How to Generate 3D Game Assets Using AI and Blender

Generative AI is reshaping creative workflows, and 3D asset creation is the next frontier. While text-to-3D still lags behind 2D image generation in practical applicability, new workflows are making it increasingly viable—especially for stylized, low-fidelity aesthetics like PS1-era graphics.

A new guide from Hugging Face walks developers through a seven-step process for generating usable 3D assets with minimal friction.

The Workflow#

Generate a base model using OpenAI's open-source Shap-E diffusion model, which takes text prompts and outputs 3D geometry.

Optimize in Blender by importing the generated GLTF file and using the Decimate modifier to reduce polygon counts—crucial for game engines and performance-constrained applications.

Add textures via Dream Textures, a Blender addon that leverages stable diffusion to generate seamless, tileable textures directly within the 3D editor.

UV map and export the finished asset as FBX, ready for game engines like Unity.

Why PS1 Style?#

The retro PlayStation 1 aesthetic is intentionally forgiving. Its characteristic low polygon counts and flat shading align naturally with the current limitations of text-to-3D models, allowing creators to ship usable assets without extensive manual refinement.

The guide notes that as these generative models improve, the same techniques could scale to higher-fidelity or photorealistic styles—opening possibilities for procedurally generated worlds at scale.