A video uploaded to YouTube has triggered a fervent debate among cinephiles, techno-geeks and filmmakers on different social media platforms. The video, a 4K high definition, upscaled ‘restoration’ of the short film Train Arriving at the Station at La Ciotat (popularly known as “Train arriving at the station”) – shot by the “fathers” of filmmaking – the Lumiere brothers.

The upscaled version of Train Arriving at the Station at La Ciotat (1895).

The 50-second-long film is the best know among the constellation of over 1,400 actuality films produced by Auguste and Louis Lumiere between 1895 and 1896. These films marked the formal beginning of film history. In Train arriving at La Ciotat, a train is shown arriving at a station in France, capturing a group of men, women and children, some disembarking and others embarking to continue their journey. Given that the movie camera was an unknown animal, the people appear strangely oblivious to the filmmakers and their filming. Lumiere’s violated the actuality code by directing the passengers. This film absorbed generations of cinephiles trying to imagine how it may have impacted viewers, who’d never seen a cinematic representation of reality before and for whom people and things only moved in real life.

The original Train Arriving at the Station at La Ciotat (1895).

Denis Shirayaev, an Artificial Intelligence developer who has been experimenting with open source video upscaling tools, had used available YouTube videos copies of the Lumiere film and other related visual material to make a version that is a huge improvement on the original. The Lumiere film is circulated largely as poor 35mm celluloid copies, not to mention the videos made from these degraded film materials. Compared to the fuzzy and blurred original clips, there is a profusion of details in Shirayaev’s version. For the first time one can discern the trees, cottages, lamp posts behind the rail track, or distinguish the faces of the men, women and children. Having remained as shadowy presences for over a century, suddenly they have become noticeably human, resurrected in flesh-and-blood likeness.

Techno-enthusiasts and digital technology reviewers have lauded Shirayaev’s smart use of Machine Learning to produce this upscaled version. The software was fed information from visual material from the same period, to make it understand and re-construct the original experience the film would have produced. The digital tools managed to generate new frames not only based on this input but also from different versions of the original film. It is not surprising that the new version has evoked intense reactions which range from fascination, to skepticism, to serious outrage. Commentators have attempted to make sense of Shirayaev’s clip, which claims to be a ‘restored version’ of the ‘original’ film. It has rekindled an old debate among historians and archivists about what we understand by film restoration and what are its ethics and limitations?

The dominant concern among commentators is whether Shirayaev’s version is really a ‘restoration’ as the term is generally understood? Does it do justice to the original Lumiere film, and hence by extension to the history of cinema? Does digital technology and Artificial Intelligence violate the ‘sanctity’ of the original? These questions become crucial as the Lumiere original was one of humankind’s first motion pictures and an inextricable part of its founding myth.


Cinephiles for whom Lumiere films have been part of a collective experience, an essential ingredient in all courses, workshops and cinema’s various anniversary celebrations, have developed a personal relationship with this ‘train’. It’s part of the ritual that is repeated a number of times every year, over decades and has accrued a sanctity associated with a holy relic. But compared to their experience of this famous film, Shirayaev’s version appears hyper-real, having been made with technological tools which belong to the present.

Those who felt that the sanctity of the Lumiere film have been violated were expressing a legitimate concern of film archivists and historians. FIAF – the federation of film archives of the world –
prescribes: “...when restoring material, archives will endeavour only to complete what is incomplete and to remove the accretions of time, wear and misinformation. They will not seek to change or distort the nature of the original material or the intentions of its creators.”

The allegation against Shirayaev, even by some of those who admired his effort and marvelled at the end result, was that the original has been distorted and its sanctity violated. Their logic is that Shirayaev has converted a film heritage into something that it never was. His 4K upscaled version is something new and different and has got little do with the original.

But the issue is far from simple, and any castigation of Shirayaev’s project cannot overlook the issues with which archivists are grappling. If an old film is subjected to processes defined as ‘restoration’, it is assumed that it has to go back to an original state. But the original is a slippery concept when one is engaging with the material medium of 35mm film. Paolo Cherchi-Usai, in 2001, wrote a fascinating book where he provocatively claims that film is a self-destructing art. Celluloid films degrade, and invariably so, because of their chemical nature, and the process is exacerbated by their projection and other factors such as the storage condition, the number of times they have been transported or shipped. The amount of degradation depends on the whether it is a nitrate film or acetate film, black and white or colour.

Hence most films from early cinema are available to us in a range of duplicates produced through myriad photochemical processes, before digital technology made it possible to duplicate them with minimal physical intervention.

True copy

The central question, therefore, is whether, after a century since its first release one could refer to the original copy? How do you restore a film back to the original state when the original does not exist? In a hundred years, all copies would have degraded, making it impossible to determine what the original could have been like. Faced with a situation like this, experts make educated guesses about what the film would have looked like when it was new. Hence, a restoration of this kind is often speculative.

According to restoration experts, a film copy is an artefact – and hence attempts must be made to see whether the signs of degradation and decay can be eliminated, or at least minimised, and the film attains an acceptable material condition. According to well-known archivist and writer Andreas Busche, the aim, however, is not to restore a film back to its original material condition but reconstruct it as a “film text”. After all, a film is meant to be viewed, heard and interpreted, and it is through this process that its purpose is fulfilled. The film is a visual text and a sensory experience – and the process of restoration should be able to replicate its experiential aspect. This experiential reconstruction, as FIAF prescribes, should be in line with the original “intent of the creator”.

The idea of the “intent of the creator” imposes another level of complexity and a conceptual hurdle. How do we determine what this intent was? Do we refer to historians for a better understanding? How do we connect the creator’s intention to the expectation of the contemporary audience?

As cinephiles, we are used to watching degraded copies of films, but this was certainly not the original intention of the creators who had made the film using the available technology of the period. If details were lost from the prints of old films, it was a deviation from and not part of the authorial intention. If the film regains some of the lost visual/aural information through the process of digital intervention, it is a legitimate attempt at renewing the cultural and textual value of the material.

In the process of digital intervention and upscaling, Train Arriving at Station at La Ciotat looks vastly different from the extant 35 mm versions. It does not have the flicker or the scratches or the grains that are commonly seen on the Lumiere prints, and exudes a sensuousness rare in early cinema. But it gives us a strong sense of time and space, which we miss in the celluloid or celluloid-derived video copies cinephiles have seen in the past.

To our contemporary senses, the film might look as if it were shot with a smartphone camera and hence stripped of ‘originality’. But one cannot claim with certainty that if the Lumieres had access to a 4K smartphone camera, they would not have used it to film the train at La Ciotat station aiming to capture the best possible image of that period. Filmmakers, largely, want to represent reality in all its little nuances. There is no reason to think that Louis and Auguste Lumiere would have intended otherwise.