2023-08-21 16:26:40
In 1994 there was a reunion with the classic of all black and white classics on the cinema screen: But this version of Michael Curtiz’ “Casablanca” was in colour, elaborately and post-colored frame by frame to breathe new life into the film. But the technology wasn’t that mature back then.
At the time, Ingrid Bergman had a fairly orange complexion, the walls in Rick’s Café flickered alternately in shades of yellow and blue, the image still looking black and white in the shadows. After all, Humphrey Bogart had lost his ashen appearance; the star known as a chain smoker also had little color on his face in real life. In the colored version, on the other hand, it was salmon pink.
The urge to experience the past in color kept blossoming: In 1998, for example, Gus van Sant dared to make a color film adaptation of the black-and-white classic: Hitchcock’s “Psycho” was just as difficult to imagine in color as Hitchcock once was could imagine putting music to the famous shower scene at all. Film history thanks him to this day for having been convinced of this, but did it really need a color version of “Psycho”?
also read
Sometimes the paint would even have gotten in the way. Billy Wilder could have shot “Some Like It Hot” in color in 1959, but the technology wouldn’t have failed. Alone: Wilder wanted the film in black and white because that corresponded better to the character of the story.
But Wilder noticed another effect of Some Like It Hot, as he noted years later in an interview book with Cameron Crowe: “The interesting thing was, most people who came out of Some Like It Hot hadn’t realized that the film was in black and white. They thought they saw a color film,” Wilder said.
In film art, the question of subsequent coloring is probably of secondary importance. In the documentary field, however, there is definitely a need for color images. The demand to colorize footage for television history documentaries is enormous.
Everything had to be colored by hand
Series like “The Third Reich in Color” are popular, which want to make the horror of war, but also the staging of the regime particularly vivid – by showing old Agfacolor color film recordings, which were already available in the hobby at least from the mid-1930s -Schmalfilm has given. But color film was expensive and therefore rare.
The reason why the subsequent coloring of black-and-white films has not really caught on so far is due to the time-consuming coloring of the individual film frames: Both by hand in the silent film era and on the computer using Photoshop – it takes a lot of time and a lot of expensive labor. Thanks to artificial intelligence (AI), that will now change.
“Psycho” in color – in the new version by Gus van Sant from 1998 that no one wanted to see
Quelle: picture alliance/
At the Technical University in Graz, researchers and computer scientists at the Institute for Machine Vision and Representation have implemented the project “RE:Color: Efficient coloring of films in cinema quality based on novel methods of machine learning”. Leading this project is Professor Thomas Pock, who, together with the Graz-based company HS-Art, which specializes in the restoration of historical films, has developed an integrated software application that combines interactive and automated coloring techniques with deep learning technologies. The result is an algorithm for a mostly automatic, yet fully user-controlled inking process.
“It is very important for me to emphasize that with this software, the human always retains complete control over the image,” says Pock, who is aware of the responsibility that the development entails. The AI software is primarily aimed at the many film archives in the world, which can not only restore their stock, but also provide the corresponding color image of the past.
Those colors of yesteryear are a tricky task for today’s researchers because it takes a lot of knowledge to be accurate. “That’s why you have to feed the AI with information that is compiled on the basis of experts or objects of the time. It always takes someone who knows from historical records what the clothes and facades looked like back then.
also read
“Was the soldier’s uniform green or blue? No algorithm can decide that. But he can learn from it,” says Pock. The algorithm must therefore be fed with a sufficiently large collection of training patterns in order to then automatically take over the coloring of historical films. “It’s about coloring the films as efficiently as possible. This can look like the human specifying the coloring for a film image and the software then takes on the coloring of further images,” explains Pock.
However, Pock qualifies, you have to intervene again at the beginning of each new scene. “At least for now,” says the researcher, who admits that AI is still in its infancy. “It is currently not possible to let the AI colorize a film independently,” says Pock, who still sees the system as ready for the market.
The core algorithm was published at an international conference, the source code is freely accessible. For efficient use, however, software based on this is required, which was developed by the project partner HS-Art: The “Diamant-Film Colorizer” was already used in the ZDFzeit documentary series “Hitler’s Power” for true-to-the-original coloring of historical recordings.
Scene from a post-colored World War I newsreel in Peter Jackson’s They Shall Not Grow Old (2018)
Quelle: picture alliance/ Everett Collection
But what does “true to the original” actually mean? Who says what colors and lighting moods were actually seen when historical black-and-white films were shot? How do we actually imagine the past? The director Michael Haneke has given a possible answer with his film “The White Ribbon” (2010).
The past is black and white in our heads because we only know of it in black and white photos and films. Following this thesis, Haneke created “The White Ribbon”, a drama strictly geared to the aesthetics of photography of the early 20th century, which formed a strikingly effective version of realism precisely from the artificiality portrayed in the photographs. But only one version.
If the AI now colors historical recordings that are created on the basis of today’s information and research, is that realism, or is it not much more – the opposite, despicable falsification of history?
Much research remains necessary
“Often you start by visiting and photographing the original locations of the recordings,” says Pock. “Then you get, so to speak, first-hand the color and light mood that used to prevail in this place”. This takes effort, but it’s not impossible.
But what if the original location no longer exists? So roughly most of Berlin in 1942? “Then you have to use archives, photos, objects or records to get information about what something might have looked like,” says Pock. It does need a lot of research, but so does the dinosaurs, which weren’t photographed, so their bones had to be sufficient to reconstruct them into primeval monsters.
Does the coloring ultimately result in a false picture of reality of the past? A wrong view of history? “Of course, the AI does not create reality, but calculates a fiction,” says Pock. He also means that taking pictures at face value is becoming increasingly difficult, especially in times of AI applications.
also read
Filmsammler Serge Bromberg
A colored film that shows Hitler in uniform at the Berghof brings the Nazi era into the present, makes more comprehensible what we had forgotten and repressed for decades as an incomprehensible, black-and-white past, and lets us realize that people in earlier times experienced their environment in colour. But this does not make these images truer or more real. They are and will remain fake and not fact.
Following this approach, however, one would have to disqualify the medium of film in its entirety as a basis for historical images. Because the medium itself is an assertion, the purest illusion: there are 24 individual images per second that use the inertia of the eye to be perceived as movement.
Based on that, you really can’t believe any moving picture. Which is why the AI cannot produce realism either. She can only perfect the illusion invented by the Lumière brothers almost 130 years ago.
#Pictures #fakes #facts