Creating a Poetic Portrait Inspired by Tarkovsky's Movie Using Blender

Introduction

Hello everyone! I’m VVXIE from Sichuan, China. I’m currently a freelancer, mainly working in the field of designer toys and figurines. I’ve loved drawing since I was a child, and after getting my first computer, I developed a strong interest in graphics-related software. Starting from basic image editing and drawing, I gradually moved into 3D software – it felt like building with blocks. Through self-exploration and scattered tutorials found online, I slowly became more skilled. It wasn’t until I started working that I was exposed to a more professional environment, which helped me further enhance my abilities and develop my own learning habits.

I’ve always had a deep passion for animated films, so my learning path has always been centered around that direction. Over time, I ended up acquiring a wide range of skills. I once hoped to create my own animated short film, but due to limited energy and time, that goal hasn't been realized. Eventually, I realized that creating and expressing myself can take many forms, so I chose to focus on character modeling, which I’m best at.

Before this, I worked in game and animation studios, mainly responsible for 3D character development. However, my true interest has always been in original design. By chance, I later co-founded a studio with friends, and that’s when I began creating work in the style I pursue today.

Inspiration & References

“驻足” – it conveys the idea of pausing, staying, or coming to a halt in a particular place. It implies not just a physical stop, but also an emotional dwelling, a temporary resting point, or even a moment of quiet connection with one’s surroundings. The inspiration for this came from a scene in The Mirror, directed by Andrei Tarkovsky. It wasn’t the film’s theme that moved me, but a single visual moment: a boy walking through a snowy field, stopping beneath a bare tree, as small birds from the branches flutter down and land gently on his head.

That image left a deep impression on me: the snow-covered ground, the leafless tree, the birds, and the quietly sorrowful boy. These elements came together to form a scene rich with life and poetic resonance.

Andrei Tarkovsky, The Mirror

Later, I began expressing my feelings through sculpting, and it was during that process that the visual elements of the piece gradually took shape. They weren’t something I had designed from the start.

As for the theme, I don’t want to overly emphasize any particular message or deeper meaning behind the work, nor do I want to explain or define it at length. I feel that doing so would be a kind of disruption; it disrupts the space where both the creator and the viewer can engage with the work through intuition and emotional response. To me, that space for emotional thinking is something truly beautiful.

The image simply exists, and in its presence, everyone is free to feel something different. That makes it open and free, rather than fixed or defined.

In this sculpture, the broken tree stump, the girl holding flowers, and the small birds each linger in their own place. That’s why I named it “驻足.”

Modeling

This piece originally began as a practice exercise, without any pre-planned design. In such cases, I usually start with a bust, meaning the head and upper torso, as it helps establish the overall feel of the character. During this process, I begin to imagine the scene she might be in. As inspiration sparks, I gradually add elements to enrich the composition. This eventually led to the following rough sculpt.

All the elements were quickly shaped using Blender’s dynamic topology sculpting tools.

I choose Blender for sculpting because it conveniently allows manipulation of objects directly in 3D space based on spatial coordinates, accommodating various modeling approaches such as sculpting, polygonal modeling, and procedural generation, making it easier to modify and work with more freely.

After establishing the overall form, I then began refining each part one by one.

The entire process was guided purely by my intuition and experience, without using any references. This includes the bird’s design. It is simply a bird, so I didn’t explore what species it might be. I just followed my feelings.

For the bird model, I separated the bird from the main sculpt to refine it as an individual “asset.” The feathers on the body were sculpted as a whole, while the wing feathers were created by inserting individual feather models one by one. I first set the bird’s pose to a neutral baseline, then manually adjusted its posture through deformation to recreate the pose from the original sculpt.

The flowers in the hand were created using the same approach. I started with a single flower, then duplicated it into multiple flowers and adjusted their overall arrangement to add variation. Since the piece is intended to be made into a physical object, the gaps between petals and stems needed to be filled in.

Because I want the overall model to maintain a certain stylistic sense of line work, I emphasize hand sculpting throughout and avoid using procedural methods or cloth simulations to shape the form.

However, I do pursue greater efficiency in other areas, such as more standardized file management and workflows, as well as software setups and plugins that suit my needs. I also develop plugins by having AI write scripts to add functions missing in the software – but this requires a clear understanding of the software’s underlying logic, so you can better communicate your functional needs to the AI. Currently, I have over 50 plugin functions created this way, covering multiple system modules. As the saying goes, “To do a good job, one must first sharpen one’s tools.”

Retopology & Unwrapping

Regarding topology, I’d like to introduce a Blender add-on called QuadRemesher. It functions similarly to ZBrush’s ZRemesher, recalculating the quad-based topology according to the model’s structure. Although its topology guidance isn’t as advanced as ZRemesher’s in ZBrush, it’s sufficient as a base for sculpting models. For areas that require specific topology, I can manually adjust those parts as needed.

Once I have a relatively uniform quad topology model, I add a Multiresolution modifier to subdivide it, then shrinkwrap it onto the earlier voxel model and apply the shrinkwrap modifier. This way, the Multires modifier inherits the corresponding mapped details for further sculpting.

After completing the manual sculpting of shapes and details, I can also add a Displace modifier with a texture to create fine surface details and textures on the model.

Once the model is completed, the next step is the UV mapping. Here is a rough outline of my workflow:

1. Quad topology
2. Multi-level subdivision sculpting
3. Low-level subdivision UV mapping
4. Reducing polygon count while preserving UVs
5. Baking high-poly to low-poly models

When I first tried the workflow of coloring using low-poly baking, out of habit, I would use the model with reduced polygon count from the high-poly model as the low-poly model, and then assign UVs. However, this approach would mean that whenever the model is modified or fine-tuned, the steps after UV mapping have to be redone, which is very inconvenient. So, I mapped the UVs for the lowest-level mesh of the subdivided sculpted model. This way, when the model is adjusted, the UVs remain intact, reducing the workload caused by modifications.

The UV mapping is done in Blender. I just need to unfold the UVs while minimizing distortion and stretching. Since the mesh is calculated by a plug-in and is not standardized, I didn’t pay much attention to the position of the UV seams, as long as they can be unfolded in a relaxed state.

For complex structures like tree roots, I split the UVs through face selection, which is relatively fast. I need to make the UV edges roughly follow the structural lines, without carefully handling whether the UV edges are continuous, and this won't affect the effect during material creation.

The UV layout is also automatic, though I still shrink some invisible parts to maximize the utilization of the texture map.

The above is the fastest UV unwrapping method I use.

I perform the polygon reduction in ZBrush because it allows for reducing the number of polygons while preserving the UV layout. It can be done simply by enabling “Keep UVs” in Decimation Master for calculation.

Finally, I use the Smooth command in the Deformation toolbar to smooth the model once, which reduces the vertex shadowing caused by polygon reduction and does not affect the model structure.

Texturing

Once baking is completed in Substance 3D Painter, managing layers becomes straightforward as long as they are set up according to material rules and organized through folder nesting. Since I strive to create materials following the actual painting processes of real objects, focusing more on the overall color sense rather than adding overly complex details or pursuing highly realistic material effects, the entire process remains relatively simple.

After setting up the mask ranges, gradient effects, and texture roughness, I spent a significant amount of time adjusting the colors in conjunction with the rendering results to enhance the color impact.

Lighting & Post-Production

For the lighting, I only used a single directional light in the scene. To create a softer visual effect, I increased the Diameter parameter so that the shadow edges appear smoother, while still preserving the feeling of sunlight. As for the skybox, I just casually chose an HDRI image to serve as ambient lighting – the main goal was to introduce a slight variation in ambient light. I didn’t deliberately set up side lighting, backlighting, or anything like that, because I was aiming for a soft and gentle look.

One directional light and one HDRI – that’s the entirety of my lighting setup.

To emphasize a soft and gentle style, I also push the look towards softness in post-processing. I always like to keep the lighting and color contrast subtle and elegant because I feel this makes the colors in the scene stand out more – almost like a painting. I can’t help but keep adjusting it this way.

In summary, I divide the image into four tonal ranges: shadows (dark tones), midtones, highlights (bright tones), and whites (specular highlights). Following my usual approach, I treat each as follows:

Whites: Pull down the brightness to prevent highlights from becoming too harsh or blown out.

Bright tones: Lower them slightly to reduce contrast in the lighter areas.

Midtones: Slightly brighten them to make the overall image feel more luminous.

Shadows: Slightly brighten and add a bit of saturation to make the shadow areas more transparent and lively.

In short, all rendering and post-processing efforts aim toward achieving a coherent stylistic direction, with the visual feel of the image as the ultimate goal.

Conclusion

I’ve always been in a state of learning and exploration, fueled by a strong passion for what I do. I’m clear about the main goals I want each of my works to achieve, and I don’t believe any piece can ever be truly perfect or flawlessly cover every aspect – that’s simply impossible.

For me, the real challenge lies in the design. Creating a 3D model is a longer and more complex process than drawing a concept design. It involves many steps, and even small adjustments can lead to repetitive workflows. When you get deeply absorbed in technical details, you might overlook design issues. After revising several times, you might become numb to the work, making it hard to grasp the key points or make accurate judgments. This raises the bar for 3D creation, requiring you to continually improve your skills. Throughout the entire process, you are the one in control – from initial sketches to final rendering, it’s all part of the creative journey. Along the way, ideas, mindset, and aesthetics all evolve, leading to different results.

So I believe that maintaining a calm and steady mindset to focus on creation is the biggest challenge. While constantly improving, you also need to balance the various conflicting problems that arise and learn to accept them – accept imperfection and face it with composure.

Regarding the piece “驻足,” I completed the GK (garage kit) unpainted model last year, but didn’t design the colors at that time. Recently, I started using Substance 3D Painter to texture the static model. Previously, for convenience, I painted vertex colors in ZBrush and rendered in Keyshot, but that method imposed many limitations on material design. Now, although the process still involves UV mapping, low-poly models, and baking – which are tedious steps – the improvement in the final effect is significant. Using Marmoset Toolbag for rendering allows me to better showcase the work and make it more complete. For the future, I hope the physical model can also achieve this level of presentation, but there is still much work to be done.

VVXIE 韦韦谢, 3D Artist