Maybe a poor idea, but what about letting AI help with the ww2online transition to UE5?

[Augetout]

I won't pretend to be at the technical skill levels of the Rats, but asking the right questions of AI seems to generate useful answers.

To wit, I asked CoPilot 'how would AI recreate the ww2online map in UE5. The answer:

 

To get from “WW2 Online’s map” to a playable UE5 level, you’d basically use AI as a turbocharged assistant for terrain, assets, and layout—while you stay in charge of accuracy and feel.

01

 

 

 

Gather and analyze reference data

Try First

You need a clear, measurable picture of the WW2 Online map before any AI or UE5 work.

Use external map tools or GIS data if the game map corresponds to real European terrain.

• Capture references: in‑game screenshots, minimaps, sector maps, any top‑down views

• Collect real‑world data if the map is based on actual geography (DEM heightmaps, OpenStreetMap roads)

• Mark key locations: towns, rivers, bridges, front lines, spawn areas

• Decide scale: how many kilometers of the original map you want to recreate in UE5

 

02

 

 

 

Generate base terrain with AI tools

AI can convert real‑world elevation or reference images into a UE5‑ready heightmap.

Follow your chosen AI terrain tool’s export guide to produce a UE‑compatible heightmap.

• For real locations, use an AI terrain tool that imports elevation data and satellite textures (e.g., heightmap_from_location‑style workflows)

• For fictional or stylized areas, convert the game’s map image to a heightmap via image‑to‑heightmap tools

• Tune presets (mountains, hills, plains, rivers) to match the strategic feel of WW2 Online

• Export the result as a standard heightmap (PNG/RAW) for UE5

 

03

 

 

 

Import terrain into UE5 Landscape

Core Step

You turn the AI‑generated heightmap into an actual playable landscape in Unreal.

Unreal Editor → Select Mode → Landscape → Import from File

• In UE5, open Landscape Mode and create a new landscape

• Choose Import from file and select your heightmap

• Match the overall size (km²) and Z scale to your design

• Apply basic landscape materials (grass, soil, rock) as placeholders

 

04

 

 

 

Block out towns, roads, and strategic points with AI assistance

AI can help place and iterate on large numbers of actors while you control layout and realism.

Connect your AI assistant to UE5 editor via an MCP‑compatible plugin, then issue scene‑building commands.

• Use an AI‑connected workflow (MCP‑style tools) to spawn buildings, roads, and props via natural language commands

• Define town centers, road networks, rail lines, bridges based on your WW2 Online references

• Ask AI to cluster houses, leave squares, and align roads to terrain

• Keep a top‑down view open to compare against the original map’s structure

 

05

 

 

 

Create and refine materials and textures with AI

AI can generate or suggest materials that match European WW2 environments.

Unreal Editor → Open Material Editor → Create materials and assign to meshes/landscape layers

• Use AI image generation to create tiling textures for cobblestone, mud roads, brick, plaster, etc.

• Import textures into UE5 and build Material Graphs (with AI‑suggested node setups if available)

• Apply different materials to landscape layers (fields, forests, roads) and building meshes

• Iterate until the color palette and wear match the WW2 aesthetic you’re aiming for

 

06

 

 

 

Populate environment procedurally (foliage, props, clutter)

AI‑driven scatter tools can quickly fill large areas with believable detail.

Landscape/Foliage Mode → Configure foliage types → Paint or procedurally scatter across terrain.

• Use procedural placement or AI‑assisted scatter tools for trees, bushes, rocks, fences

• Define biomes: dense forest, open fields, riverbanks, urban outskirts

• Add war‑specific props: sandbags, trenches, barbed wire, destroyed vehicles

• Keep performance in mind—use foliage instances and LODs for large‑scale maps

 

07

 

 

 

Set lighting, atmosphere, and time of day

Lighting sells the mood—dawn offensives, foggy mornings, or overcast battlefields.

Place lighting actors in the level and tweak via the Details panel.

• Add a Directional Light, Sky Atmosphere, and Volumetric Fog for the desired time of day

• Use AI suggestions for color grading and post‑processing to match historical photos

• Test visibility from key positions (sniper nests, artillery spots, open fields)

• Adjust shadows and fog density to keep gameplay readable but tense

 

08

 

 

Optimize and segment the world for gameplay

Performance

Large WW2‑style maps need streaming and optimization to stay playable.

World Settings → World Partition / Data Layers; then use profiling tools to optimize.

• Use World Partition or level streaming to break the map into cells

• Set up Data Layers for different detail levels or scenario variants

• Profile performance (CPU/GPU, draw calls) and reduce overdraw and foliage density where needed

• Add collision, navmesh, and gameplay volumes (spawns, capture zones) to make the map functional

 

Sources:

In practice, AI won’t magically “import WW2 Online” for you—it accelerates the grunt work: terrain generation, mass asset placement, texture ideas, and iteration. You still decide scale, authenticity, and gameplay flow.

If you tell me which specific theater or sector of WW2 Online you care about, I can help you sketch a more exact layout plan.

 

I am just thinking out loud.

[aismov]

@madrebelhas been doing some of this and the problem as with any AI out there is always the details and edge cases. You can get it to "make a map" but "make an accurate map of circa1940 France" and it will choke.

Ford recently re-hired a lot of their QA engineers precisely for this reason. AI companies promised that AI can do their prod QA for them. So they fired a bunch of ppl, put in AI, which then promptly proceeded to screw things up. CEO said in rather diplomatic terms that the AI didn't have a good enough data set to give them the QA they needed. That is a fancy way of saying that AI doesn't do what the companies are hyping: taking a generalized data set and using to achieve a specific outcome.

At this point with my own experience with it, I'm wondering if even the "frontier" AI (i.e. the good stuff they are cooking up just now but isn't public) is just something to keep the hype train going and pumping more hot air in the balloon. The synetheic benchmarks they use to measure AI error is nowhere what is happening int he real world. Now the next hype is agentic AI so they can get all the companies that shelled out cash for LLMs to now come back to the trough for a second time to get their agents as well. we'll see...