Geometry Nodes

Why not incorporated Socar for Geometry Nodes and grow from here ?
I mean anyone who use Socar hope Geometry Nodes is just beginning and will grow.

Otherwise geometry nodes will stay some very limited subset of features, and people already familiar with Houdini won’t care a lot about Geometry Nodes and it’s very limited features.

I do not know why integrating Sorcar (which would likely require the shoehorning of a different design into the current one) would be a requirement.

For now though, the devs. are back to working on core functionality (which includes recent commits like support for multiple input nodes on the top level (the type you build your modifier with).

1 Like

Did anyone successully modify the uvs inside a geonodes network? I am not able to modify an existing uv map. I see I am affecting values in the spreadsheet but those values are not used in the actual shader.

It seems like Eevee can’t get the geonodes modified uv coords but modifying the uvs inside the geonodes work for Cycles. I originally was checking this with Eevee. This works only in 2.93 nightly, not in the current 2.92 release.

WIP curve support (being worked on by Hans).;c95b2cefab25252469e50404f04293db681df825

So far, it looks like the initial patch will have full support for attributes as well as a new ‘curve to mesh’ node. I wonder if this also means scattering along curves as well?


It looks like Hans created the branch just, 3 days ago.

Task Design details are on wiki.

If there is a functional ‘curve to mesh’ node, that should be possible to obtain a scattering on vertices from conversion.
There is also a resolution attribute. So, that should be possible to obtain different scatterings modifying this attribute.
But that means that are evenly distributed instances. To obtain a random one, we probably have to mess up the resulting mesh.
There is probably an inability to obtain a specific position like with a follow path constraint with absence of Curve Trim node.


Quick question: is there a way to apply a data transfer or any other method to transfer normals from one object to another on geometry node scattered objects?

I ask, because when I’m making my trees, it’s a bit of a chore having to make my instances real, combining all my separate parts into one object, then doing a data transfer, only have to undo everything if I feel the need to readjust something.


I think at the moment it’s not possible, but it’s planned.

1 Like

Hey guys. Working in the 2.93 Alpha, I’m testing geo nodes on a production where I need a forest on a huge photogrammetry landscape. I’ve been trying to find a way to modify my painted weights with a procedural noise, so that I paint in the edges of the forest, but then add some variation with the noise in the denser areas. My idea was to use an Attribute Sample Texture node, but then somehow multiplying the weights with the noise using Attribute Mix. I did manage to get something to happen, but nothing predictable. Any ideas?

Hey guys! Here’s my new challenge. I want to spread objects on a curves surface and keep them aligned with the original geometry. By default, they will aligned with the normals and that’s fine but as you can see in the attached image, the cubes are randomly rotated in their Z axis. Any ideas?

EDIT: see it like a tire tread.

1 Like

Something like this?

Using an attribute randomize node with a rotation attribute set to 0 on mix and max will give all the cube instances on the curve the same rotation as the source.

The particles normals are aligned fine, You just need to align their tangent, aka their local Y axis
you can do that with the align to vector node

the default tangent alignment is a bit weird indeed

Nop, not what I’m looking for. See it as tire treads. The cubes should follow the object’s normals but without a weird rotation, they should stay aligned with the “tire”

1 Like

I tried the aligned vector node but couldn’t get it to do what I’m looking for.

Yeah there’s not going to be any consistent alignment possible without UVs I think

well for me i always try to align them with Local Y axis and it looks fine enough
but the default tangent algorithm is a bit weird, i don’t really get how he is working, seem like it’s looking for the highest neighboring vertex

I don’t know how this relates to Geometry Nodes, but: Curves have settings for aligment under ‘Twist Method’, which does imply it has an inate direction/normals that can be used, right?

The mesh primitive nodes are far faster now (the patch is now in master)

There is an interesting bit about how Bmesh implements them, the primitives in Bmesh are actually macros that use modeling operations in the background. No wonder editmode is slow, a lot of tools may need to be rewritten if the easiest way out was preferred to a harder to code, but far more efficient method.


That’s not the correct interpretation of the bmesh comment. It was directed just at primitive creation where bmesh uses rotate/extrusion operations behind the scenes to place the verts in the right place like say for a cylinder or uv sphere (e.g. first place the circle verts then use the extrude code to complete the cylinder). This new code just rotates/translates directly as it builds the primitive.


So, in the last few days, I played around in sorcar because I heard good things about it from other sources, and I wanted to try it out. For one, sorcar is not a procedural generation tool, it is an editing tool. With sorcar, your node tree ends up in a somewhat linear fashion, which is not a bad thing, but it makes it more limited. All of your normal editing tools are there, and you use them in the way you would usually edit a mesh: making selections, extruding, insetting, transforming, etc etc. To me, it looks more like editing with unlimited undo steps, which I really like, but it’s not going to replace geometry nodes. Geometry nodes is attribute based, as opposed to selection based like sorcar, and it is very useful for actual generation, or close at the moment. So in my opinion, comparing geometry nodes and sorcar is like comparing apples and oranges. They are completely different and they are used for completely different things. I agree that it is fairly limited at the moment, but everything takes time to evolve, and geometry nodes just started.

Earlier, @Ace_Dragon and I were talking about exposing and hiding node values, only showing the most basic values by default, until you click the expose icon in the properties window. Sorcar has that exact feature, and it works really well. I think that is a pretty good example.

1 Like

I found it more easy to make procedural builindings in Socar, while it’s a bit work around in actual geometry nodes and lot less capable.
Geometry Nodes just starting, too many missing features everyone already pointed, many Houdini Unity examples just not possible.

I also hope it will evolve and get more features complete, allowing to make all those small Houdini tools examples (mainly related to game assets).