Researching AI tools for Level Design

Project SplitWar began as an experimental project to explore how artificial intelligence tools can assist game developers, especially Level Designers, in their workflow.
The main goal was not to create a full game, but to analyze which parts of the development process AI genuinely helps with, and which parts still require human creativity, intuition, and design sense.

PS: This Blog post was written on 1 November 2025. So the content might not be accurate (AI is evolving every day).

Objective

Research Tools

My first step was to dive deep into the world of AI tools that can help game developers. I explored how tools like ChatGPT, Google Gemni, Gamma AI Meshy, NEO AI and Cursor can speed up documentation and coding tasks, while procedural tools like Houdini and Unreal’s Behavior Trees and EQS can shape smarter, dynamic gameplay systems. This phase helped me understand what AI can and can’t do for designers, especially in creative, spatial thinking.

Writing a GDD

Once I had my tools ready, I moved on to building the Game Design Document. I used Project SplitWar as the base, a concept where time dilation and dual-reality mechanics drive the gameplay. The goal was to design a game that uses AI in developmen This stage taught me how AI can assist in organizing thoughts, improving clarity, and

Prototyping

With the GDD done, I jumped into Unreal Engine and started building the first prototype. Using Cursor in Unreal was a nightmare but it helped me for smart coding assistance, I streamlined scripting and debug workflows w This helped me test how practical AI support really is in an actual production environment from idea to implementation.

Level Design

Finally, I analyzed where AI truly fits into my role as a Level Designer. While AI tools can speed up documentation, automate testing, or even suggest layout ideas, the actual design, the flow, pacing, and player experience, still relies on human judgment. That became my biggest takeaway: AI can assist, but not replace, the designer’s creative vision. At least not yet

Tool Research

Below are some of the tools I experimented, to understan dthe capabilities AI in Game Development

ChatGPT (a.k.a. Chief)

ChatGPT has been assisting people in everything from nutrition to relationships, but I wanted to see if it could actually help in game design. So, I created a fresh account, no prior context, no bias, and started from scratch. I shared my game idea, and within seconds, it gave me a whole list of AI tools to try out. From documenting ideas to suggesting level layouts, ChatGPT proved to be a great creative assistant and an amazing brainstorming partner.

Pros:

Great for ideation and refining thoughts
Speeds up documentation and design structure
Easy to bounce ideas off and get quick clarity

Cons:

Can’t create entirely new, original ideas on its own
Needs a solid foundation or spark from the designer first

Overall, it’s definitely here to stay, not replacing designers, but supercharging them.

This the image Chat GPT generated when prompted "if you have to represent yourself in an image what would it be (Generate image to showcase)"

Google Gemini

Gemini, the bread and butter for quick concept art... or is it? It struggles with accuracy (can’t tell left from right sometimes), which makes it frustrating to use. However, being completely free and not credit-based gives it a big advantage. It helped me turn my hand-drawn sketches into clear, beautiful concept images. For instance, I asked it to create an image of sci-fi soldiers in snow with war bags, and it absolutely nailed it.

Pros:

Free and accessible
Great for rough concept visualization
Converts messy sketches into usable art

Cons:

Inconsistent results
Weak control over small details (especially direction/orientation)

Leonardo AI might outperform it, but Gemini’s no-credit system keeps it valuable for quick visualization.

Asked Gemni to envision a boss battle with a time-dilated mutated enemy

Asked Gemni to envision a split time frame of the player character, as you can see on the left-hand side, it's the past, and the right side is the future

I wanted a direct comparison of the past and the future to give me a sense of what the blockout should look like. So this image helped me a bunch to visually see how my game should look.

It was a nightmare to remove the Sword from the left hand as GEMNI does not know the difference between left and right

Gamma AI

This one’s a hidden gem. Gamma AI takes rough drafts, even simple ChatGPT prompts, and turns them into complete, clean presentations. It’s not the most user-friendly tool, but the output quality is chef’s kiss. My workflow is to let Gamma build the initial layout, then tweak it in PowerPoint to my liking. I can see myself using this for game pitch decks, class presentations, and even internal documentation.

Pros:

Excellent presentation generator
Saves massive time in formatting and structure
Great for quick idea sharing

Cons:

Limited editing tools
Works best when paired with manual tweaks later

Cursor

The GOAT. This tool genuinely feels like the future. Cursor can create fully functional code, from HTML to Unreal blueprints, almost instantly. The credit system is generous, and it works shockingly well for prototyping. The only problem? It’s not Unreal-friendly yet, and setting it up was a pain. But once I figured it out, I had working prototypes in seconds.

My workflow: tell ChatGPT my idea → it writes a Cursor prompt → paste it → boom, functional code.

Pros:

Super fast coding assistant
Great for solo devs and rapid prototyping
Smooth integration with AI prompts

Cons:

Unreal setup is rough
Still early for complex engine tasks

If Cursor ever integrates natively with Unreal, it’s game over (in the best way).

Cursor WorkFlow

Meshy AI

A paid tool I didn’t get to dive deep into, but from what I saw, it’s crazy powerful. Meshy generates fully rendered 3D models in seconds. Need a cracked brick wall? Multiple variations? Picture-to-model conversion? It handles it all. I’d love to use it more if the budget allows. Perfect for solo devs who want art quality without hiring a full team.

Pros:

Instant 3D model generation
High-quality results
Great variety and realism

Cons:

Paid tool, not accessible to everyone
Limited hands-on experience (for now)

Meshy WorkFlow

NEO AI & Ludus AI

These are upcoming Unreal-specific plugins designed to help automate blueprints and fix issues. Still in early development, but I can see the potential. They can already create blueprint snippets in seconds and are getting better at assisting with logic setup. Definitely worth keeping an eye on.

Pros:

Built for Unreal users

Fast and promising for automation

Cons:

Still experimental

Limited real-world usability (for now)

Dash AI
Dash is more of a worldbuilding companion. It helps with setting, lore, and environmental storytelling, which is useful when planning the tone and pacing of a level. It’s not perfect, but combining Dash with ChatGPT gives surprisingly deep world concepts.

Pros:

Great for lore and narrative setup
Complements creative writing tools

Cons:

More of a side tool than a core workflow piece

Writing a GDD

Ideation and Foundation

When I first shared my concept for Project SplitWar with ChatGPT, it immediately helped me expand on it, not just narratively, but also mechanically.
It suggested ways to blend time dilation into gameplay, proposed themes around “split realities,” and even recommended how to pace the player’s experience through mission-based flow.
The brainstorming felt natural — like talking to another designer who never gets tired of throwing ideas back and forth.

Key Takeaway:
ChatGPT helped me go from a raw concept to a structured game idea, making it much easier to visualize the tone, goals, and core loop of Project SplitWar.

Project Split War - GDD

Prototyping VS Level Design
using AI Tools

Project Split War - GDD

Time DIlation

Concept:
Slows down (or speeds up) the flow of time to give the player more control or add dramatic emphasis.

How it works (in-engine):

Controlled via Unreal’s Global Time Dilation or Custom Time Dilation on specific actors.
When activated, the time scale drops below 1.0 (e.g., 0.3 = 30% speed).
Player input and animation stay synced, but everything else feels “in slow motion.”
Often temporary, returns to normal after a timer or player action.

Gameplay Use:

Dodge bullets, pass through closing gates, or react quickly in puzzles.
Create cinematic time-slow moments when switching between Past/Future.

Flicker Zone

Concept:
A localized area where time auto-switches between Past and Future, independent of player control.

How it works (in-engine):

Implemented as a Trigger Volume (your ATimedFlickerActor).
When the player enters:
- Their manual time-switch input (T) is disabled.
- The zone automatically cycles between timelines (Past ↔ Future) based on a configurable sequence (like every 3 seconds).
- Optional post-process (glitch effect or color shift) indicates flicker activity.
When the player exits, control is restored and normal time-shifting resumes.

Gameplay Use:

Environmental storytelling (unstable timeline zones).
Challenge areas where players must time their movement with automatic timeline swaps.

Reverse Time

Concept:
Lets the player rewind objects (or themselves) a few seconds into the past, reverting their state to how they were before.

How it works (in-engine):

Each object that can rewind has a TimeRewindComponent.
This component continuously records its transform, velocity, and other properties over time (using an array of snapshots).
When the rewind is triggered, the component starts stepping backward through those stored states, smoothly interpolating position and rotation.
Rewinding can be global (everything rewinds) or localized (only specific actors).

Gameplay Use:

Undo mistakes (like moving platforms or fallen boxes).
Solve puzzles by returning an object to a previous position.

This is a simple Turret, the first code Cursor wrote for me. The turret only works in the futur eso the player must travel to past to avoid it.

Yes, all of this can be achieved using blueprints, but Cursor was able to complete all of this in seconds. This is a great workflow for non-technical designers to communicate their ideas easily

Reshaping Level Design WIth AI

Throughout this project, AI tools like ChatGPT and Gemini became creative partners in my design workflow.

ChatGPT helped me refine puzzle mechanics, enemy encounters, and spatial flow for Project Split War. The process was iterative, I’d describe gameplay intentions or player abilities, and ChatGPT would suggest new layout ideas, pacing adjustments, or environmental storytelling opportunities.

It kept the momentum going when I got stuck, helping me visualize how gameplay, tension, and rhythm could evolve across spaces. Gemini, on the other hand, was my visual jumpstart.

As a designer, I often face what I call the blank canvas syndrome, those early moments where you know what you want to feel, but can’t quite see it yet.

Instead of scrolling endlessly through Pinterest boards for the right tone or setting, Gemini generated quick concept art references that instantly aligned with my ideas. That immediate feedback loop made ideation faster and more inspired.

Together, these tools became an extension of my process. ChatGPT sharpened my logic and design thinking, while Gemini filled my visual imagination. They didn’t replace creativity, they removed the friction around it, letting me focus on what really matters: designing spaces that tell stories through play.

Conclusion

This entire experience reshaped the way I see game design. I’ve always believed creativity and technical skill go hand in hand, but after exploring AI tools, I’ve realized there’s now a third hand in the mix: intelligent assistance.

From ChatGPT helping me brainstorm and document my ideas, to Cursor generating prototypes in seconds, AI became an invisible teammate that accelerated my workflow and boosted my confidence as a designer.

However, I also learned that AI can’t replicate the soul of design. It can’t feel pacing, emotion, or narrative flow the way a human can. It’s powerful for prototyping and iteration, but when it comes to crafting meaningful experiences, the spark still has to come from me.

Overall, this project taught me how to balance innovation with intuition. AI isn’t here to replace designers, it’s here to amplify us. As a level designer, I now see AI not as a shortcut, but as a creative amplifier that frees me to focus on what really matters: designing worlds that make players feel.

Back