ETIENNE JULIEN - PORTFOLIO

In this project, I developed a Minecraft mod capable of dynamically generating custom items using data from several neural network. The goal was to allow players to create unique items, enriched with personalized data, without having to pre-register them in the game. This technical challenge allowed me to create a java algorithm for create a Minecraft mod

The algorithm works as follows:

• The player inputs two items through an in-game command (later, a more integrated system will be developed).
• A Java script sends the request to a Python script that hosts both an LMM model and a GAN model. The LMM model generates a name for the new item and, based on this name, determines its category (weapon, armor, food, etc.). Then, the GAN model uses the name along with several prompts to generate a sprite that visually fits the item.
• All outputs from the Python script are sent via PHP to a specific page on my portfolio website. The Java mod then retrieves these elements and reintegrates them into the game as a newly generated item.

This project allowed me to discover Java, particularly with the Fabric modding framework for Minecraft. It also taught me how to deploy and interact with Python scripts over the internet, managing data flow between different technologies and ensuring proper communication between game client and Server.

One of the biggest challenges was managing dynamic content integration within Minecraft, which is naturally limited by its static item registration system. As of now, I have not fully completed the integration back into the game, but the entire pipeline from taking two input items, generating a name and a sprite, and creating a final new item outside of Minecraft is fully functional. You can find the github projet of Infitems at the following link : https://github.com/Crodaliox/MinecraftItemGenerator