This is a pretty loose interpretation of the prompt, and it's generative in the "how" of what's being drawn, but not of the "what" of what's being drawn. I went to the Art Institute of Chicago's page, and found a couple impressionist CC0 paintings.
I generated a blue noise point grid, and a flow field for each painting. I generated particles at each point of the point grid, and traced their motion along the flow grid flowing in and out of their designated point.
It looks pretty good as an animated GIF - frames overlay previous frames, so you don't see the gaps between particles, except for at the very beginning.
To contrast, the video on YouTube:
You can totally see the gaps between the particles.
Tools Used: Art Institute of Chicago's web page to download the CC0 source images. Pillow, ImageMagick, ffmpegLanguages Used: PythonDevelopment Time: ~4 hoursDrawing Time: ~15 minutes (and then further processing to make the GIF and MPG animations)What's Generative Here: The flow fields and underlying point grids are random, but that's not where the eye lands. You see the familiar paintings, so I'm not 100% satisfied that this is as generative as it could be. I could generate images to use as my source material, which would be more generative. Or, maybe if you think of the individual particles as having a chunk of data each, on each frame, the particles collaborate to generate the image. Feels like a stretch.
No comments:
Post a Comment