at every level.
Instead of asking a human to identify and remove each defect, AI systems are trained on enormous datasets of damaged and clean film pairs. Through deep learning, these systems develop a mathematical understanding of what film damage looks like across an infinite variety of contexts — different film stocks, different types of damage, different lighting conditions.
The result is software that can automatically detect and remove defects across thousands of frames per hour, with a consistency and precision that human artists cannot match over long projects.
At Algosoft, we have spent over 25 years developing VIVA — the leading automatic film restoration software, now used by the Library of Congress, Hollywood studios, and archives across Europe and Asia. VIVA uses AI-based complex mathematical algorithms that track everything happening from frame to frame, enabling picture improvements that have to be seen to be believed.
What AI Restoration Can Do Today
Modern AI film restoration handles defects that once required armies of artists:
Dirt and Dust Removal
Blotches, emulsion damage, and frame contamination are detected and removed automatically, without affecting the underlying image. VIVA’s Total Clean AI and Scratch & Dust AI can distinguish between genuine film content and contamination at the pixel level.
Scratch Removal
Vertical scratches — both dark and light — are among the most technically demanding restoration challenges. They run the full length of a frame, can vary in width and density, and often resemble genuine vertical structures in the image background. VIVA’s dedicated DeScratch neural network, introduced in 2017 as the first application of neural networks to scratch removal in film, handles this automatically.
Flicker Elimination
Exposure flicker — that characteristic pulsing brightness of early film — is caused by inconsistent frame exposure during original photography or projection. AI algorithms analyze the luminance of each frame in sequence and correct inconsistencies invisibly.
Stabilization
Unwanted shifts and rotations and warping errors are corrected through temporal stabilization algorithms that track the movement of the image as a whole and compensate automatically.
Film Grain Preservation
Critically, VIVA does not simply smooth or blur the image to hide damage. Film grain — the natural texture of the photographic medium — is preserved as a fundamental part of the film’s visual character. Destroying grain in the name of “restoration” produces the artificial, plasticky look that has rightly attracted criticism when applied to beloved classics.
Advanced Capabilities
Recent releases of VIVA have introduced capabilities that address even the most complex damage: Spider Web mode for crazing — the dense networks of fine cracks caused by emulsion contraction or water damage; Color Stain Removal for the chemical staining that appears as green or other colored blotches; and DeWarping tools for correcting localized image distortions at the pixel level.
A Real Example: Metropolis (1927)
The removal of a dense “forest” of vertical scratches — more than a hundred per frame — from the 16mm dupe negative copy of Metropolis discovered in Buenos Aires stands as one of the most striking demonstrations of how modern artificial intelligence is transforming digital film restoration.
For decades, Metropolis existed only in incomplete and fragmented versions. This changed in 2009, when a nearly complete nitrate print of the film was discovered in an Argentine archive. Its condition was critical: severe flicker, a dense forest of vertical scratches, dirt, and mechanical damage made viewing nearly impossible.
In 2010, a specialized deflicker in VIVA was used to address the flicker, allowing the newly discovered footage to be integrated into the existing restoration. Yet only the AI-enhanced evolution of VIVA has, for the first time in more than 80 years, enabled a level of image recovery deep enough to present these scenes as they were originally envisioned by Fritz Lang — removing scratches at a density and complexity that no previous technology could address automatically.
Metropolis (1927): Rediscovered Argentina Reel — scratch removal with VIVA AI
Not Just for Studios: AI Restoration for Everyone
One of the most significant developments in recent years is the democratization of film restoration. AI has made professional-quality restoration accessible to individuals and small organizations that could never have afforded manual restoration services.
VIVA Auto/Lite — the consumer version of the same technology trusted by the Library of Congress — is free to download and use for experimentation. Users pay only when they are satisfied with the results and wish to release footage without a watermark.
For those without compatible hardware, DigiFilm Restore (digifilmrestore.com) provides the same VIVA technology as a cloud service: upload your footage, select a preset, and receive a professionally restored file. The first 30 seconds are always processed free of charge.
This means that the 8mm home movies sitting in your parents’ attic — the birthday parties, the summer vacations, the faces of people you love — can now be restored to a quality that would have been impossible just a decade ago.
The Limits of AI Restoration
AI film restoration is not magic, and it is worth being clear about what it cannot do.
Severely damaged footage — frames that are physically missing, or material so deteriorated that there is no underlying image information remaining — cannot be restored by any technology. AI works with what is there; it cannot invent detail that has been lost.
For the highest-stakes archival restorations, human oversight remains essential. VIVA is designed to work alongside restoration professionals, not replace their judgment. The best results come from combining the speed and consistency of AI with the expertise of experienced restorers who can evaluate results and adjust parameters for specific material.
Looking Forward
The pace of development in AI film restoration has accelerated dramatically. Capabilities that seemed impossible five years ago — automatic color stain removal, dewarping for elastic distortion, spider web mode for crazing — are now standard features of professional restoration software.
As AI continues to improve, the barrier between damaged film and its restoration will continue to fall. The world’s film heritage — from Hollywood classics to home movies, from national archives to local newsreels — is within reach of restoration in a way it never has been before.
For archives, studios, and individuals with footage that deserves to be seen, this is a remarkable moment.
VIVA by Algosoft is the world’s leading automatic AI film restoration software, used by NBC Universal, China Film, Prasad, the Library of Congress, and other restoration facilities worldwide. Download VIVA free or try DigiFilm Restore to see what AI can do for your footage.