Blender Asset Tool

Inspiration

In the time between finishing my A-Level Diploma and starting university, I wanted to keep improving my programming skills. So I started learning Unreal Engine by building my own game. At some point I needed assets. Since the game is mainly a hobby project and may never be released, I wanted to use free assets. The problem was that they all looked different and never really fit together. I asked ChatGPT for ideas. The obvious answer was to create everything myself in Blender. But creating every asset manually would have meant a lot of repetitive work. Instead, another idea came up: build a Blender tool that generates the assets for me. That was never supposed to become its own project. My only goal was to create consistent assets for my game. But after the first working prototype I realized that the tool itself had become an interesting project, so I decided to continue developing it as a standalone Blender add-on.

What it does

Blender Asset Tool is a Blender add-on for creating reproducible game assets through configurable procedural generators.

Instead of manually repeating the same workflow for every asset, the tool provides generators, style presets, material roles, material profiles, and a controlled export pipeline for Unreal Engine. Assets can be generated, regenerated, modified, and exported with consistent naming, materials, pivots, collision data, and export metadata.

The current implementation includes several procedural street lamp variants as production-ready example assets, while the underlying system is designed to support many more asset types.

How I built it

As a solo developer, I designed the workflow, defined the requirements, reviewed every implementation and used GPT-5.6 and Codex as development partners throughout the project.

GPT-5.6 helped me plan the architecture, break larger problems into smaller tasks, and review design decisions. Codex implemented individual features, which I continuously tested, reviewed, and refined inside Blender until they worked as intended.

Challenges I ran into

The biggest challenge was not writing code but designing a workflow that stays maintainable while supporting different asset types

Another challenge was translating the asset designs from my ideas into Blender while keeping them procedural, configurable, and reproducible instead of creating one-off models.

Accomplishments that I‘m proud of

The project went from an idea to a working Blender add-on in only a few days.

After building the core architecture, adding new assets became significantly easier. Instead of starting from scratch every time, I can now extend the tool with new generators using a short Codex prompt together with a reference image.

What I learned

• AI becomes much more useful when the requirements are precise
• Breaking large problems into many small tasks produces much better results
• GPT-5.6 is especially useful for planning and reviewing architecture, while Codex works best on focused implementation tasks
• Building the right system first saves much more time than optimizing individual assets

What's next for Blender Asset Tool

• Finish the core framework so additional asset types can be added quickly through the existing generator system
• Expand the library with many more procedural game assets
• Continue improving the Unreal Engine export pipeline and asset consistency
• If I decide to release the tool publicly in the future, I would like to make adding new assets possible without relying on Codex

Built With

• codex
• gpt-5.6
• python
• sol