Geometry Nodes From Scratch

[Tyger2]

I wanted to learn geometry nodes, so I started a course called “Geometry Nodes From Scratch” from Blender Studio. Here’s how it’s going:

.

Project 1: Rock generator

Fairly simple. Done by rotating ico spheres around and using boolean mods. These four different results are created by tweaking these few settings:

.

Project 2: Rock Scattering

This project takes the rocks created in the first project and makes a system to scatter them around a plane. The deformation of the plane was also done with geometry nodes.

The density of the rocks is roughly tied to a vertex group which can be weight painted. Don’t know why I did a ‘G’.

.

Project 3: Tree Generator

I added the trunk myself using the concepts from the videos.

.

Conclusion

Its interesting stuff and I’ve enjoyed it so far. The theory videos took me a couple viewings to fully understand, but the projects exemplify the concepts well. It is a paid series, since you have to be subscribed to Blender Studio, but I figured I might as well throw them a few bucks since Blender is free and they’re the people developing it. There’s one more section I haven’t done yet, so I’ll post my results after I get to that one.

[FedPete]

@Tyger2

Don’t know why I did a ‘G’.

I saw a curved arrow …

[FedPete]

@Tyger2

The theory videos took me a couple viewings to fully understand, but the projects exemplify the concepts well.

I had the same feelings, when I tried to follow the Material shaders course, understanding the 3D aspects. But I stopped, not because of the technical knowlegde, but my hardware couldn’t cope with the huge data calculations involved, with the examples used.

[NP5]

I have followed some tutorials on these geo nodes. The scattering is the most useful. But ultimately not that different from hair particles just a new way of doing it that by node inputs is a bit easier to control.
The trouble is I would have to start over if wanting to use them again as it is not used regularly so not remotely a habit.

I still consider most of the ‘making’ stuff with them to be pretty pointless and much harder work than just modeling. OK accepted that done by nodes things are still adjustable. But frankly not being some big game studio there are few if any times I want a dozen variations of the same basic thing. Even if I did I would rather just make a different version!

Basically, it is visual programming not modeling. Certainly clever and has uses, just not the be all and end all obsession and concentration on them Blender is making at the moment. imo.

My scattering project. Plant models made normally then used to scatter.

[Tyger2]

You make a fair point. I think using them interspersed with traditional modeling is probably the sweet spot. Definitely a powerful tool for procedural work though.

[zeRgenTa]

Looking good!

@NP5

The scattering is the most useful.

It’s useful but, I’m not sure if I agree, with it’s the most useful. In my experience, working with curves is where geometry nodes really shines.

[NP5]

Don’t think I have come across, well even looked for what the curves do. Oh maybe I did a rope thing that probably counts. Yes it works but so does the same sort of thing with the real curves rather than nodes. So much seems to depend on having any point to you in being able to make variations on the theme.
In fact I suspect and Blender weekly thing by Pablo has talked in terms of expecting geo nodes to be used to make ‘modifiers’ that you just download to use. Perhaps loaded in the asset browser thing. As they are a form of programming not really modeling.

I think part of the problem is the continual adding to geo nodes, perhaps when it settles down as basically ‘done’. Then some ways of doing some things might become more ordinary.

Still, probably my age, and making background, a turn towards programmers taking makers functions in life, as ai encroaches too.
May have to revert to whittling a stick with a flint tool.

[zeRgenTa]

While geo-nodes doesn’t usually do things that you can’t by other means, sometimes you can use it places that regular geometry can’t be.
For example, you can add binding to clothing while keeping the simpler geometry for the cloth simulation.

A few things that you can’t use curves for without geo-nodes, as far as I know.

• Making generators for repetitive things like fences (Preferably using pre-made models).
• Deforming a landscape, for a river, road, etc.
• Array an object along a curve without deformation.
• Converting a curve to a volume for quick smoke that looks decent.
• Rigging a curve for animation, or just easier control. (I think you could use drivers for this, but it would take longer.)

I agree that using geo-nodes as modifiers is the way it’s probably going to go. I don’t think it’s worth making something with geo-nodes that you won’t be able to reuse in a verity of situations. If they start going in the direction of Houdini, I think that I’ll have to re-evaluate that opinion.

[Tyger2]

I’m using my entry for this week’s collab as a chance to practice these concepts. Here’s my idea:

And here’s how it’s going so far:

The basic concept is a ground mesh (modeled the old fashioned way) with hand-drawn curves that are
manipulated by geometry nodes to make them more interesting and organic feeling, and are then converted to a mesh and used as a boolean to cut holes in the ground mesh.

I also plan to add other details later, like grass, sky, and of course ants. Lots of ants. Doing different things and with stuff in their nest, like eggs, food, etc.

I think both of you guys are right, @NP5 and @zeRgenTa . Some stuff just doesn’t make sense to do using nodes. I tried to make the ground mesh using nodes at first, because I thought that would help with my workflow when I went to use the boolean later on. But it was unnecessarily complicated when I could just model it in a few minutes.

But when it comes to the curves, I don’t know how I would get them to look like that without the nodes. They just give you a different level of control that is not present without them (or at least that I know how to access).

[Tyger2]

Here’s what the tunnels look like before they become a boolean object.

Looks like some kind of alien digestive system or something.

[Tyger2]

You can see the end result at the Collab page for this week.

If I were to continue it, I’d like to add more details to the tunnels. More food chunks or other ant activity. What the hell do they do in there? I’m also not totally satisfied with the lighting and textures.

It’s far from complete, but it’s as much time as I want to spend on it, and I think I accomplished my objective of using the skills I learned from the geometry nodes course on a real project. And with that I’m considering this venture finished for now.

Thanks for stopping by and for the good discussion!

[FedPete]

The idea is good, Maybe place this in a toy like construction as in a “Ant farm”.

Also a different view, to see the ants close by … which are the most important “Non-human” aspects of the scene.

Have fun.