Generative Art Homework

| | Comments (0) | TrackBacks (0)
So, years ago, while sitting next to Disco Stu at work, I kept day-dreaming about writing software that would create art for me from web input. It wouldn't be very sophisticated. It would, at a minimum:
  • take a list of URLs as input
    • attempt to synthesize the content and context of elements of the URLs into the subject matter of the imagery.
    • spider the URLs when more input or refinement was required
  • set its own style based on the content or a user's style overrides.
  • output a static image of user-selected size
But, after a minimal amount of browsing from the topic GenerativeArt, it's clear that these day-dreams were topical at best. So, this post is an attempt to gather together my thoughts on some of the readings today.
What is dull?
  • Random or selective input is not new or novel.
    • Nor is adding random variation.
    • To select input from the web with intent or at random is stale too.
      • To feed back the output last made back into a pipeline - again, quite boring.
        • To end a piece when the linear renderer/pipeline is done, well, boring.
  • Creating a pipeline for input->generated output really stale.
  • To codify elements of painting style and create unique brushwork, alas, stale.
    • To write filters that re-create or re-mix styles? Ditto.

What is interesting?
  • Self-defacing, morphing paintings made of executable shader code and inputs are pretty interesting.
    • Always different at the time-slice they are viewed in.
    • Always adhering to some hidden ruleset.
    • Static reproductions at some timeslice valuable if they are non-deterministic.
    • Force artist to ask what artistic elements are interesting, and see if the results really are.
    • A toolkit for brainstorming, as posited on TomsHardware, is interesting.
      • For example, the Adobe Photoshop variations interface is interesting in that it jogs the user while editing.
  • The idea of presenting an image as the portrait of the URL input or their topics is slightly interesting still..
    • To create a machine-language program that has artistic intent, adds insight, adds commentary on a subject matter? Not merely abstract or aesthetic.
    • To provide a machine-language program with the ability to describe and defend their works?
      • Or at least log why each decision was made.
  • To find the hidden parts of painting and authorship that seem to set people apart from machines and embed them in machines?
  • To let the machine pick up a topical painting again
    • Intentional series re-creation or themes, as opposed to an endless process that never stops, or a user-defined stopping point.
      • Implied self-awareness of previous topics, weighting them by self-interest level (and other-interest level), and uniqueness or boredom with technique and subject matter.
        • Peer feedback -weighting the pieces for an audience of non-human painters (by their interest levels?)
  • A plugin architecture for handling and extending resource types (2d, 3d, music, text) is a requirement.
  • a filter-stack architecture for selecting or applying transforms to input and setting resultant output file writers, is probably flexible enough to achieve original goal.

So, to pick this up again:
  • would have to have more subject-orientedness and painterlyness than webgobbler.
  • could include fractal content brushes like fleen
  • should save, resume its state.

There's additional reading available, especially about processing (also introduced to me by Mr. Fortress Crack Master):

0 TrackBacks

Listed below are links to blogs that reference this entry: Generative Art Homework.

TrackBack URL for this entry: http://skull.piratehaven.org/~epu/mt/mt-tb.cgi/56

Leave a comment

Recent Tweets