Check Out Scrypted’s AutoRPG Tech In Action

Tim Cottenformer Ultima Online developer and Citadel Studios co-founder, now heading up a startup game development company called Scrypted — just uploaded a video to YouTube showcasing his company’s newest technology: a Python script called AutoRPG that (as the name implies) generates an RPG level (with points of interest and NPC stories) using a variety of AI-backed “Autonomous Task Agents”.

Per the video description:

We developed AutoRPG as an in-house tool to speed up the development of our own game: Niftiez. By leveraging the power of Yohei Nakajima’s BabyAGI project, as well as MiniGPT-4, Pinecone, and over 20 years of game development expertise, we’ve made something that goes well beyond procedural generation.

Welcome to the world of Autonomous Task Agents.

You’ll witness the AI making logical decisions about world-building based on how things work in the real world and how time passes. You’ll also see it generate various objects and elements that it may or may not use as it tells a compelling story.

While AutoRPG is only an early prototype and is not yet ready for commercial release, we believe that it has incredible potential for both game developers and user generated content (UGC). One of the most exciting aspects of this technology is that it doesn’t require the AI to draw art, but rather uses the default art tileset from the RPG Maker engine (with a lot of embedding information for the AI to understand what the tiles mean to each other!).

The system is modular, and our next demo will be in 3D!

The video is pretty impressive to watch; AI decisions are echoed to screen as the scene evolves in front of you, and the end result is something that doesn’t just add an area to a world, but makes the world feel lived-in, as though it and the people that live there have a history and a story to tell.

This is currently the only video on the Scrypted YouTube channel, but consider following it anyway; it’s likely more such demonstrations will be published in the near future.