tweegeemee (트의기미)

A Clojure library designed to breed images via twitter and other social media.

Since it was written in 2015, this library has grown and matured into a larger system of parts. There is now a website http://tweegeemee.com, an underlying database and as of December, 2022 posting to Mastodon has begun. While those other pieces are not open source, this code will remain open source.

Adding Mastodon support is forcing me to rearchitect a few things to use a real database instead of the current gist-based "database" that I have historically used. I'm calling this "Version 2.0" and we wil be deprecating the gist file use going forward. Docs probably won't be updated to keep up with these changes...unless I get requests from interested developers.

UPDATE: as of Version 2.3.0 I no longer include the code to post to Twitter and I no longer read data from gists either. I'm using a real DB for the gist function and just put images into the images directory. Look back in the git revision history to before V2.0 if you want to see that code.

Background

Each tweet has a name based on the timestamp of creation with a suffix at the end, a link to a github gist and a hashtag (currently #ProceduralArt).

The link to a github gist highlights the code used to create the image.

Follow, Favorite and Retweet tweegeemee images to select them for future generations.

Currently posting a set of two new images every hour. The suffix at the end of the image name tells you generation-type: a,b are random; C,D are sexually reproduced and M,N are mutants with only one parent.

For reproduction, the last 200 tweets are considered. The top 5 are allowed to produce offspring.

Oh, what is "트의기미"? That is the Korean (Hangul) characters for twee (트의) gee (기) mee (미). I believe it is just a nonsense word and liked the look of it.

Let's see where this goes...

Background

A while back, I saw Carin Meier's "How I Start" post about running a Clojure-based twitter bot on Heroku. She showed how easy it was to get this working, I was intrigued and decided I'd like to give it a try--but with my own twist.

In the back of my mind, something I've always wanted to do was experiment with image generation via lisp s-expressions as Karl Sims did in his '91 Siggraph paper. But, how to decide what constitutes a "good" image to breed? Well, how about using Twitter retweets and favorites for scoring?

I found the Clisk clojure image library was perfect for recreating Sims' imagery. I also took some inspriration from https://github.com/nodename/clevolution, http://www.thattommyhall.com/2013/08/23/genetic-programming-in-clojure-with-zippers/ and http://blog.raynes.me/blog/2011/12/02/waving-our-tentacles/.

So far, this has been quite fun to work on.

Usage

You should be able to clone/fork this library, follow the "How I Start" post and get this running on Heroku. Note that as of version 1.3, due to Heroku free dynos now requiring 6 hours of sleep in a 24 hour period, I switched to using the Scheduler add-on. See Issue 11 for more information.

If you just want to create & breed your own imagery, see the bottom of core.clj for snippets to generate & show a random image, breed images by hand and mutate images.

Making your own images

Perhaps you'd like to reproduce one of the images posted? Maybe, you want a desktop background or something like that?

First, click that tweet's link to go to the github gist data structure. For example,

 { :name "150613_061332_D.clj" :hash -504503564 :image-hash -1696445894
   :code (clisk.live/vdivide (clisk.live/vfrac (clisk.live/vmin (clisk.live/v+ (clisk.live/alpha clisk.live/grain) [-0.9438 0.4027 2.3753 1.7962]) (clisk.live/sigmoid (clisk.live/vfloor [0.4416 -2.6627 -1.6566])))) (clisk.live/gradient (clisk.live/square (clisk.live/v- [-0.2226 -2.2105 -2.7124 -1.7799] clisk.live/vsnoise))))
 }

Second, copy the info after the :code keyword to the end of the line. e.g. (clisk.live/vdivide ... )

Next, load up a repl and paste the clipboard contents in to define a variable code after a (important!) single quote like so:

=> (def code '(clisk.live/vdivide (clisk.live/vfrac (clisk.live/vmin (clisk.live/v+ (clisk.live/alpha clisk.live/grain) [-0.9438 0.4027 2.3753 1.7962]) (clisk.live/sigmoid (clisk.live/vfloor [0.4416 -2.6627 -1.6566])))) (clisk.live/gradient (clisk.live/square (clisk.live/v- [-0.2226 -2.2105 -2.7124 -1.7799] clisk.live/vsnoise))))
   )
=> (show (eval code) :width 720 :height 720)

To save an image, do

(write-png "nice_one.png" (image (eval code) :width 720 :height 720))

License

Copyright © 2015-2022 Roger Allen

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.