Creating 3D Environment Models and Props for Games

We faced a scenario where due to an unoptimized UV atlas, the draw calls on a mobile RPG reached 350—after atlasizing and batching it dropped to 120, ensuring stable 60 FPS. Such cases are industry norms: a player views the environment 90% of the time, and every extra draw call is critical. Over five years, we've worked on more than 50 projects, from indie to AAA, and developed a pipeline that saves time and resources.

Our Process for Creating 3D Environment Models and Props

The process starts with analyzing requirements: engine, platform, asset role in the scene. For mobile games, polycount (up to 500 triangles for props) and draw calls (no more than 150 per frame) are critical; for PC and consoles—texture quality and LOD. These parameters are defined at the modeling stage.

Non-Obvious Technical Challenges

How to Ensure Seamless Tiling of Modular Elements?

Modularity requires planning at the modeling stage. If tiled floor sections are not aligned to a grid with steps matching the texture size (e.g., 1024 px), visible seams appear. For Unreal Engine 5 the standard grid is 10/100/1000 units; for Unity—depends on the project scale. We customize grid snapping for your pipeline.

Why Manual Lightmap UV Is Better Than Automatic?

In Unreal Engine, every static mesh requires a separate UV channel for lightmap. Automatic generation via Generate Lightmap UVs often produces overlapping, unequal distribution, and baking artifacts. We unwrap lightmap UVs manually, ensuring optimal texel density (e.g., 20.48 px/m) and no artifacts. Manual unwrapping gives uniform texel density, which is critical for baked lighting.

What Is a UV Atlas and How Does It Help Optimize Draw Calls?

A UV atlas merges multiple meshes into a single texture, reducing draw calls. This is especially useful for environment kits with repeated elements. We create atlases considering texture bitness and compression format for the target platform. For example, a set of 20 props went from 20 draw calls to 2 per scene using atlasization. Using UV atlas and batching can reduce draw calls multiple times.

Besides atlases, we use Physically Based Rendering (PBR) texturing for all assets. This ensures realistic material appearance under any lighting. Our work is based on PBR standards described in the specification, guaranteeing compatibility with most engines.

How We Build an Asset

Everything starts with context: engine, platform, asset role (hero prop or background filler), static or interactive.

1. Concept review. If no concept exists—we request references. Modeling without a brief is the most expensive way to get the wrong result.
2. Blockout. Quick shape blocking in 3ds Max or Blender. We agree on proportions and modular fitting.
3. High-poly or low-poly. For PBR pipeline we create high-poly with chamfer and details; for stylized games—low-poly with clean geometry directly.
4. UV + Bake. UV layout with correct texel density, separate lightmap channel, full PBR set baking in Marmoset Toolbag.
5. LOD chain. Minimum three levels (0/1/2); manual LOD for key assets delivers 2–3 times better quality than automatic while preserving silhouette. For hero props we manually refine LOD, controlling polycount from 2000 down to 150 triangles.
6. Export. FBX with correct axes, smoothing groups, named material slots. For asset sets we use UV atlas to minimize draw calls.

For Unreal Engine we configure import with automatic collision and LOD group creation. For Unity we use Asset Postprocessor to automate tiling and pivot settings. This cuts asset integration time by 20–30%.

Comparison of Automatic vs Manual LOD

Timeline Guidelines

Pricing is calculated individually; for sets—package terms. By ordering an asset set you save up to 30% compared to per-item pricing.

What's Included in the Work

• Source files (blend/max, ztl, high-res textures)
• Import settings for your engine
• Documentation on asset usage (grid, pivot, LOD levels)
• Team training if needed

How to Get Started?

To let us evaluate your project, prepare a brief with engine, platform, style, and LOD requirements. Contact us to discuss details and deadlines. Request a consultation to determine the optimal pipeline for your game.

3D Modeling for Games

An artist delivers a character at 120,000 triangles and insists it “looks better this way.” In a mobile scene, ten such characters tank FPS below playable. We solve this daily: every model passes strict polygon control and per-platform optimization. We have shipped over 50 game projects — from mobile hyper‑casual to PC action titles — and guarantee each asset fits technical constraints without losing visual quality.

3D modeling for games is not “make it look good.” It is prioritising within a limited budget of triangles and draw calls. Every pipeline stage affects runtime performance. Skipping retopology or producing bad UVs forces rework during integration — we eliminate that from the start.

We design models for specific target platforms: mobile uses low texel density (512 px/m), PC up to 2048 px/m. All decisions are backed by engine benchmarks and profiling on real hardware.

How Are Polygons Distributed for Different Platforms?

This aspect is often overlooked at project start — and later causes rework.

Mobile Platforms (iOS / Android)

Mobile is the toughest environment. Character budgets:

Environment objects (single prop):

More critical than polygon count are draw calls and batching. A 200‑triangle model with a separate material can be more expensive than a 2,000‑triangle model that batches with the scene. SRP Batcher and GPU instancing cut draw calls by up to 40% — two to three times more efficient than per‑object rendering — giving headroom for richer assets.

PC / Console

Budgets are looser but not unlimited:

• Main character in a first‑person shooter: 15,000 – 60,000 triangles (enemy constantly visible).
• Character in a third‑person RPG: 8,000 – 25,000.
• Vehicle in a racing game: 30,000 – 80,000 (requires LOD).

Key tool — LOD (Level of Detail). In Unity it is LOD Group, in Unreal — automatic HLOD plus manual LOD levels in Static Mesh Editor. Rule: LOD0 → LOD1 loses ~50–60% polygons, LOD1 → LOD2 another ~50%. Occlusion culling and clustered rendering further reduce GPU load.

VR

VR renders for two eyes at 90 fps (Quest) or 120 fps (PSVR2). Any freeze is physically felt — the player gets dizzy. Character budget: 4,000 – 12,000 triangles (depends on simultaneous NPCs). Environment requires aggressive optimisation: portal system, occluder culling, clustered rendering. We use Frame Debugger and RenderDoc early in the pipeline.

Why Is Retopology an Artistic Task?

Retopology — creating a clean polygon mesh over the sculpt. Many see it as a boring technical step and rely on auto‑retopology (ZRemesher, Instant Meshes, Auto Retopology in Maya). That is a mistake.

Auto‑retopology fails in three critical areas:

1. Joints — knees, elbows, wrists need even edge loops perpendicular to the rotation axis. Auto‑retopology produces chaotic diagonals; during animation the joint collapses unattractively.
2. Face — concentric loops around mouth and eyes are mandatory for correct blend‑shape deformation. Auto‑retopology ignores muscle structure.
3. Silhouette edges — shoulders, chest, large forms. Polygons must be consciously distributed so the silhouette reads even on LOD2.

Manual retopology reduces skinning artifacts by 60% compared to auto‑retopology. Tools: Maya (Quad Draw), Blender (BSurfaces + Shrinkwrap), 3ds Max (Graphite Modeling Tools), ZBrush (ZRemesher with guide curves). Density is uneven: face and hands get more polygons, back and lower legs fewer — because the camera spends more time there. This is not arbitrary; it is driven by silhouette and animation needs.

How Does UV Unwrapping Affect Texture Quality?

UV is another underestimated step. Texel density must be uniform across the entire character. If the boot has twice the density of the face, the texture looks inconsistent. In Maya we use UV Toolkit with normalization; in Blender — the TexTools addon.

UDIM (U‑Dimension) combines multiple UV tiles into one object, allowing 4K–8K textures without single‑UV‑space limits. Used for characters with high detail — cinematic games, hero assets. Pipeline then requires Substance Painter in UDIM mode and separate export per tile.

Baking — transferring high‑poly detail to maps for low‑poly. Tools:

• Marmoset Toolbag — industry standard, cage‑based with fine tuning. According to its documentation, batch baking saves up to 30% time.
• xNormal — free, stable for batch baking.
• Substance Painter — built‑in baker for quick results within texturing, less flexible.
• Blender Cycles — slower, but free.

Typical baked map set: Normal Map, Ambient Occlusion, Curvature, Thickness, Position, World Space Normal. Curvature and Thickness serve as generator masks in Substance Painter for automatic procedural texturing.

What Does 3D Modeling for Games Include?

We also provide one hour of support after delivery — helping with LOD Group setup, collisions, and animation. Source files (.zpr, .blend, .ma, .fbx) are fully transferred.

Common UV mistakes we prevent: different texel densities on adjacent parts, overlapping UVs, insufficient padding between islands, missing UDIM for complex characters.

Reach out for a detailed project assessment. We'll analyze your requirements and recommend the optimal pipeline — from mobile hyper‑casual to PC action and VR. Contact us to discuss your project and receive a tailored commercial proposal.