Jesse SpielmanSome tinkering inspired by @bitartbot ; first art thing in a long while. Feels nice to be a bit creative again.
#generative #generativeart #proceduralart #noai #processing #ps5
@bitartbot I'm going to bed now but one more cool one (zoom in!).
red: f(x, y) = (x & (x + (((((4 | 14) * (y | y)) * x) + (y & x)) + (x & (x | (x - x))))))
green: f(x, y) = ((x & (x % x)) | (y * ((y * y) * x)))
blue: f(x, y) = (((-11 | (y + ((x & (y & 1)) * x))) | (((-5 + -1) | ((y - ((x % y) % x)) | (x | x))) % x)) + (x - (y | x)))
red: f(x, y) = ((y - (y - (2 * (y - y)))) & (y | (x | y)))
green: f(x, y) = ((x - (y * (y - y))) & y)
blue: f(x, y) = (((x | x) & (x * ((y % (y + x)) | x))) + 3)
green: f(x, y) = ((x - (y * (y - y))) & y)
blue: f(x, y) = (((x | x) & (x * ((y % (y + x)) | x))) + 3)
Stefan Ihringer@heavyimage There’s y-y in there 🤔
@compfu well observed! The equations are generated randomly so this can happen. There’s already a TODO in the code to do an optimization pass to remove or simplify nodes or groups of nodes :-)
@compfu eg in addition to algebraic identities (such as x-x=0), a set of nodes that are all constants could be collapsed into a single value.
@heavyimage the result looked very pretty though.
@compfu yeah -- having parts of the tree resolve to constant values doesn't break the technique -- it's just inefficient.
@compfu Got a simple version of this working -- so satisfying:
f(x, y) = ((((floor(tan(tan(-1))) % sin(cos(2))) % tan((x ** ~(1)))) | (((y | sin(-2)) // (x + ((1 | y) / ((floor(floor(-1)) // floor(-1)) * floor(max(((min(-(x), 0) ** floor(1)) ** y), (y ** y))))))) ** ~(-1))) % -2) (length: 217!)
SIMPLIFIES to:
f(x, y) = (((-0.005818788 % tan((x ** -2.0))) | (((y | -0.034899496) // (x + ((1 | y) / (NaN * floor(max(((min(-(x), 0) ** 1.0) ** y), (y ** y))))))) ** 0.0)) % -2) (length: 164!)
@compfu each node in the equation tree gets traversed and if the children are all constant (eg, don't contain x or y) they can be replaced with a single value node containing the value they (always) resolve to.
@compfu just noticed that NaN ah well :-)
@heavyimage sneaky. What does the result look like? Or does it fail to render?
@compfu In the event of an arithmetic error, we just return 0. So likely that entire equation will just return 0 for all points. Since I create 3 functions (one for each of RGB) that'll probably result in an image without one of those channels.
2rad0Here's one I stumbled across, not quite in the same stylish league as your work, but maybe someone with better skills than I can spice it up.
@2rad0 @bitartbot Yeah this is cool! For sure there are more optimal ways to do what I'm doing now that I see the shape of it a bit better. That's what I love about environments like processing -- fast to just kinda bash something out.
Essentially I'm generating a random function to run per-pixel in the output grid. This is both embarrassingly parallel and essentially how GPUs work. So def some options for speeding this code up! I would love to see it animated...
@heavyimage @bitartbot
Animations are nice to add a little extra flavor, but the patterns alone are the real prize. I wish I could intuit how the math works out, I feel a bit like a monkey mashing on a keyboard until something miraculous happens. I think it's similar to a signed distance function but rather it's a square distance. Your star/spirals pattern is close to what I'm doing, i've seen it before, maybe one or two operations different but I can't recall how to recreate it. How rude of me not to post the animation:
Animations are nice to add a little extra flavor, but the patterns alone are the real prize. I wish I could intuit how the math works out, I feel a bit like a monkey mashing on a keyboard until something miraculous happens. I think it's similar to a signed distance function but rather it's a square distance. Your star/spirals pattern is close to what I'm doing, i've seen it before, maybe one or two operations different but I can't recall how to recreate it. How rude of me not to post the animation:
@2rad0 @bitartbot Oh shoot, @2rad0 -- I didn't realize that was your code! When you said "one I stumbled upon" I thought you meant the code project and not the image! Now I understand!
Totally agreed re-animation -- been thinking about that all day! Should be easy to add some additional context to each node for time and record some animations!
And yeah -- I've seen some of the structures in my images in other contexts but I'm not totally sure how they're appearing. That's kind the fun of generating the formulas randomly!
@2rad0 @bitartbot I like your animation a lot! def has some of the structure of SDFs, agreed. Where is the color coming from?
Here are some early animation tests using my approach with a time dimension.
#animation #generative #generativeart #proceduralart #noai #processing #ps5
Thanks for sharing, I love this type of art and now have some new hashtags to follow. I setup a video wall system a long time ago and this is the type of visual content they go crazy over. I forget the name of the program we had, but most of their presets were garbage compared to this.
The coloring (I think) is based on distance from top-left and r/g/b are slightly offset a bit further. The patterns that emerge I'm not so sure about... Here's what the coloring looks in zoomed closer to t=0.
Bonus image included because those stars just now decided to show up in a completely unrelated render engine a couple hours ago when I randomly toggled off screen clearing while rotating the camera! My best guess is an artifact from how my screen rays are being cast against the rotating floor texture, but I've seen it so many other places the mystery remains.
@2rad0 @bitartbot Very cool. Re the stars I think they might be related to moire patterns?
Moiré pattern has to be it, case closed finally!!! after years of feeling crazy seeing it everywhere. That's related to twisted graphene too which is an area of research in super conductors, amazing!