Historical Deepfake

Historical Deepfakes, Part 2

In Part 1 of historical deepfake blog post, we covered photographic manipulation that occurred in the 19th century. In Part 2, we will cover image manipulation from 1900 onwards, and the rise of video manipulation.

Advancements in Media Manipulation Technology

The 20th century saw significant advancements in the history of media manipulation with the development of more sophisticated film and photo editing techniques. Similar to today’s Instagram filters, many early photo retouching techniques were used for vanity’s sake. Since photo editing required significant effort and craftsmanship at that time, it was primarily reserved for actors , political figures and the few portrait photographs one might ever have taken.

The USSR under Joseph Stalin frequently put experts in this craft to work. Stalin frequently removed purged political enemies not only from the physical world, but also from the photographic record. One of the famous historical deepfake examples starts with Stalin and four of his deputies. Subsequent reproductions cut the figures down to three, then to two, then one, and finally with Stalin all alone.

Political image manipulation
Soviet flag raised in Berlin, 2 May 1945 © Photo by Yevgeny Khaldei/Getty Images

Towards the end of World War II, the U.S. captured a famous flag-raising moment in a picture taken at Iwo Jima. It wasn’t a staged photo, though it was the second flag-raising in that spot on that day.

The USSR – determined to create its own Iwo Jima moment during their taking of Berlin – not only staged a shot of a flag-raising that had happened earlier, but had the image retouched in the darkroom to make it look better and to hide a second wristwatch on the soldier’s arm to avoid showing signs of looting. By the late 1970s, their techniques had gotten quite good.

Mikhail Gorbachev had a famous birthmark on his balding head, but it was airbrushed quite convincingly out of official pictures.

The Soviet Union was not the only country using such manipulations. The United States got an early start on manipulated video with the Spanish-American War in 1898. To show battles that had taken place in Cuba and the Philippines, they recreated them in New Jersey, and then mixed in those films with real video of troops on the move and other easy-to-film activities. Given the fake battle footage, film manipulations were technically simpler than photographic deepfaking, as they required only splicing together different scenes. 

Moviola Model D video editing machine
A Moviola video editing machine from 1927 with a microscope attachment from 1940. (HaeB)

As Hollywood movie editing advanced, films of newsworthy events could be adjusted in kind. The Moviola came out in 1924, making it possible to view a film while editing it. In the 1950’s, newer editing tables made the film much easier to see, simplifying what had been a tricky editing process with the Moviola.

Magnetic video tape, the primary medium for capturing television shows, was also launched widely in the 1950s. It offered even easier editing, opening up the possibilities for creative advances like the French New Wave in cinema, which made use of rapid edits and jump cuts despite relatively low budgets.

Computerization of Deepfakes

Computers finally entered the picture with the introduction in 1971 of the CMX 600. This device enabled editors to quickly jump to different frames in the video, and go back and forth in the video easily. Since this was an era in which disk drives were the size of washing machines, however, these tools were hardly ready for widespread use.

With the shrinking of technology and the rise of personal computers in the 1980s, we began to see the democratization of image and video manipulation. In 1987, brothers Thomas and John Knoll created Photoshop, and sold it to Adobe shortly thereafter. By the early 1990s, a new wave of photographic manipulations based on Photoshop and other digital image editing tools was taking hold. Digital alterations were gracing the covers of magazines, being used in television, and helping to artificially portray racial diversity at the University of Wisconsin.

deepfake historical figures
George W. Bush speaking in front of some real US troops and some digitally copied ones at Fort Drum. July 19, 2002 White House Photo by Tina Hager.

In 2004, we saw political photoshopping in the US presidential election. Old photographs of candidate John Kerry and actress Jane Fonda, each speaking at separate anti-war rallies, were merged to make a composite picture that was made to appear as if it was in an old newspaper. Incumbent President George W. Bush, meanwhile, was made to look more presidential during wartime by having more soldiers duplicated behind him. The latter used a Photoshop tool called “clone stamp,” showing the increasing power of the software for fast and easy manipulation.

Digital editing of video grew just as rapidly. The Avid Media Composer for personal computers was introduced in the late 1980s, and has risen to become the dominant professional editing system for video. Adobe created Premiere video editing software for home use in 1991; it was followed by many other software companies in the years since. These tools enable regular users to make so-called cheapfakes, media manipulations without the use of deep learning.

A popular example is the 2020 viral video of Speaker of the US House of Representatives Nancy Pelosi that was slowed down to make it seem as though she is drunk and slurring her speech. More recently, President Joe Biden has been edited in a series of deepfake videos to appear to say things that make him seem confused (e.g., that getting vaccinated would protect people from hurricanes), act strangely (e.g., sitting in an invisible chair), read speaking cues (“end of quote”) mindlessly off a teleprompter, and to fall asleep in public. The common signs of manipulation. These modifications are simple to create with video editing software.

Of course, some cheapfakes do not even require any video editing. Videos of tragic scenes from other countries in 2016 and 2018 were simply mislabeled as being part of the aftermath of the Hamas attack on Israel in 2023.

Molding public emotions through this kind of manipulation is an old tactic, but its use seems to be growing, even as we enter the age of AI.

Some of our related blogs you may find useful: