3D Modelling and VFX 2

[denysbatsmanov]

Project Idea: Asset Interaction

As an optional feature, I might implement a basic interaction system that allows users to inspect objects. This system will allow selected assets to be picked up, rotated, and viewed up close. While simple, it adds a certain level of engagement, which again makes the environment feel more alive. Also being able to inspect objects closely highlights material detail. And it allows assets to be showcased individually, which is useful for demonstrating work.

This involves basic input handling, object selection, and transform control within the Unreal Engine node system.

[denysbatsmanov]

Project Idea: Niagara

I want to try to avoid a static environment, so using Niagara can help with small movements, which can significantly improve realism. I can add particles such as dust, debris, and leaves.

I also plan to simulate hanging clothes reacting to wind using cloth physics. This involves weight painting and adjusting gravity and wind settings to achieve believable movement. Weight painting allows control over how different parts of a mesh react. For example, parts of the cloth can remain fixed while others move freely.

The goal is to create a responsive environment where everything works naturally.

[denysbatsmanov]

Project Idea: Textures and Shaders

I'm planning on making real/semi-realistic materials with a focus on surface detail.

I will use Substance Painter to texture assets and Unreal Engine 5 for shaders. One of the features I plan on using is Parallax Occlusion Mapping. This allows surfaces like brick walls to appear more detailed without increasing polygons.

POM creates the illusion of depth using height maps. This is especially useful for close-up surfaces where flat textures would otherwise look unrealistic.

I will also use mesh blending to improve transitions between assets, for example, where walls meet the ground. This helps avoid visible seams and repetition and makes the scene look more realistic. And I will use decals to add variation, including dirt, puddles, and posters. This adds realism without requiring unique textures for every asset. I may use AI tools to generate base textures and refine them after.

[denysbatsmanov]

Project Idea: Assets

I will probably create assets using a modular workflow. This allows me to reuse components across the environment and build scenes efficiently.

The main assets include walls, windows, doors, and ground pieces. These will form the structural base of the environment. I will also create supporting props such as crates, barrels, pipes, bins, and signs to add detail and variation.

I will model most assets in Maya. For more detailed elements, I will use ZBrush to add surface variation. I may also experiment with AI-generated base meshes. However, these will only serve as a starting point. All assets will be refined manually after.

[denysbatsmanov]

Project Idea: Victorian & Industrial London

Alongside game references, I researched real-world Victorian and Industrial London. Streets are pretty crowded, and buildings mostly use bricks. Everything is tightly packed, streets and alleys are narrow, light is limited, and a lot of buildings look alike.

What stood out most was the amount of visual noise. Posters, signs, pipes, cables, and dirt are everywhere, surfaces are rarely clean. This level of imperfection adds realism and storytelling. Streets also often feel enclosed, with tall buildings and limited visibility, which should add depth to the scene

I will use some of these references to model the assets/props, adjust proportions and create material variation.

[denysbatsmanov]

Project Idea: Inspiration and Research

I looked into how modern games build atmosphere through environment design rather than relying only on narrative or characters. My primary focus was on games like Assassin’s Creed Syndicate, The Order: 1886, and Dishonoured 2 because of their time periods and style. These games consistently use dramatic lighting, scale, and material detail to create believable worlds. Environments feel alive, surfaces are not clean, materials show age, wear and tear.

In Assassin’s Creed Syndicate, Victorian London is dense, industrial, and layered. Smoke, bricks, and narrow streets create a strong sense of place. In The Order: 1886, lighting plays a major role. Soft fog, warm lamps, and cool shadows create contrast and depth. Dishonoured 2 uses stylisation and semi-realism, yet still features very strong lighting and composition.

My goal is to create a narrow alley environment that feels believable and grounded. I will focus on controlled lighting, layered materials, and prop placement to achieve this.

[denysbatsmanov]

Activity: Nanite Foliage

Nanite is UE's 5 virtualised system that allows very detailed meshes to be rendered in real time. It partially removes the need for traditional optimisation methods like manual LODs/Retopology. Very recently the added support for foliage, which is a significant step forward for environment scenes.

Foliage has always been one of the most expensive and trickiest parts of a scene. It usually relies on alpha cards, instancing, and aggressive LOD setups to stay performant. This often limits how detailed assets can be. Nanite changes this by allowing more complex geometry to be used without the same level of optimisation.

This is important because it shifts how environments are built. Instead of spending time creating multiple LODs and heavily optimising assets, artists can focus more on the quality, and especially the density, of the scene. It also allows foliage to feel more natural by using real geometry instead of flat textures. Nanite foliage is still a new technology, especially in areas like transparency and wind animation, but it is already being explored in UE5 projects. It shows the improvement towards removing technical limitations from the art pipeline.

So there is less focus on optimisation and more focus on asset quality, composition, and lighting. I will most likely try to use this technique for my future projects

Source: https://dev.epicgames.com/documentation/unreal-engine/nanite-foliage

[denysbatsmanov]

Activity: Heterogeneous Volumes & Volumetrix

Heterogeneous Volumes is a feature that allows more realistic volumetric effects, such as smoke, fire, and fog to be rendered in real time. They support complex simulations like VDB files, which are often used in film and high-end VFX. Volumetric effects have traditionally been difficult to use in real-time environments. They are either simplified using particles or pre-baked into textures, the same as lightning before RayTracing. This limits how realistic they can look. Heterogeneous volumes allow for more detailed density, lighting, and shadow interaction within these effects.

This feature is important as it brings high-quality VFX volumetrics closer to real-time scenes. Artists can now use more accurate simulations without fully relying on rendering. It also improves how effects interact with lighting, making scenes feel more believable. Heterogeneous Volumes is still developing and, at the moment, can be performance-heavy. But it still allows for much higher-quality VFX in real-time engines.

This feature greatly improves what can be achieved in real time. It allows for more complex atmospheric effects and reduces the need to fake volumetrics.

Source: https://dev.epicgames.com/documentation/unreal-engine/heterogeneous-volumes-in-unreal-engine

[denysbatsmanov]

Activity: Neural Rendering

Neural rendering is a new method that combines computer graphics with machine learning (AI) to generate images. Instead of relying only on usual rendering techniques like geometry, materials, and lighting calculations, neural rendering uses trained models to predict how a scene should look. Generally, rendering pipelines are based on physically accurate simulations, which can be quite heavy and expensive. Neural rendering offers a different approach by learning visual patterns from data. Once trained, these systems can reproduce complex lighting, materials, and geometry with much lower computational cost in some cases. This is somewhat similar to AI Upscaling like DLSS 4-5, but a touch beyond.

One of the most well-known examples is Neural Radiance Fields (NeRFs). NeRFs can reconstruct a 3D scene from a set of 2D images by learning how light behaves at different points in space. Similar to Photogrammetry. Instead of storing geometry in a usual way, the scene is represented as a continuous function. This allows for a smooth switch between views and can produce highly realistic results.

Neural rendering is also used in real-time workflows. Similar techniques, such as AI Upscaling, denoising, and frame gen, are already integrated into modern pipelines. For example, AI denoisers can clean up noisy ray-traced images much faster than traditional methods, allowing for higher quality rendering in real time.

This can change how assets and scenes are created. Artists may no longer need to build everything manually, as some elements can be generated or enhanced through learned data. It also introduces new workflows where captured data, such as photographs or scans, can be turned directly into usable 3D content.

However, neural rendering also has the usual AI limitations. It can lack direct control compared to traditional methods, and the results depend heavily on the quality of the training data. It can also be difficult to edit or art-direct in the same way as standard geometry and materials.

Neural rendering might shift how content is produced. It does not replace traditional workflows, but it adds a new layer of tools that can speed up production and enable new forms of visual creation.

Source: https://www.matthewtancik.com/nerf

https://eureka.patsnap.com/article/how-neural-rendering-is-changing-the-future-of-cgi

https://www.microsoft.com/en-us/research/blog/renderformer-how-neural-networks-are-reshaping-3d-rendering/

[denysbatsmanov]

Activity: Real-Time Rendering

Unlike offline rendering, which can take minutes or hours per frame, real-time rendering can maintain a stable frame rate, usually 30 to 60+ fps or higher. This usually requires a balance between visual quality and performance so something LoD helps a lot. Usually, high-quality rendering was done offline using physically accurate calculations for lighting, reflections, and shadows. While this produces very accurate results, it is slow and not interactive. Real-time rendering uses optimised techniques and approximations to achieve similar results much faster. And recent developments have reduced the gap between real-time and offline rendering a lot. Tech like Ray Tracing allows for more accurate reflections, shadows, and global illumination. Or systems like Lumen can provide dynamic global illumination without the need for baking, while Nanite allows for highly detailed geometry without traditional optimisation methods. Together, these systems allow scenes that were previously only possible in offline workflows.

Real-time rendering is now used across multiple industries. In games, it is the standard for rendering interactive worlds. In film and television, it is used for virtual production, where environments are rendered live. This allows directors to see final-quality visuals during filming instead of waiting for post-production. In architecture and design, real-time engines are used to create interactive walkthroughs and presentations. So instead of waiting for renders, artists can see the result of their changes immediately. This allows for faster iteration and more experimentation with lighting, composition, and materials. It also shifts the focus towards optimisation and performance, as scenes must run in real time.

However, real-time rendering still involves trade-offs. Techniques are often approximations rather than fully physically accurate simulations. For example its still quite complicated to fully simulate large bodies of water in real-time. Artists need to understand how to balance quality and performance, especially when working with large or complex scenes. Still, for environment and lighting artists, real-time rendering has become the standard way of working. It allows for interactive scenes, faster feedback, and more control over the final result. It also requires an understanding of both artistic and technical aspects, as decisions directly affect performance and visual quality.

Source: https://fortes.vision/blog/real-time-rendering/

https://developer.nvidia.com/rtx/ray-tracing?sortBy=developer_learning_library%2Fsort%2Ftitle%3Aasc

[denysbatsmanov]

Activity: Autodesk’s AI

Autodesk has been integrating AI into its software to support and speed up parts of the creative workflow. Tools like Maya and 3ds Max now include AI-assisted features that help with tasks such as retopology, animation, and scene setup.

Many parts of the 3D pipeline are repetitive and time-consuming. Especially tasks like retopology, UVs, or adjusting animations often require a lot of repetitive manual work. AI tools can automate or assist with these processes, allowing artists to focus more on creative decisions rather than these tasks.

One example is AI-based retopology, where the software can generate clean topology from a high-resolution mesh. This reduces the time needed to prepare assets for use in engines. Another area is animation, where AI tools can generate or adjust frames based on input data, helping to speed up the work.

Autodesk has also introduced machine learning tools that assist with scene organisation and asset management. These systems can analyse patterns in a scene and suggest improvements or automate certain tasks.

This development is important because it changes how artists approach production. Instead of building everything manually, parts of the workflow can be assisted or accelerated by AI. This can lead to faster creation and more efficient pipelines.

However, these tools are not a replacement for artists. They require input, direction, and refinement. The final result still depends on the artist's judgement and understanding of form, composition, and detail.

For environment and prop artists, Autodesk's AI tools can reduce technical workload and allow more time to focus on modelling, materials, and overall scene quality.

Source: https://www.autodesk.com/autodesk-university/class/Accelerating-3D-Workflows-With-Generative-AI-2023

https://adsknews.autodesk.com/en/news/autodesk-flow-studio-launches-wonder-3d-gen-ai-model-to-help-creators-make-3d-characters-and-objects-faster-and-easier/

[denysbatsmanov]

Activity: FLIP Fluids & Dynamic Water Simulation

FLIP (Fluid-Implicit Particle) is a simulation method used to create realistic fluid behaviour. It combines particle-based and grid-based approaches to simulate how liquids move, interact, and collide with objects. It is commonly used in tools like Houdini and is widely used in film and VFX.

Water and fluid simulation are some of the most complex effects to reproduce accurately. General real-time methods often rely on simple shaders or animations, which can look convincing from a distance but lack physical accuracy. FLIP simulation allows for detailed behaviour such as splashes, waves, and interactions with the environment. In high-end workflows, FLIP is used to pre-simulate water, which is then baked into meshes or textures for use in real-time engines. More recently, there has been interest in bringing parts of this behaviour closer to real-time applications.

This development is important because it shows the gap between offline simulation and real-time rendering is getting smaller. Artists now have more tools to create believable water, whether through full simulation or hybrid approaches that combine simulation data with real-time techniques. FLIP fluids are mainly used in the creation stage rather than directly in-engine. The results can be used to generate high-quality assets such as splash meshes, flipbooks, or cached animations. This allows realistic fluid behaviour to be integrated into real-time scenes without the full performance cost of simulation.

Source: https://www.sidefx.com/docs/houdini/nodes/sop/whitewatersource.html https://www.sidefx.com/tutorials/houdini-tutorial-introduction-to-flip-fluids-simply-explained-part-1/

[denysbatsmanov]

Activity: Fluid Simulation Shader System

Fluid simulation shader systems are real-time techniques used to simulate fluid-like behaviour directly in the engine using shaders, rather than full physics-based simulations. A good example of this can be seen in Half-Life: Alyx, where liquids react dynamically to movement and interaction without relying on heavy simulation.

Usually, fluid simulation methods like FLIP are quite heave/expensive and are usually pre-baked. In contrast, shader-based systems approximate fluid behaviour using textures and distortion. This allows fluids to respond instantly to player interaction while remaining performant.

In Half-Life: Alyx, liquids inside bottles and containers react in real time to motion. This is achieved by using Pixel Shaders, Double Reflections, masks, and surface textures to create the illusion of moving liquid. The result is very exciting. This allows for interactive and responsive effects that would otherwise be too expensive to simulate.

Fluid shader systems are now commonly used for water surfaces, small contained liquids, and stylised effects. They can be combined with other techniques such as screen-space effects or simple physics to improve realism without a large performance cost. So artists can create interactive fluid effects to be created directly in-engine, without relying on external simulation tools. It also encourages a more efficient workflow.

Source: https://www.youtube.com/watch?v=9XWxsJKpYYI

[denysbatsmanov]

Activity: Mesh & Texture Blending

Mesh and texture blending are techniques used to create smooth transitions between different assets in a scene. Instead of objects sitting on top of each other with visible seams, blending allows them to merge more naturally. Environments are usually built from modular assets. When meshes intersect, they often create hard edges or visible gaps that break realism. Blending techniques solve this by softening the transition between objects and materials.

For example, an addon I used before called MeshBlend. This addon uses methods such as vertex painting, decals, and distance fields to blend one mesh into another. For example, a rock can gradually blend into the ground instead of appearing as a separate object placed on top. This removes harsh intersections and helps assets feel more grounded in the scene.

Texture blending works by allowing materials to transition smoothly. Artists can blend surfaces like dirt, grass, and stone using masks or height-based blending. This adds variation and avoids repetition, especially in large environments. It allows for more flexible and efficient workflows. Instead of creating custom transition meshes for every asset, artists can reuse assets and blend them directly in the engine. It also improves the overall quality of the scene without increasing complexity. It helps create more natural environments and reduces the need for manual fixes. It also allows for faster iteration, as assets can be adjusted and blended directly in the scene.

Source: https://www.fab.com/listings/1f4abe73-4cda-42db-995a-c9f8ca4790e5

[denysbatsmanov]

Activity: Parallax Occlusion Mapping with Sihouettes

Parallax Occlusion Mapping (POM) is a rendering feature used to create the illusion of depth on flat surfaces by using a height map. Instead of adding real geometry, the shader offsets texture coordinates based on the viewing angle and simulates how light interacts with the surface. This creates a strong sense of depth while keeping the mesh simple. Compared to normal mapping, which only simulates lighting detail but does not change the perceived depth of a surface, POM simulates self-occlusion and perspective shift, making surfaces like bricks, stones, or cracks appear much more three-dimensional.

However, standard POM has a limitation. While it can create convincing depth inside the surface, it does not correctly affect the outer shape or silhouette of the object. At grazing angles, the illusion breaks, and the surface still appears flat. This is where silhouette-based POM comes in. Techniques such as Prism Parallax Occlusion Mapping extend the method by modifying how the surface edges are rendered. Instead of relying only on texture offsets, the algorithm traces view rays more accurately through the height field and adjusts the visible edge of the surface. This allows the silhouette to match the perceived depth.

These methods work by performing per-pixel ray tracing inside the height map to find the correct visible point on the surface. This improves depth accuracy and allows for effects such as proper occlusion, shadowing, and more realistic edge definition. This technique pushes texture-based detail closer to real geometry. It allows highly detailed surfaces without increasing polygon count, which is especially useful in real-time rendering where performance is important. It also improves visual quality in close-up views, where traditional flat textures often fail.

However, silhouette POM is more expensive than standard POM and is not always used in production. There are examples such as Crimson Desert, which recently used this technique in their game, showing that it is becoming more viable with modern hardware. In many cases, artists still need to balance between performance and visual quality, often choosing simpler techniques when necessary.

In conclusion, not all details need to come from geometry. Advanced shader techniques can create convincing depth and improve realism while keeping scenes efficient. Understanding when to use these methods is key to building high-quality real-time environments.

Source: https://inria.hal.science/inria-00606806/file/Dachsbacher-Tatarchuk-Prism_Parallax_Occlusion_Mapping_with_Accurate_Silhouette_Generation_SI3D07_.pdf https://www.cryengine.com/docs/static/engines/cryengine-5/categories/23756816/pages/29450143 https://chetanjags.wordpress.com/2013/05/10/parallax-occlusion-mapping-with-silhouettes/

[denysbatsmanov]

Project Idea

Here, I shifted towards a later time period, focusing more on early 20th-century urban environments. These concepts still have an industrial influence, but the setting feels more structured and less dominated by heavy machinery. The environments are based around streets and alleys, with brick buildings, storefronts, and urban layouts. Compared to the previous concepts, this direction is cleaner and more readable, with clearer composition and spacing.

Lighting is more natural and balanced, with softer shadows and less extreme contrast. This makes the environment feel more grounded and realistic, while still maintaining a certain atmosphere.

I equally like this style, it keeps the industrial influence but simplifies the scene, making it easier to manage while still allowing for detail, storytelling, and strong composition.

[denysbatsmanov]

Project Idea

Here I looked at an industrial city direction. These concepts focus on heavy industry, dense architecture, and large amounts of machinery integrated into the environment.

The key elements here are factories, smoke, pipes, and layered structures. The environments feel busy, with a strong emphasis on production and scale. There is also a darker and more atmospheric tone, supported by low lighting and strong contrast between warm industrial light and shadow.

I also looked at smaller spaces within this setting, such as narrow streets and back alleys. These areas still carry the industrial feel but allow for more detail and storytelling through props, materials, and lighting. In my opinion, this direction is quite strong visually because of the complexity and density. However, it might make it more difficult to build as a full environment.