Public Domain Cut-up Bot

This week the MET release 375,000 images free to use without restriction. NYPL had a similar release about year ago. With so much cc0 cultural heritage content now available I wanted to try to connect the two releases somehow, especially between these two New York institutions.

A familiar sight to any New Yorker who rides the subway are the billboards found on the walls throughout the stations. As new ads get posted, which are basically large stickers, the previous ones are not removed, the new addition just contributes to a sedimentary layer of advertisements. You’ll often see defaced billboards where people have sliced into this composite layer and removed one or more of the upper layers. The result is an often surprising juxtaposition of the current and past ads. This aesthetic echoes the literary cut-up technique that emerged with the dadaists in the early 20th century and is one I really find striking and wanted to apply to this corpus of images.

The result is a Twitter bot that slices through the layers of public domain images creating new confusing and often interesting combinations. The bot works by pairing up two works starting with the metadata released by NYPL and MET. It looks through all the titles and finds one from each that have the smallest levenshtein distance or the fewest changes needed to make the two titles the same. This is fairly primitive way to pair up works, and often falls flat, but sometimes it does a decent job, for example:

Both works have the word “Canyon” in them, so at least they are about the same type of thing.

The bot then layers the two images together and takes a digital X-ACTO knife to it, cutting an irregular polygon into it with only a pseudorandom number generator guiding its hand.

It also has another trick it sometimes employees. It can find the most common color in the upper layer and slice out those pixels exposing the underdrawing which has a much more “all over” glitchy effect. The result is a new image, an aggregation of two works that often span decades and cut across artistic styles and mediums.

This is the first twitter bot I’ve made, which pales in comparison to some of my ingenious colleagues at NYPL who are prolific bot makers. The one interesting tech feature that people might be interested is that bot is powered by an Amazon AWS Lambda that is triggered on a set internal schedule. This way I don’t have to worry about a server running 24/7. If I had more computer vision chops I would like to explore making the cuts based on superpixles which would likely make it more contextual. But regardless, with so many images now in the public domain made available by NYPL and MET this little bot will have plenty time to refine its craft.