Combining 143 World War II movies into one series
My process for creating the World War II Supercut
Intro
For more than a year I’ve been working on the World War II Supercut, a video project that combines 143 World War II movies into one 12 hour series, with historically significant clips pulled from the movies and ordered chronologically.
Watch the Supercut on Google Drive
Watch the trailer on YouTube
The goal is to use popular movies to show the timeline of the war, providing an educational tool to see what was happening across the world at similar times, and make the entire context much more intuitive to a viewer.
The posts for each video can also be found on the fantastic /fanedits subreddit, where the community has been kind enough to follow the project and support it.
I’ve collected a few programming scripts that were useful and posted them here as well: World War II Supercut - Scripts
Results
I’m writing this post to share some of the process to help others who may want to make a similar historical edit of popular movies, and to warn about some of the pitfalls I ran into. Even just watching 150 movies would take about ten 40 hour work weeks, and that’s without taking notes, historical research, editing, etc. When building a cut like this it’s important to avoid perfectionism, and instead to break the task into multiple small projects, each with their own achievable due date.
I tried to use a variety of films, some higher quality than others, from as many different countries and points of view as possible. Unfortunately, most big budget movies are American or European, so the Western point of view is heavily over represented. The full list of films is below, along with how much footage from each ended up in the final cut:
Western movies are over represented, mirroring the bias in the entertainment industry.
Finding Movies
The first step in the process was creating a spreadsheet of World War II movies (available here, for the curious). I started from memory, but quickly needed other sources. Wikipedia has a large list of films, which made a good starting point. ChatGPT is also a good tool if you’re looking for something more specific:
ChatGPT is great at generating lists of ideas, but often misses obvious possibilities, or makes up movies that don’t actually exist
The criteria for including a movie was difficult, but in general I looked for a wide variety of viewpoints and events. If I had the choice between an older and a newer film, I usually took the newer one, with the goal of keeping the cut engaging and as modern as possible. Sometimes there were 2 films about 1 event that I thought would be interesting to see combined, for example Oppenheimer and Fat Man and Little Boy are both about the development of the atomic bomb, so I included both.
Oppenheimer watches the Trinity test. The same event but shown in two films with two different styles.
To watch all these movies, but to still finish the project in a reasonable amount of time, I watched several of them at 4x speed in VLC, so a two hour movie would take only 30 minutes to finish. While watching I took notes of any interesting scenes or important events depicted. Even with this aggressive approach, I simply didn’t have time to watch many great movies, and they didn’t make the cut, but I would recommend them as followups to watch. Here’s a few examples, and check the spreadsheet for more great options:
Young Churchill (1972)
Army of Shadows (1969)
God on Trial (2008)
Miracle at St. Anna (2008)
Fury (2014)
Land of Mine (2015)
Dieppe (1993)
The Hiding Place (1975)
Dunkirk (1958)
Several important events aren’t shown often in film, if at all. I found very few about Montgomery’s campaign in Africa, or about Hirohito’s early years. Sometimes the only option for a famous event was a low quality, made for TV movie, like the Wright Brothers’ first flight:
Many of these movies were hard to find on DVD, and it was important to me to purchase a copy of every movie used. Many of these films are no longer being sold, and so the only option is used copies, or prints in other regions. Also, the Battle of Neretva and The Battle of Sutjeska are both Yugoslavian films, but their distributor no longer exists, and neither does their country! Most of the surviving copies are low quality, but I was able to find a Spanish language release of one, and salvage enough footage to use.
Schindler’s List was used in 8 out of 12 of my final episodes, the most of any movie, showing what a wide time period the film spans. Most of the films were only used for a few scenes, and only appeared in a single episode. Here’s the breakdown for the most used films:
This chart shows which movies were used in the most episodes of the Supercut. Schindler’s List appeared in 9 out of the 12 episodes.
Video editing
DaVinci Resolve is a widely used, free video editor, and I wanted to give it a try, but unfortunately it kept crashing on my computer. After experimenting with a few other Linux editors, I decided to use the open source 3D modelling tool Blender instead. I’m familiar with Blender from other work, and knew it had a basic video editor. A few tests showed it could handle the requirements, with the added benefit that it’s easily scriptable with Python, so I could automate my workflow and save a lot of time.
Blender’s video editor is simple, but capable, and has built in Python scripting.
Almost immediately I started to run into problems with Blender. First, Blender isn’t able to combine video clips with varying frame rates, and each of my DVD rips had a different frame rate. I eventually decided to convert all my video files to the same frame rate to work around the issue, accepting the potential loss in quality. ChatGPT wrote a script to do this:
#!/bin/bash
# Check if the input directory is provided
if [ -z "$1" ]; then
echo "Usage: $0 /path/to/mp4_directory"
exit 1
fi
input_dir="$1"
output_dir="${input_dir}/converted"
# Create output directory if it doesn't exist
mkdir -p "$output_dir"
# Set the desired frame rate
frame_rate=29.97
# Loop through all MP4 files in the input directory
for input_file in "${input_dir}"/*.mp4; do
# Get the file name without extension
file_name=$(basename "${input_file}" .mp4)
# Set the output file path
output_file="${output_dir}/${file_name}.mp4"
# Convert the video to the desired frame rate
ffmpeg -i "${input_file}" -r "${frame_rate}" -c:v libx264 -crf 23 -c:a copy "${output_file}"
echo "Converted ${input_file} to ${output_file}"
done
echo "All MP4 files in the folder have been converted to 29.97 FPS."The second problem was that Blender has no easy way to export multiple video files from a single project. I wanted to load the movie into Blender, cut it into clips, and export those clips to combine later. I found this script to handle it, but it had a bug. The fixed script is here.
Eventually I started using Blender’s library linking features to import clips. This let me cut the movie into clips in one project, and then import that project into my main file, without having to render individual clips. This saved a lot of time, and improved how quickly I could iterate on which clips to bring in.
Blender supports linked projects, making it easy to reference clips stored in another project
Another issue with this was that Blender mixes absolute and relative path imports. I usually just fixed this manually, but there’s also this option to handle it:
Blender isn’t great with transitions either, and there’s a lot of other improvements it needs. But in the end the scripting capabilities were a life saver. Here’s a few scripts I ended up writing and using (often with AI help):
World War II Supercut - Scripts
To store the 150 video files, and all the render artifacts and work in progress, I used an NFS storage drive that I had previously set up. This worked very well, although some backups would have kept my heart from stopping a few times. Here’s a link for how to setup your own, but you’re probably better off using an external hard drive.
Historical Research
This took up a lot of time. Not only did all the clips need to be in chronological order, but I needed to add title cards to each clip so the viewer would know when the clip took place, what movie it was, what event they were watching, the historical characters shown, and the location. Without these title cards, the story was too hard to follow, since the Supercut jumps between many movies, and usually only includes short clips from each.
A closeup of one of hundreds of title cards. Information included:
Date, Location, Major Characters, Movie, Event
This was a lot of information to research though, with the added burden of needing to fact check that events shown in the movie actually happened.
The facts here are inaccurate. You should never trust a ChatGPT answer, but it can point you to better information, and give you an idea of where to look.
When the movie contained dates or locations I could use those as a starting point, although they were often inaccurate. Any reference to a historical event in the movie was also helpful, since any clip shown after the reference presumably took place later than the event. When a movie didn’t have these clues, I was forced to either dig in deeply to the actual story, if it was a true story, or dig into the historical context if it was a fictional story.
An example of the notes I took on the films Race and Midway (2019). This spreadsheet contains thousands of similar notes
Wikipedia was often a good starting point, although not fully trustworthy. I could also use AI tools to find links to useful sources, although all the information provided by AI tools is suspect and must be verified.
This answer may be wrong, but it gives you a name to lookup, and a starting point.
An example of a true story, and a movie that appeared in 7 episodes, was The Pianist. The film is based on actual memoirs, so I was able to buy the book and use the dates inside as a reference. Some events had a specific date, for example the family listening to the radio broadcast announcing Britain and France joining the war, but others were more vague, for example the restrictions on Jewish families holding currency. The film John Rabe had a similar issue, where some events were historical and could be accurately dated, and others were fiction added for drama.
In John Rabe, Minnie Vautrin was replaced by the fictional Valérie Dupres
An example of a movie that uses a fictional story as a representation of many true events, is Come and See, one of the best depictions of Nazi Germany’s ethnic cleansing in Eastern Europe. Millions of people died in similar conditions to those shown in the film, but very few films have been made covering the events or countries involved. Since the story is fictional, it was difficult to know where to put it in the timeline. But the story is set in Belarus, and researching the history of the resistance in Belarus, and lining that up with when Nazis were burning villages in the East, pointed to early 1943 as the best estimate for when it would have happened. I had to do this estimation with many films, for example the battle at the end of Saving Private Ryan takes place in a fictional town, even though the events are representative of many similar battles that actually did take place.
History vs. Hollywood: Saving Private Ryan
Some movies purported to be true events, but were riddled with inaccuracies. One of the worst offenders was the Mussolini miniseries starring George C. Scott. While I used many clips from the series since it’s one of the few depictions of Mussolini in film, researching it was a nightmare because the series would portray events out of order, and sometimes would make up events for drama, for example it has a scene where Mussolini assaults a British journalist, even though this never happened.
“Fine, my response is they’re all idiots. Fire them and use the savings to fund my machine.”
In the film Turing says this right before going to Churchill for funding, on his own. In real life, Turing’s coworkers cosigned the letter to Churchill and supported him.
Another disappointment was the Imitation Game, a movie about Alan Turing and the British code breakers who cracked Enigma. While the film had beautiful shots of the Bombe machine and a few accurate conversations, it was deeply misleading about the story. First it showed Turing as an antisocial loner who wouldn’t work with a team and cracked the code on his own. In reality the work was a group effort, and Turing collaborated effectively. It also minimized the work of Polish code breakers who had already cracked an earlier version of Enigma, and passed their work on to the British.
History vs. Hollywood: The Imitation Game
The research ran into all sorts of other problems. Many films are deeply political, based on their country of origin, so to avoid this problem I tried to show opposite viewpoints as often as possible, but it’s a fine line between showing a viewpoint because some people believe it, and contributing to propaganda by repeating clear lies.
Posting all the sources used in researching the Supercut is out of scope of this blog post, but to give an idea, here are a few of the many sources I referenced when researching Episode 11, which covers the start of 1945 to August 1945:
Releasing the Videos
I broke the project into 4 stages, and at each stage verified the idea was sound before deciding to continue. The stages were:
Build a prototype of the first 2 episodes and see if the idea is possible, and the result is comprehensible. Show it to a few people to verify. (6 months)
Finish episode 3, with more care. (2 months)
Finish rough cut of all 12 episodes. (5 months)
Remaster all 12 episodes (4 months)
After finishing phase 3, I created a trailer and released it on Twitter and the fanedits subreddit.
And then released a teaser on the D-Day anniversary:
Overall the release went as expected. YouTube removed significant parts of most of the episodes, but I posted videos there anyway to try to get more traffic. Twitter (or X if you prefer) didn’t cut clips out of the videos, but made it difficult to promote the project without spending a lot of time gaming their algorithm. Fanedits was by far the most useful way to promote the project.
Several people have kindly watched the Supercut and shared their experience, and hearing stories like this are what makes the effort worth it:
Because of this cut I finally got a good idea of what the beer hall putch looked like. My elderly parents love this. It is a phenomenal teaching device and very entertaining.
Plus so many of these scenes come from otherwise mediocre TV movies and films that one need not slog through. Even Gandhi which is so good is so very very long it's wonderful to see these important scenes yet they are still moored to context.
Here’s a list of all the Reddit posts:
Other Considerations
I wanted the finished project to be accessible to a wide range of people. The goal was to make the war feel real, and to understand how much of the modern world is actually an outcome of what happened in World War II. I spent a lot of time writing English subtitles manually (auto_subtitle and Subtitle Composer helped, but it still took around 4 hours per episode). For the same reason, I also made the decision to remove as much adult content from the cut as possible, although some violence is unavoidable without minimizing the actual events.
Many people also have a tendency to look down on people in the past, and fail to understand how similar the problems they faced are to the problems we face today. I didn’t want to shy away from the moral complexity or provide easy answers, but I also wanted the truth to shine through when it was clear. For example, the decision to drop the nuclear bomb was based on the best information available at the time, but it’s impossible to deny how horrific the result of that decision was.
The cut is hard to follow in some places. I wanted to create a viewing guide with additional historical context for each episode, but unfortunately ran out of time. I do encourage anyone watching the Supercut to keep Wikipedia open while you watch and look up any events or people who seem interesting. Much of what you’ll find is stranger than fiction, and almost all of it is still relevant in global politics today.
Final Thoughts
Building this project was an absolute blast, and I’m lucky enough to have had some support and viewers. At the end of the day, I built this project because I wanted to watch the end result, and I’m hoping my work inspires others to do the same thing with other historical periods. I’ve learned so much about the war through this project, and about the brave men and women who gave or risked their lives to fight oppression and evil in their day. Hopefully we can learn from their example, and notice the same patterns in the world today, when propaganda, fascism, racism, and hatred rise up, and learn the lessons from history to fight against them.
Many Americans can’t name a concentration camp or ghetto
The full World War II Supercut is available on X or Google Drive at the top of this post. Thanks for reading, I hope you’ll check it out!
Known errors:
* Episode 12: The Saving Private Ryan cemetery scene takes place in Normandy American Cemetery and Memorial in France, not in Arlington National Cemetery.