(Higgsas) Geometry Nodes Groups Blender 3.3/3.4/3.5/3.6/4.0/4.1

Hi, thank you.

You can render it by converting curve to mesh and use attributes in shader editor to color the mesh

Example file:

curl_noise_text (1).blend (153.4 KB)

9 Likes

I think blend file was from a 3.6 alpha and there was a some changes in simulation nodes so it won’t recognize sim nodes in current version. The @Benny_G screenshot shows how it should be connected.

I decided to rework the node the setup, now the circle packing converges way faster compare to the old method

circle_packing_new.blend (169.5 KB)

12 Likes

Hi. this is normal?
reverse effect

Hey, It shouldn’t cull geometry behind the camera view, I think problem is that your camera object scale has negative value

1 Like

Hi higgsas! Thank you for your nodegroups! Is it possible to add Instance Index to Geometry Visualizer nodegroup?

1 Like

Hey, I will update it with instances index, also I want to add option to visualize any float/integer values.
But in meantime you can use Instances to Points to view index.

4 Likes

Attempt to create 3d straight skeleton, based on this thread: Skeleton curve implementation based on VDB medial axis

3d_skeleton_test2.blend (1.3 MB)

Updated version using single Sample Nearest Surface node instead of 6, with same results

3d_skeleton_2.0.blend (1.3 MB)

34 Likes

Theres something quite soothing about watching that form, like running a hot bubbly bath…backwards!

5 Likes

hey higgsas can you make an custom unsubdivide node (like the modifier version but better handling topology, because the modifier is buggy in most of cases), thanks in advance if this is doable…

1 Like

Hey, I don’t think is possible do it with geo nodes atm, don’t really have any ideas on how to do it

1 Like

Ok no problem, is quite strange they haven’t added this yet hmmm…

Pretty cool! if this possible to autounwrap mesh along this skeleton lines as a post-step (U=along, V=around)? would have many practical implications…

5 Likes

Thanks, it is possible but there will be ugly/incorrect seams, I don’t really know how to fix it

uv_skel.blend (1.4 MB)

9 Likes

the uv need to be calculated on the GPU I think, to remove the seam all except 1 pixel width,

using something like signed distance field

edit - are you passing in closest point on curve / normal per vertex and then using the shader to do the distance around?

something like this but per fragment of a material

3 Likes

I use closest point and do the distance around all calculated inside geo nodes

2 Likes

Wow, thanks again!!! Useful already. Seams are expected, this is still much better starting point than other defaults

3 Likes

uv_skel_shader.blend (1.3 MB)
here is getting the position per pixel instead of per vertex,

I think it could still be better if we could do a shader that got the closest point on the curve per pixel, and if we have 2 equidistant points we end up with a artifact only 1 pixel in width

3 Likes

sure you could do this.

I’ve actually wrote it up in regards to hair curves but you should be able to use it for this case.

https://blender.stackexchange.com/a/293219/165102

if you have any trouble implementing it in this case let me know.

I am already doing all of this - what I was talking about is somehow using shaderSDF math to get the closest point on the curve per pixel in the shader, (so curve_pos, curve_tan, and curve_factor extracted in the shader without using geometry nodes except maybe to pass in the closest curve section )

you could probably get it good enough within geometry node with something like sample nearest surface or something with interpolation and using sample curve.

though, if you want to go that root [in the shader] you could do linear interpolation between the two closest points in the shader with some vectors math (the calculation of the two closest points [positions] would from geometry node), or do a cubic (bezier) interpolation for bezier curves. (for the tangent you would have to do the calculation twice).

hmm but now that im thinking about it im not sure how to get the cubic factor of a point (or pixel) since it’s not linear hmmm.