from.. thecolorsofmotion.com/faq
How does this all work?
We take a movie, then we extract the frames at at a regular interval based on the length of the movie (For a 1h30 movie, it's around once every 10 seconds). Next, we calculate the average color of each of those frames. Finally we stack those colors as thin bars from top to bottom in chronological order. The result is a color timeline of the movie.
But like, how does it really work?
Ah! The nitty gritty. The entire process of extracting the movie data and generating the prints is automated using a bunch of node script. To extract the images, we use a node wrapper around ffmpeg, then a series of custom scripts extract the average colors. The frames are uploaded to amazon S3 and the color data stored in mongoDB. Relevant movie data is pulled for each movie using an unofficial IMDB api. The generation of the print files is also automated, using a node canvas wrapper.
if each frame was a pixel then a 360 x 360 picture would do 129,600 frames
(, Sun 2 Jan 2022, 21:08, Reply)