I woke up yesterday with a simple idea for generating a picture based on an input string. I don’t exactly know where the idea came from, but i think i’ve been influenced by Nick Huggins, whose abstract work i adore, and also in a way by QR codes. I’m not a particularly artistic person, but i figured i could come up with an algorithm and let the computer do the creative bit for me! :)
So i installed the open source tools ImageMagick and RMagick, learned a bit about the RVG library, and set about trying some ideas. I fiddled and tweaked the algorithm until it seemed to consistently output something that was reasonably pleasant. Here is the picture for my name, and for my twitter id.
Having tried random numbers and obvious inputs like my name, i searched for other input sources. Being interested in community generated content, i wrote a script to fetch the current top twitter trends. Here are the results.
As you can see, some come out better than others. I’m adding the input string mostly for debugging purposes so that i can see how the image was seeded. When i get one that i like, i can increase the blur and remove the input string. For example, i really like the images produced by “Kim Hee Chul” and “#bsr_tousounow”, so let’s try with a bit more blur.
Nice, hey?! Not sure they’re ready for a gallery just yet, but certainly an interesting experiment.
Everything in the picture is generated from the input string: the size, colour, number of boxes, box sizes, opacity, border style. It is extremely unlikely that any two input strings would ever generate the same picture. However, the algorithm is not random. Given an input string, you’ll always get the same picture, though i may choose to do some post-processing on it (such as blur, frame, lighten or darken).
In the interests of sharing knowledge here is the main structure of my generator, but the really creative part is in how it comes up with the numbers, which is going to remain a secret, sorry!
I am willing to generate images for anybody who asks nicely! :)