How To Create DIY Radio With Blender

Introduction

Hello there! My name is Fabio Paiva, and I've been working with 3D for over 20 years. I'm the creator of 3D Masterclass, an online learning platform for 3D artists (more on that later).
 
I actually started doing 3D pretty late in life and had to switch from a civil engineering background. I initially did a lot of archviz, mostly because I couldn't draw anything and didn't even know tablets existed. Over time, I slowly gravitated toward anatomy and character work, which ended up becoming the focus of most of my career.
 
During that period, I studied anatomy extensively because I genuinely loved it, while working as a freelancer on various projects, including a game you can find for like ten dollars on Steam (Ryse: Son of Rome). Around that time, I finally felt confident enough in my skills and ended up landing a job in New Zealand about ten years ago.
 
After two years working at different game studios, I joined Wētā FX, formerly known as Weta Digital, as a 3D Modeler. That experience pushed me into a more senior level, simply due to the scale of the studio and its very tight and complex pipeline. There, I worked on film projects such as Avatar, Alita: Battle Angel, Guardians of the Galaxy, and many others.

That path eventually led me to Wētā Workshop, where I worked for the last six years. This was a major perspective shift in how I think about creating 3D assets. Making objects for the real world is a completely different beast. During that time, I worked on a wide range of projects, from films (creating assets such as armor, weapons, prosthetics, and masks) to premium collectibles and museum installations around the world, producing giant humans, animals, and props.
 
Having recently moved to Australia, I decided to invest more time in 3D Masterclass since I create everything for it. I'm the owner, the editor, the marketer, the web designer, and the janitor. It's a lot of work, and I'm genuinely proud of it.
 
Long-format videos may not be the internet’s recipe for short-term success, but I won't save time or resources when it comes to explaining all the hows and whys. I know that people who prefer this format will greatly appreciate all the effort I put in.
 
And finally, this is where this project comes in.

DYI 3DMC Radio

This project is part of a 9-hour course on 3D lighting I made for 3DMC. It belongs to one of the 10 demos in the course, where I walk through my lighting and rendering workflow, including editing, composition, and a section on lighting and rendering inside Unreal Engine.

You can watch a quick trailer here:

I modeled everything from scratch and tried to use as little SubD as possible. I've always wanted to work with everyday mechanical parts like these, but never had the right project to place them in, so this became the perfect context.
 
The DIY theme was inspired by works from artists such as Sabrina Garcia and several other pieces I've come across on Behance.

Being a DIY object, it is fairly easy to find perfect references and ideas for it. I first modelled the cassette tape, and from there it started the slow process of designing the whole box. Here are some of the references I used:

Modeling

I'm fascinated by how intuitive Blender feels once you really get a sense of it, especially coming from Maya, which I used for most of my career. It's very easy to switch workflows, and the UI is surprisingly friendly.
 
As I mentioned earlier, I wanted to use as little subdivision modeling as possible, not only as a challenge, but also to see whether the remaining methods are acceptable in terms of speed and quality. This essentially leaves us with variations of a Boolean-based workflow.
 
And the answer is: yes, they are, but it depends.

The TL;DR is that for objects based on simple shapes (cubes, spheres, cylinders, and so on), this modeling technique is extremely fast. There's no need to worry about edge flow or support loops. You focus on the form and add bevels only where they're needed. Similarly, if you want to have full control of curved and elegant forms, a CAD/NURBS workflow might be better.
 
To me, the paradox is that this method is both easier and harder at the same time. It's easier because you work directly with the form instead of mentally predicting how subdivision smoothing will behave. But it's also harder, because committing to bevels too early can be destructive and make it difficult to return to the base shape if needed.
 
In that sense, you actually need to be more experienced to fully benefit from this approach and know when to commit to things and move on to the next stage.

It's perfectly possible to work using only vanilla Blender, but I do recommend a few add-ons. These are the main ones I use: Hard Ops, BoxCutter, MESHmachine, and MACHIN3tools.
 
These are very powerful tools, and it can feel like you need to fully integrate all of their features to justify using them, but that's not true at all. I want to focus on a few specific tools that, on their own, already justify their cost:

MESHmachine: Fuse. Create bevels from chamfers.

MESHmachine: Unbevel/Unchamfer. Recover sharp corners.

MESHmachine: Symmetrize. Quick symmetry in edit mode.

MACHIN3tools: Many QOL tools. Quick vert/edge/face operations, more streamlined pie menus, align and focus commands, and many more. Here's the align command in action:

BoxCutter: Custom boolean tools built on top of Blender's native engine. Any boolean operation becomes easier and faster with this tool, even something as simple (and very common) as a Subtract.

You can create cutters manually or use the tools to create them. Booleans on Booleans is also a very powerful technique to achieve complex shapes. There's a lot more to it, but I use it very simply like this most of the time.

Hard Ops: A library of varied tools that's too large to list or categorize. There's probably a tool for everything Blender can do, but with a little bit of sauce on it. These are the ones I use the most, and I hope their names are almost self-explanatory:

• Reset Axis;
• Face Extract;
• Weighted Normal;
• Shade Solid;
• Smart Apply;
• Clear Sharps;
• Clean Mesh;
• Toggle Modifiers.

The list is short, but there's so much more to it, as almost every command has modifiers (with Shift, Ctrl, and Alt) for additional functionality.
 
These commands reflect how I work, where the idea is to create shapes as needed, either with booleans or by cutting and extruding as usual. As this often results in a messy mesh, I then run cleaning operations to ensure the shading is visually correct, especially when adding bevels at the end. Rinse and repeat.

UVs

UVs were required because I decided to texture this project in Substance 3D Painter for higher quality. In the end, I split all objects into groups, with each group having its own UDIM layout. Some objects, such as cables and smaller metal parts, didn't need a texture pass, so they were shaded procedurally in Blender.
 
I used approximately 65 textures in this project, all of which are 4K in resolution. I knew my RTX4090 could handle it, so I did it this way.

These are the groups that required UVs:

• Metals – 3 UDIMs;
• Dielectrics – 3 UDIMs;
• Wooden Box – 4 UDIMs;
• Cassette Tape – 4 UDIMs;
• OLED Display – 1 UV tile.

The bottom three are their own groups because I wanted to organize the Blender scene accordingly and treat them as "special objects", even though they are composed of more than one material.

Keep in mind that technically, it makes no difference. You could just as well have a single group with 30 UDIMs if you wanted to, and then make all the selections and masks in Substance 3D Painter directly.

Blender's native UV tools aren't the best, but they were more than sufficient for this project. I just did it all manually.
 
One free add-on I recommend is Texture Density Checker. I only used it to copy and paste UV scale values between UV islands to keep texture density consistent. To be fair, doing this by eye would also have been fine, since small differences are hard to notice unless the texture has a very clean, well-defined pattern.

Before exporting the FBX, I apply a Triangulate modifier. This modelling workflow naturally produces a lot of N-gons, and in this way, the triangulation is controlled on my side instead of being generated automatically by SP. It also makes it easier to track potential issues.

Texturing

Most materials fall into two main categories: dielectrics and metals. For both, the two most important channels are Diffuse and Roughness. For metals, there's an additional channel: metalness, set to 1. This channel tells the PBR shader how to interpret the other channels correctly in order to physically represent metallic surfaces.
 
I have a personal method when texturing, and I try to do it every time, unless I'm dealing with something very alien or experimental. I start by building the base material, which is essentially a perfectly clean version of the material I'm trying to represent. From there, I introduce complexity into the base color, even if it becomes slightly unrealistic. Two or three additional color variation layers are usually enough. Next, I add layers of height information if needed, and finally, I work on the roughness channel, using the previous layers as masks or anchors.

Here are the full diffuse and roughness channels. You can visit 3D Masterclass to watch a full 13-hour course on texturing, where I go into all of this method in detail. There's also a shorter course focused on materials and shaders, which the texturing course builds upon.

OLED Screen

Its structure is simple, but crucial for the effect to work. A solid (with thickness instead of one-sided) refractive object is needed in front of the text to sell the look. 

The OLED screen shader was created by the great artist Cody Evan. With it, all I had to do was create a black-and-white texture, and the shader handles all the pixelation automatically.

The refractive panel itself uses a very simple transparent glass material with a roughness texture. The added dust and hair particles also contribute significantly to the final effect.

Lighting

As mentioned earlier, I've already covered lighting in depth in a dedicated course, so here I'll focus only on the fundamentals. The lighting setup is built around a very subtle ambient light (using the free Easy HDRI add-on) combined with a large number of Area Lights.

The final placement and orientation of the lights is never solved in a single pass, nor do I start with a perfectly clear idea of the final image. Instead, I follow a simple rule: by controlling contrast, I can guide the viewer's eye to exactly where I want it to go.
 
This usually means a lot of tweaking, adjusting light size, shadow softness, intensity, color, and especially the distance to the subject. Creating flags, which are simply objects used to block or shape the light, is also part of the process. Whatever helps control contrast is fair game.
 
Manipulating the scene itself is just as important as tweaking light settings. In a professional photo studio, it's common to block lights, cut shapes, or even take multiple photographs and composite the best parts of each. In 3D, we have the same level of control, but we often forget to use it.

I also had specific cameras for each beauty shot. In this way, I was thinking and working as a photographer, controlling the environment at all times.

Compositing

I like to handle compositing directly in Blender for most of my projects, mainly because it's well integrated and powerful enough for what I need.
 
My most basic setup is simply a combination of bloom, sharpen, and lens distortion. If color correction is needed, I add it at the end of the chain or do it later in Photoshop. As you can see, for the bloom, I use the glare node to isolate the highlights and then add them back on top of the original image. This approach gives me much finer control over both its intensity and overall effect.

Again, if you want to learn everything about render passes, 32-bit workflow, and both basic and advanced compositing in Photoshop, Blender, and DaVinci Resolve, you can check my courses over 3D Masterclass.

Conclusion

You can see all the final images of this project here.
 
This project was very fluid and experimental, with the design evolving and changing along the way. The only idea I tried to keep from the beginning was to keep the main form extremely simple, so what's easier than a cube?
 
In isolation, it took me about a week to complete. However, since it was part of a course, the process stretched over several months, even after the asset itself was finished, due to scripting, recording, editing, and producing additional renders and images.
 
In any case, I hope you found this useful. Feel free to ask me anything, and you're welcome to join the 3DMC Discord community and find me there as well.

Fabio Paiva, Senior 3D Artist

Interview conducted by Amber Rutherford