nodevember 2022! procedural art event month! [56k no]

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > nodevember 2022! procedural art event month! [56k no]

«‹›10 »

Bluemillion: Aug 18, 2008; I got your dispensers
right here

fart simpson posted:

agreed op you managed to do some cool stuff this year. even more cool than a spiky ball

That's how far my skills progressed in a year. You better not be slackin'! :c00lbert:

# ? Dec 2, 2022 02:54

Adbot: ADBOT LOVES YOU

# ? Apr 19, 2024 09:42

Trabisnikof: Dec 24, 2005

so this nodevember i have been pleasantly given more desirable activities than blender to engage with mostly so i didnt get that much time to node. but i have been lurking this thread and in the interest of not spamming quotes of all the incredible work posted itt

Komojo - you clearly have a good eye for composition, you have a bunch of excellent execution of ideas within the constraints given. i really want to see what you do next with these skills.
Wheany - cheers for jumping in and being helpful btw
distortion park - your planet and ocean renders are exactly the kind of cool poo poo that it doesnt matter if theyre prompts or not theyre cool and good investigations into particular images and the process of simulating them
fart simpson - thanks for being op, your poo poo is dope af and very pleasing. keep at it, all your poo poo was amazing but the hyperrealism of the MIRV and the meteor shattering are s tier
FalseNegative - excelent work! good vision and ability to select the key features to focus on.
toiletbrush - you really went for it and that freaking rocks. some of the best detail work this nodevember. doing escher in a 3d modeling event is a metal move
Bluemillion - great work!!! like the futuristic garden is like 90s mainstream movie level of cgi already and the spooky depths bring the spooky

# ? Dec 2, 2022 05:23

fart simpson: Jul 2, 2005; DEATH TO AMERICA

digital ruins

# ? Dec 3, 2022 01:54

fart simpson: Jul 2, 2005; DEATH TO AMERICA

blend file here https://mega.nz/file/gJgDlIzB#rJCXzicjP3B94EpCtT6qcLyN8WQSNqTZI6VmEmyELLc

i felt kinda stuck on this one for a bit. i set it up with just a few cylinders and plans and got some general shapes i thought looked alright and then decided to make the cylinders into cell towers because theyre digital. the trees are cool but they have like 10 million verts each so it kills my computer. the trees in the background are very low poly versions because you cant see them. basically just made a curve, and curve to points it into more curves going out radially, then curve to points to get more curves going out radially, then on those i make like 50 green cubes each

# ? Dec 3, 2022 02:05

Jenny Agutter: Mar 18, 2009

looks great, and it�s easy to see how you could parameterize the cell tower generation

# ? Dec 3, 2022 16:46

fart simpson: Jul 2, 2005; DEATH TO AMERICA

Jenny Agutter posted:

looks great, and it�s easy to see how you could parameterize the cell tower generation

i did, but i only made two of them so it sort of doesn�t matter for this work

# ? Dec 3, 2022 16:50

toiletbrush: May 17, 2010

fart simpson posted:

digital ruins

loving hell

are those trees manually nested stuff?

# ? Dec 3, 2022 19:09

Komojo: Jun 30, 2007

Trabisnikof posted:

Komojo - you clearly have a good eye for composition, you have a bunch of excellent execution of ideas within the constraints given. i really want to see what you do next with these skills.

Thanks for the kind words! I challenged myself to do everything in the geometry nodes, but I think I could do some fancy effects if I combined it with modeling by hand.

Here are my final 3 entries:

Ominous Skies

Digital Ruins

Asteroid Village

# ? Dec 4, 2022 03:10

Bluemillion: Aug 18, 2008; I got your dispensers
right here

Komojo posted:

Thanks for the kind words! I challenged myself to do everything in the geometry nodes, but I think I could do some fancy effects if I combined it with modeling by hand.

Here are my final 3 entries:

Ominous Skies

Digital Ruins

Asteroid Village

1. Nessnessnessnessness
2. lol
3. Bitchin' Unreal Tournament map.

# ? Dec 4, 2022 04:42

fart simpson: Jul 2, 2005; DEATH TO AMERICA

toiletbrush posted:

loving hell

are those trees manually nested stuff?

i dont know what you mean by that but, its not manual so i guess no

i started with a curve line

curve to points, instance new curve lines on those points with a random direction but biased in a general upward direction

realize all those, then curve to points and instance new curve lines on those points with random directions biased outward a bit

curve to mesh to give it some thick branches

then curve to points that (right before the curve to mesh) and instances little green cubes with a tiny bit of subsurface scattering on each of those

all of this is scaled down along the spline parameter factor of the original base curve

# ? Dec 4, 2022 05:51

fart simpson: Jul 2, 2005; DEATH TO AMERICA

Komojo posted:

Thanks for the kind words! I challenged myself to do everything in the geometry nodes, but I think I could do some fancy effects if I combined it with modeling by hand.

Here are my final 3 entries:

Ominous Skies

Digital Ruins

Asteroid Village

all of these are great. lol at the twitter one.

great work this month op, congrats on finishing. i also agree that the real power of geometry nodes lies in mixing it with more traditional modeling and assets. nodevember is kind of a degenerate case but its fun to see how far you can push it and it forces you to learn stuff you wouldnt have otherwise

# ? Dec 4, 2022 05:57

fart simpson: Jul 2, 2005; DEATH TO AMERICA

Trabisnikof posted:

so this nodevember i have been pleasantly given more desirable activities than blender to engage with mostly so i didnt get that much time to node. but i have been lurking this thread and in the interest of not spamming quotes of all the incredible work posted itt

Komojo - you clearly have a good eye for composition, you have a bunch of excellent execution of ideas within the constraints given. i really want to see what you do next with these skills.
Wheany - cheers for jumping in and being helpful btw
distortion park - your planet and ocean renders are exactly the kind of cool poo poo that it doesnt matter if theyre prompts or not theyre cool and good investigations into particular images and the process of simulating them
fart simpson - thanks for being op, your poo poo is dope af and very pleasing. keep at it, all your poo poo was amazing but the hyperrealism of the MIRV and the meteor shattering are s tier
FalseNegative - excelent work! good vision and ability to select the key features to focus on.
toiletbrush - you really went for it and that freaking rocks. some of the best detail work this nodevember. doing escher in a 3d modeling event is a metal move
Bluemillion - great work!!! like the futuristic garden is like 90s mainstream movie level of cgi already and the spooky depths bring the spooky

thanks for following along and even doing 1 of them. and for the reviews

# ? Dec 4, 2022 06:00

Komojo: Jun 30, 2007

Now that I've got a bit more time to write, I can share some of the things I learned while making these.

Polar Mystery

I wanted to replicate the look of field lines coming off of a magnet. It would be nice if I knew the actual equations to make it scientifically accurate, but I decided to just approximate it using some basic vector math.

First I spawned some iron filings (simplified UV spheres) on random points in a square.

I can make them all point to the same location by subtracting a constant vector from each instance's position and then using the "Align Euler to Vector" node.

Now I want it to look like it's pointing to 2 points for the different poles, so I'll take each of those difference vectors and divide by the distance squared, then add them together.

Hmm. that doesn't look right. What am I doing wrong?

Ahh, there we go. One of the poles is a - b and the other is b - a, so they point towards one and away from the other.

Impossible Pathway

This one is entirely a shader effect on a flat polygon. My main idea involved this distortion effect with two circular horizons.

It's feeding each pixel's position into a checkerboard texture, but adding a vector to the position to distort the pattern. Once again I am subtracting a constant vector from each point, then scaling that vector and adding it to the input position. I subtract a constant from the length so it equals 0 at the circle's radius, then I divide another number by that value so it goes off to infinity at the horizon.

I can color the circle's interior differently by checking if that values is greater than zero, and if I add some noise textures I can make the pattern more interesting.

Now I wanted to add a Julia fractal between the circles, which I was able to set up just using the node editor. I found pseudocode for a distance estimation function so I could get a continuous color change.

Here's a handy function that takes two vectors and multiplies them as complex numbers.

Here's a single iteration of the Julia fractal with extra variables for the distance estimation. I added an empty object that drives the parameters for the Julia constant, so I can get different Julia fractals by dragging it around with the mouse.

I can't use for loops in Blender shaders, but I can chain these groups together to get the same effect.

Then I can put THOSE into a bigger group and chain those together to make a huge monstrosity of a shader with 128 iterations.

Psychedelic Alley

I used this one to learn how to use the volume-to-mesh node. I wanted to do some sort of blobby helix shape.

Starting with the basics: The "Volume Cube" node evaluates a density at each point in a region, then the "Volume to Mesh" node approximates the surface where that value equals 0.
So, for instance, if you take each point in the volume and compare its distance to a fixed point, you can generate a sphere with a given radius.

The "Volume to Mesh" node has some settings that I'm not familiar with, but basically there are separate settings for the resolution of the volume cube (pre-generating the density values) and the final output mesh.

Let's take that idea but stretch it out by making it 2-dimensional (mapping (x, y, z) to (x, 0, z))

If I take the Y coordinate and plug it into a "Vector Rotate" node I can turn it into a helix. I can control the twisting amount by scaling that Y coordinate.

Now to make it look blobby I can subtract a noise texture or a Voronoi texture, multiplied by some constant.

Finally I added a little bit more variation by scaling the density by a sine wave based on the Y coordinate.

The material is just an emission with the Y coordinate plugged into a color ramp node and a glowing outline from a Fresnel shader. I enabled bloom in the Eevee settings to make it look more glowy, and used an irradiance volume around the helix combined with a few extra background lights.

# ? Dec 4, 2022 06:01

fart simpson: Jul 2, 2005; DEATH TO AMERICA

Komojo posted:

Now that I've got a bit more time to write, I can share some of the things I learned while making these.

Polar Mystery

I wanted to replicate the look of field lines coming off of a magnet. It would be nice if I knew the actual equations to make it scientifically accurate, but I decided to just approximate it using some basic vector math.

First I spawned some iron filings (simplified UV spheres) on random points in a square.

I can make them all point to the same location by subtracting a constant vector from each instance's position and then using the "Align Euler to Vector" node.

Now I want it to look like it's pointing to 2 points for the different poles, so I'll take each of those difference vectors and divide by the distance squared, then add them together.

Hmm. that doesn't look right. What am I doing wrong?

Ahh, there we go. One of the poles is a - b and the other is b - a, so they point towards one and away from the other.

very cool writeup. i didnt know you were doing it like this. i had assumed you placed bezier curves or something and instanced filings along the curves. but this makes a lot more sense now that i look at it

Komojo posted:

Impossible Pathway

This one is entirely a shader effect on a flat polygon. My main idea involved this distortion effect with two circular horizons.

It's feeding each pixel's position into a checkerboard texture, but adding a vector to the position to distort the pattern. Once again I am subtracting a constant vector from each point, then scaling that vector and adding it to the input position. I subtract a constant from the length so it equals 0 at the circle's radius, then I divide another number by that value so it goes off to infinity at the horizon.

I can color the circle's interior differently by checking if that values is greater than zero, and if I add some noise textures I can make the pattern more interesting.

Now I wanted to add a Julia fractal between the circles, which I was able to set up just using the node editor. I found pseudocode for a distance estimation function so I could get a continuous color change.

Here's a handy function that takes two vectors and multiplies them as complex numbers.

Here's a single iteration of the Julia fractal with extra variables for the distance estimation. I added an empty object that drives the parameters for the Julia constant, so I can get different Julia fractals by dragging it around with the mouse.

I can't use for loops in Blender shaders, but I can chain these groups together to get the same effect.

Then I can put THOSE into a bigger group and chain those together to make a huge monstrosity of a shader with 128 iterations.

lol yeah this is the way to do it without loops. i did the same think with i think up to 256 iterations on my julia fractals i was doing earlier this year. one more trick you can do is to also pass in an iterator index integer, with the first iteration its set to 0. you pass this through and increment it by 1 in part of your inner loop. then on the output of the group, you can use a switch node with the boolean value set to check if the index is greater than some number. if it is, then it switches into just passing through all its inputs without doing any more work. this way you can also adjust the number of iterations on the fly and the switch node ensures none of the excess computations are actually being performed so its fast

Komojo posted:

Psychedelic Alley

I used this one to learn how to use the volume-to-mesh node. I wanted to do some sort of blobby helix shape.

Starting with the basics: The "Volume Cube" node evaluates a density at each point in a region, then the "Volume to Mesh" node approximates the surface where that value equals 0.
So, for instance, if you take each point in the volume and compare its distance to a fixed point, you can generate a sphere with a given radius.

The "Volume to Mesh" node has some settings that I'm not familiar with, but basically there are separate settings for the resolution of the volume cube (pre-generating the density values) and the final output mesh.

Let's take that idea but stretch it out by making it 2-dimensional (mapping (x, y, z) to (x, 0, z))

If I take the Y coordinate and plug it into a "Vector Rotate" node I can turn it into a helix. I can control the twisting amount by scaling that Y coordinate.

Now to make it look blobby I can subtract a noise texture or a Voronoi texture, multiplied by some constant.

Finally I added a little bit more variation by scaling the density by a sine wave based on the Y coordinate.

The material is just an emission with the Y coordinate plugged into a color ramp node and a glowing outline from a Fresnel shader. I enabled bloom in the Eevee settings to make it look more glowy, and used an irradiance volume around the helix combined with a few extra background lights.

idk if you already know this, but this volume cube stuff you're doing is essentially the same thing as "signed distance functions" or SDFs. like, your voronoi noise is in effect a field providing a distance to its own surface at every point in space. subtracting one SDF from another is a known way to "round" things in SDF land. for example, you have a box, and if you just subtract a constant value then you get a rounded box. you can read all about this stuff on this guys website if youre interested:
https://iquilezles.org/articles/distfunctions/
heres a visual explanation why subtracting rounds off corners and makes shapes bigger:

in fact, what you plugged in to the density of the volume cube there... you derived an equation for the SDF of an infinitely long cylinder. you could take that exact same equation and use it in whatever software you wanted that works with SDFs. heres another equation to do basically the same thing from that website:

i use this kind of stuff all the time with the new volume cube. i have a node asset set up to convert any mesh into an SDF (based on raycast and geometry proximity) and another to convert any SDF into a mesh (based on volume cube) and some others to do the basic SDF boolean operations. a cool trick thats useful all the time to me is to use voronoi noise distance to edge plugged into an SDF subtraction boolean using the smooth maximum for a smoothness slider. it makes any mesh look like its cracked up and made of cobblestones or something

fart simpson fucked around with this message at 06:25 on Dec 4, 2022

# ? Dec 4, 2022 06:19

toiletbrush: May 17, 2010

im trying to do some of the prompts I missed and coming across a problem ive seen a few times where 'float curve' doesn't seem to always apply. I'm trying to make some barnacles with the fac output of voronoi going into float-curve so I can 'clip' everything past a certain point. if I plug it into the color input of my shader for example it works exactly how I want, but if its plugged into a bump node, it's like the float-curve is ignored.

is this a bug or am I missing something?

# ? Dec 5, 2022 21:04

fart simpson: Jul 2, 2005; DEATH TO AMERICA

toiletbrush posted:

im trying to do some of the prompts I missed and coming across a problem ive seen a few times where 'float curve' doesn't seem to always apply. I'm trying to make some barnacles with the fac output of voronoi going into float-curve so I can 'clip' everything past a certain point. if I plug it into the color input of my shader for example it works exactly how I want, but if its plugged into a bump node, it's like the float-curve is ignored.

is this a bug or am I missing something?

i saw this on older versions of blender. not in 3.3 afaict. although i did figure out a bug where if i click and drag from an output and get the search box and type in "float curve" it was automatically plugging into the fac instead of the value, which made it look like it wasnt working

# ? Dec 6, 2022 01:35

fart simpson: Jul 2, 2005; DEATH TO AMERICA

and i think im done! asteroid village

# ? Dec 6, 2022 01:56

fart simpson: Jul 2, 2005; DEATH TO AMERICA

i wanted to add a few boats anchored there in the foreground but doing those well would have taken a lot of time. so i just didnt do them

i was working from the following reference photos

# ? Dec 6, 2022 01:58

fart simpson: Jul 2, 2005; DEATH TO AMERICA

blend file is here: https://external-content.duckduckgo...a47a&ipo=images

basically just blocked out the scene with big shapes like this

then i worked on the right type of emission shader i wanted for the airburst. i was having a problem with way too many reflections because it was so bright, especially reflections on the water. so everything was blown out. so for the first time ever i used the light path node to reel those reflections in

added a extremely light fog around the camera for some extra glare and smearing around the asteroid airburst because otherwise it looked like this

and my last "trick" was that the island in the background is just a cylinder with a lot of subdivisions. i ran it through 2 layers of noise for large scale and fine details and structure. then made a musgrave texture with tons of detail driving a bump node with absurd distance and strength. and then i just faded in a flat light brown color when it was close to the z=0 plane of the water so it looked kinda like a shore

fart simpson fucked around with this message at 02:08 on Dec 6, 2022

# ? Dec 6, 2022 02:05

echinopsis: Apr 13, 2004; by Fluffdaddy

gently caress mate

the post processing.. I thought it was real for a bti

# ? Dec 6, 2022 09:52

fart simpson: Jul 2, 2005; DEATH TO AMERICA

echinopsis posted:

gently caress mate

the post processing.. I thought it was real for a bti

i did a lot of compositing on this one. i found that applying a blur filter and then sharpening again made it look more "real" somehow? i dont know why. anyway after all that i added in some film grain and did some basic color correction in capture one

# ? Dec 6, 2022 10:15

echinopsis: Apr 13, 2004; by Fluffdaddy

that is honestly the best pic you have made ever imo

# ? Dec 6, 2022 10:42

echinopsis: Apr 13, 2004; by Fluffdaddy

good poo poo mate

# ? Dec 6, 2022 10:42

fart simpson: Jul 2, 2005; DEATH TO AMERICA

thanks echi.

you know whats funny is i just realized that last year i ended nodevember on an asteroid hitting earth too

# ? Dec 6, 2022 10:47

echinopsis: Apr 13, 2004; by Fluffdaddy

ive said it before and by god i'll say it again

you're one of hte best nodesters in the world

# ? Dec 6, 2022 11:30

fart simpson: Jul 2, 2005; DEATH TO AMERICA

echinopsis posted:

ive said it before and by god i'll say it again

you're one of hte best nodesters in the world

im not sure that�s true especially when u consider all the houdini guys. i think anyone could do what im doing with these nodes, it�s just about practice and putting in way too much time lol

# ? Dec 6, 2022 11:56