USSR Headphones AG-2

Prop Breakdown

Eugene Karmazin


Eugene Karmazin

Hard-Surface Artist


Hello, my name is Eugene.
My journey into 3D graphics was initially winding, as I wasn't entirely sure what I wanted to achieve and how seriously I wanted to pursue it.
At first, I thought I could just do it as a hobby, but over time, I realized that I wanted to engage with 3D on a professional level.
For myself, I chose three main areas that I would find most interesting to work on: props, weapons, and vehicles.


While working on my latest project, a medieval helmet, I reached a point where I felt stuck, and my skills weren’t improving.
There was hardly any difference from my previous work.

I started thinking about how to address this issue, and the solution turned out to be simpler than I thought. I sought feedback from a skilled artist on ArtStation.

He thoroughly analyzed my work and provided comprehensive feedback, which significantly elevated the quality of my work. Essentially, I had been neglecting what I already knew up until that point.

Learn from my mistakes: don’t hesitate or be lazy to ask for feedback.


I wanted to achieve: This led me to the realization that to progress further, I needed not just any random person from ArtStation, but preferably a professional from the industry.

I found such a person; his name is Dmitriy Bely, a fantastic artist with years of experience in the field.
We agreed on him mentoring me for my new project, which I will talk about next.

The main goal for this project was to solidify all the knowledge I already had, identifying strengths and weaknesses.


The first thing I needed to do was to figure out the name of these headphones. Since I initially compiled a huge reference board with various objects I was interested in creating, many of which I didn’t know the names of, I used reverse image search.

I recommend using for image search; they have an excellent image search feature. Personally, I find it even better than Google’s.

After learning the name, I utilized all available resources. Don’t limit yourself to just Google and Pinterest, for example.
Your best friends also have various websites for selling all sorts of stuff, both new and used, such as eBay, PicClick, Amazon, and others.

The key is to gather images from every possible angle. One perspective is usually not enough; multiple viewpoints enrich your perception of the form. Also, don’t forget about references for textures; this can be a separate stage in the search for references.

It’s great when you have excellent photos of your object with delicious and interesting textures, but unfortunately, the opposite situation often occurs. In that case, it makes sense to look for similar objects or objects with similar materials.



During my time working with 3D, I noticed that while modeling, whether it’s a small prop or an entire scene, at some point, I start flying around the object and jumping from one part to another, not giving them the attention they deserve.

This often leads to mistakes, such as proportions or topology issues. To overcome this problem, I gradually learned to set subtasks for myself within the scope of the object. Most importantly, I avoid getting distracted by other things until the subtask is completed.

This helps me concentrate my attention and prevents me from missing errors caused by distractions during the modeling process.

My modeling process didn’t differ much from what you can find online; it followed a classic pipeline: blocking, draft, mid-poly, high-poly, low-poly, UV unwrapping, baking, and texturing. I work in Blender, and here, turning mid-poly models into high and low poly is relatively easy due to modifiers.

Then comes UV unwrapping, baking, and texturing. During the mid-poly stage, I refined all the forms in their detailed state. All necessary booleans are applied, but wires are not yet converted from curves. Complex geometric transitions are simplified, areas are not merged, and bevels are virtual.

At this stage, I try to create geometry with the same number of faces as the final low-poly version right away. There’s no point in doing this work multiple times.


At the draft stage, I initially struggled to get the proportions right in the upper part of the headphones. The headband looked too thin and fragile, creating a sense of an unrealistic object or poorly designed one.

In such cases, aligning the viewport camera approximately with the same perspective as the real object in the reference helps. This immediately clarifies all the incorrect proportions.

Below images are the initial iteration & final midpoly.

High Poly

At the high-poly stage, for the hard surface parts, I used subdiv modeling. The ear cushions, tape, and leather parts of the headband were created by Marvelous Designer. I decided to try sculpting the strap in ZBrush.
One of the most interesting features of this object is the large voluminous folds on the ear cushions.

Creating them wasn’t too difficult, but figuring out how to make them look realistic was a bit more challenging. Their construction is quite simple: a fabric disk is sewn to the inner surface of fabric strips.

These strips have elastic bands on the opposite side from where the disk is sewn. The elastic bands are there to ensure that the fabric fits snugly against the plastic part. Then, all of this is stretched over a foam ring that holds the shape.

Recreating this was straightforward, but achieving the large folds right away was a bit challenging.

I had to experiment with the settings and make use of the ‘pin’ tool. In the settings, I increased the values for shrinkage warp and weft.

I also physically enlarged the fabric pieces so that the excess length could fold into the desired creases later. I used the fabric preset ‘trim fusible rigid,’ which eliminated the unwanted fabric yellowing effect present in the default setting.

With the pin tool during the simulation, I grabbed a part of the fabric, pulled it to the side, and then pinned it in place. This ensured that the resulting folds didn’t smooth out during the simulation.

The below images are Marvelous’s first & final iteration.

With the leather headband and tapes, I didn’t face any significant challenges. The key was to be patient and adjust the necessary parameters until the result satisfied me.

After achieving the desired result in Marvelous Designer, you can export this geometry to ZBrush if you need to make adjustments in specific areas or create a more refined mesh.

For the strap and damaged parts, I first created a base mesh of the strap in Blender and unwrapped it.

In ZBrush, I increased the polygon count and then, in the ‘Surface’ tab, clicked on ‘Noise.’ In the opened window, I selected ‘UV’ and adjusted the sliders as needed.

After that, I added damages using brushes like Standard, Inflate, ClayBuildup, and Trim Dynamic. DamagedHSPart. Strap. FinalHP.

Below images are the Damaged HS part, strap and final HP.

Low Poly

I created the low-poly model while adhering to all the optimization principles commonly used in game development.

However, I slightly exceeded the geometry to ensure high quality, as this project was primarily for my portfolio, and I didn’t want to heavily reduce the geometry.

For the ear cushions, I optimized them first using the Decimate modifier in Blender and then made some manual adjustments to the geometry.

UVing & Baking

I use RizomUV for unwrapping; I appreciate the tools it offers as well as the algorithms that allow flattening and optimizing shells with minimal distortion.

For packing, I use the UVPackmaster addon in Blender. In my opinion, it currently has the best packing algorithms in the industry. For this prop, I decided to use a single texture set at 4K resolution.

I utilized overlaps to achieve the highest texel density, resulting in just over 100px/cm. Here are a few important points I always consider:

  • Minimize seams wherever possible.
  • Decrease texel density for areas not prominently visible.
  • Apply symmetry for repeating elements.

Before baking, I group objects into bake groups and name them with ‘low’ and ‘high’ suffixes. For baking the normal map, I used Marmoset.
It’s straightforward and quite fast, and it has an excellent skew tool for real-time editing of baked maps.

I checked the bake results both in Marmoset and Maya. In Marmoset, I adjust the material by cranking up the metallic setting and reducing roughness to its minimum values.

These manipulations help spot black spots and other artifacts that might have occurred due to mismatches between high-poly and low-poly versions or shading issues on either the high-poly or low-poly meshes.

I prefer baking AO and curvature maps in Substance Painter, as they tend to turn out better there than in Marmoset.

I bake multiple versions with different cage sizes and average normals settings, then use these maps in layers within Substance Painter.
I edit any artifacts using masks.

Below is an image of the AOandCurveMapsFix.



Texturing, for me, is one of the most interesting yet challenging stages. I can spend quite a lot of time on it, often redoing several stages over and over again.

Here, I might revisit the search for additional references if I’m not entirely sure how to convey the properties and damages of a specific material.

I primarily texture in the metal/roughness workflow.

Once all maps are correctly baked and adjusted, and additional references are found, a couple more things need to be done, namely preparing custom stencils in Photoshop or finding ready-made ones online. I convert images I find online into alphas in Photoshop.

Some of them were of lower quality, so I enlarged them using the AI Image Upscaler website, which seemed very useful to me. This can also be done using Stable Diffusion without any quantitative restrictions.

I started with the base color, then added color variations. For some of the materials, I used ready-made textures from Quixel Bridge, like leather. I didn’t need to find the exact leather texture as in the reference; I just found a similar pattern and adjusted the color in Substance Painter.


Also, at the very early stage, when breaking down individual parts into base colors, I tried to consider the possible history of these headphones. I learned that they were created in the mid-1970s of the 20th century but are still used in some countries of the former Soviet Union.

I was interested in imagining the journey they could have gone through over all this time. One way I decided to try to convey the object’s history is by breaking down parts into different shades, as if these parts changed over time and some appeared older, faded, or worn, while others seemed newer.

I often notice in the works of other artists how they damage, scratch, smear, telling the story of the object, but they do it too simply, and it doesn’t look very appealing.

I try to make something interesting with each layer I apply to the textures, keeping in mind a few questions.

First, what material is it: plastic, fabric, paint, etc.?

Second, where are the damages applied on the material: edges that are chipped off, or sweat stains on the ear pads?

Third, how did certain damages, dust, and other external influences get on the material or were applied to it? Returning to the ear pads, there are several layers of dust I placed in the fabric’s recesses and between the threads, as well as various sweat stains, both old and fresh, as if they had been used recently.

Fourth, it’s the interaction of the material with the external environment, like places where the leather rim has been rubbed, which remain from constant wear.

Rendering & Lighting

The rendering stage is crucial, as it can either enhance or diminish your work. I often notice artworks where textures might not be very interesting or overly simple, but an interesting angle, well-placed lighting, and an intriguing environment make the final result much better than it actually is, or sometimes the opposite.

For my final render, I prepared eleven scenes; many shots didn’t make it into the final result. I usually use Marmoset Toolbag for rendering. In the fourth version, ray tracing support was introduced, allowing for an even more realistic image.

However, I noticed that it slightly blurs the image, but it’s not a problem and can be easily corrected in post-processing.

I start by setting a suitable HDRI for the sky and reduce the brightness almost to a minimum.

This provides a subtle background light, so there are no completely black shadows on the object.

I begin with classic three-point lighting but never stop at just three lights because there’s always something missing, and additional parts need to be illuminated.

Always enable all maximum parameters for shadows and reflections in the rendering settings; this significantly affects the image quality. During the rendering stage, I continue to develop the object’s story.

I was interested in creating a render reminiscent of an old radio shack photograph, where the headphones are part of the overall composition but remain in the foreground.


Here, everything is quite simple. I try to make the render in a way that doesn’t require final processing, but a little adjustment never hurts.
I do this using the Camera Raw filter in Photoshop.


In conclusion, I’ll note a few commonplace things that we often forget. Watch and study more tutorials. Always strive to find your visual references.

Feedback is important, and don’t forget to take breaks.

Set goals for yourself and achieve them. If you don’t succeed, analyze why and try again. Develop new skills and, if possible, dedicate time to creative hobbies.