8/27/2026, 7:05:55 PM

DRAGONFLY

3D Models
Sculpting
Texturing

My second contribution to Dragonfly online drawing school. Read about the proccess behind the first course here. Just like last time, I was responsible for creating the images used to decorate the course webpage.

The whole project was completed in one week totaling 58 hours of work.

~BREAKDOWN SECTION~

Last time, I was given a lot of models based on a primitive geometry. To make their shape eye-catching, I implemented displacement mapping into my workflow, creating height variations on dense geometry, adding micro-level details and making them more realistic. For this one, my initial plan was to stick to the same approach, however..

When I got first concepts, I saw that some of the models were made of repetitive parts (diskoball and glass pieces, brush and bristles and so on). Doing this kind of stuff manualy is a pain for a modeler, cause you waste enoumorous amount of time on a repetative task (which is not good for deadlines and your mental health).

So that's when procedural modeling and geometry nodes come in. I had experience working with node-based modeling inside Houdini, but was curious to try the approach inside Blender and keep the work within a single piece of software.

In a nutshell, procedural modeling is similar to programming: you define “what” appears “where” under which “conditions”. The approach reduces modeling time and helps with fast iteration. But don’t forget that writing these “conditions” takes time too, so use cautiously.

In gamedev and VFX, procedural modeling is a go-to aproach for creating non-repetative and catching enviroments from pre-made assets relatively fast.

PALETTE

Really proud of this one. To make the palette I used pretty much every 3D related knowledge I had (except photogrammetry).

There was a little thought just to find old palette, scan it and call it a day. The only problem is that the philosophy of the model lives in its texture (it’s meant to depict carefuly subtle variations in color and how they blend together).

With an entry-level scanning setup like the one I have, the only usable thing you get is the geometry. Textures won’t be production ready (blurry as hell) - not the quality I’m trying to achive.

Also, while my setup is good at getting the overall object’s shape, it can’t catch micro height variations (which were the things I wanted most from the start).

Draft Version

So everything was made the good-old way: by hand. LOTS of masks for displacing and fluid sims. I spent most of the given time on it, but when I saw the result on my screen.. no regrets.

Provided Concept

TERMOMETER

It’s one of those objects you see so often throughout your life path that you could probably make it with your eyes closed, knowing exactly how and where it naturally wears over time.

The only complication was to come up with the right refraction values for the glass parts.

DISCOBALL

The first one on the project made proceduraly. My philosophy was to make every piece of glass the ball consists of unique, otherwise the model would feel sterile and render-like.

Geo-nodes did the job. I created one glass piece, scattered it across points extracted from a simple sphere, applying random transformation tweaks to each piece.

For an extra level of variety I also projected roughness map from camera’s POV on the ball.

COMPACT DISK

There is a saying in VFX world: “If you want something to look real, the easiest way is to JUST copy reality”. So I measured the box I had within my reach and recreated it in Blender at 1:1 scale.

The disk itself was also made to match real life reference (multiple layers of geometry), which added volume to the model.

Choosing the right amount of wear was tricky. The box I had the pleasure of spinning was very scratched and had fingerprints on it. However, for the final version, we decided to keep it to just a few scratches.

PAINTBRUSH

Another model that wouldn’t be possible to create non-procedurally. The lower part of the brush (just like the diskoball) is made from variations of a single curve, using which final cylinders/bristles were created.