Geometry nodes - hopes and disappointments

I had great hopes for geometry nodes. I imagined they might be perfect for modelling regular systems in which there is a strong underlying geometry, modified by secondary variations in scale, position, orientation - things like flowers, leaf patterns, scales, feathers etc. In the picture below, geometry node would be great for modelling the dust and dead flies but when it comes to the main elements (flowers and leaves) right now it looks like I will have to stick with particle systems (which are tricky).

If I am missing something about the capabilities of geometry nodes and someone can put me on the right track I would be grateful. Alternatively I can see I might have to wait and see how things evolve.

Geometry nodes isn’t strong as a modeling tool in 2.92 as much as a scattering/instancing-aid.

Your example scene should be achievable with 2.92 geometry nodes, but would require a lot of pre-modeling of assets. Even so, people are doing incredible things with current limitations. You should be able to do the flower-instancing in your example and even add variation to the “position” attribute of the flower-meshes, but you’d struggle to “generate” the flower.

In the current 2.93 alpha build things are looking a lot more promising: It allows you the sorely needed ability to “chain” geometry node modifiers… this means you can start with a base-geometry of a line, array that, put that through a twist, do some point instancing of some plane-object using geometry nodes, then solidify that, then array that, then do another geometry nodes where you do some weird-time-based animation… etc… Just that addition will be game-changing, I hope.

Cheers

Many thanks for your response. I am sort of reassured to hear that geometry nodes is still somewhat ‘specialised’ right now and that it wasn’t just down to my lack of understanding.

I have been playing with 2.93 but am really struggling to work everything out. It seems right now:

  1. that things are on the move and
  2. that there is very little useful documentation.

But you are right - modelling the flower itself is what I want to do. So far I have been doing it with a combination of cloth modelling + wind (to generate petal variations) and particle systems for instantiating the petals with a degree of randomness.

What I feel I need to be able to do this more effectively is a version of geometry nodes that is much more versatile with respect to local co-ordinate systems including polar co-ordinates.

The key milestone will be when I can model something like a convincing pine cone.

All the best …

Not sure you saw this, for inspiration:

This was just a super quick test but you can already do pinecones and anything like this where it’s just a distribution of something like a pinecone scale or a flower petal etc

2 Likes

Hi,
The pinecone and your Summer Evening composition that Boder drew my attention to are really encouraging - though it seems the key component is your ETK toolkit, which I hadn’t heard of before.
I have taken a look and plan to download a copy in the next day or two.
I am not a Blender developer myself, but I have many years experience of writing complex geometrical code in C and C++. This makes me want to understand the basic components, co-ordinate systems behind all this stuff but I am finding hard information very difficult to come by. Looks like I will just have to work out what is going on by experiment.

It’s a similar situation with sverchok which is another tool/system that gets talked about a lot but which I don’t really understand.

Thanks for your help though - really useful.

2 Likes

Currently the devs are really focused on the low level, core tools. Things like ensuring we have a working spreadsheet and distribution options and functioning booleans and node sockets etc. Still very early in development.

My toolkit is just aiming to create some high level functionality as single nodes - it is all just regular maths inside though so you’d probably be pretty familiar with it from your coding. Phyllotaxis for example is just a line primitive that’s bent into a specific spiral with sin and cos. Having presets just helps hit the ground running really.

I’ve always found Sverchok to be a little short on English resources (understandably since it’s a Russian addon) but their documentation is generally very thorough and the developers of that are great for fielding questions.

Shout if you have any geo nodes questions though, I’ve been sticking pretty close to the developments.

2 Likes

to omnivorist (sorry for “off-topic”):
I have WebGL procedurally generated gardens at https://www.ibiblio.org/e-notes/Splines/garden/lathe/lathe4.htm and WebXR flowers at https://www.ibiblio.org/e-notes/webxr/garden/garden.htm (cartoon-like yet). How can I improve these procedural models (any suggestions)?