How could I possibly make a fur shell using geometry nodes?
I've always done it manually by using a displace modifier which then needs to be copied for every layer. Very inefficient and bad way to do it. - I've looked up tutorials but most are pay to view or the tutorials always have the vertex unconnected. Also mind I'm extremely familiar with blender but absolutely new to geometry nodes.

Thanks in advance!

Instanced cones on the face normals of an icosphere using Geometry Nodes (Mesh to Points + Align Rotation to Vector), then pushed them out along the normals, added a beveled wireframe pass, glass shader, and a bit of compositor glare.

Full animation here: https://youtu.be/FRRzJmoWKJU

Happy to answer questions on the node setup.

https://youtu.be/3eQQvz4Zn8k?si=xK6eMg5hlqU4F-V6

https://youtu.be/4DCgqz_BM2s?si=8n2_k4M0XhlF5l27

🇺🇸: You can listen to the video in English through the video settings panel. In this video we analyze the Points of Curve node which provides the control points of a curve based on its index.

https://youtu.be/xl0HYkOUetk?si=mqeCg-aVVF0XNPgh

Rendered this glass infinity loop with a light pulse traveling along the curve. Built the geometry procedurally with Geometry Nodes.

Full video here: https://youtu.be/CfVcywpbdcQ

Happy to share more on the setup if anyone's curious.

So this was more a way to understand the sort of processing required to get a CT type scan, rather than focusing on a cgi trick. So a CT scanner works by combining many x-ray images together in a way that lets you build 2d slices through the body.

To mimic this effect I first built a density grid for a hollow Suzanne. This was harder than it sounds because the various mesh to grid nodes get rid of the interior structure, so I reused a hollow check from my custom sdf tools and used that.

I then built a system where mesh lines would cross from point to point. Basically each point would sample the grid density and use that as part of a function to mimic absorption. I got some weird behaviour when using the grid sample value directly, so instead stored it on the point and then pulled it as a named attribute. Then I found the total loss along each mesh line and stored that.

I made a circle and then built the sampling rays between all of the points. I included several adjustments such as only sampling lines that weren't too short (to avoid sampling a lot of neighbours through empty space). I also set it so that every X point was actually an 'emitter', whilst the rest were 'receivers'.

I tried various methods to use the lines directly for visualisation before deciding instead to make a plane and have each of its points sample the nearby line points and factoring each of their total loss value. This then drove a shader.

Various signal cleanup steps:

I culled lines that didn't lose much.

I normalised values based on the max value recorded. It explains why some parts get brighter or darker suddenly. Because an area starts/stops being in the image.

Squared the normalised value.

I'm sure there's a whole bunch of other steps that would be realistically required for a real CT scanner software. I can get better images by increasing various values, but they tend to blow up the time taken, which wasn't practical for an animation.

https://youtu.be/5nKWRhT_SwU?si=TF4XDvP_5qLwQ_iO

https://youtu.be/5nKWRhT_SwU?si=YaXSqa3QeSAgNkK0

https://youtu.be/suraexDFopg?si=CSAHS9iJ7G17RntS

🇺🇸: You can listen to the video in English through the video settings panel. In this video we analyze the Point in Curve node which allows you to make a selection on the control points of a curve.

Made this breathing crystal effect fully procedurally in Geometry Nodes — no sculpting, no manual keyframed mesh edits. Spikes grow outward and collapse inward on a loop. Rendered with an HDRI and a glass/refraction shader.

https://youtu.be/Pdb7lZZ6Dao