๐๏ธInitial Vision 2023: Text-To-OpenWorld Engine
We are a startup working on AI-powered game engine. Here are what you can expect in a long-term from RPGGO.
One-sentence
As a technology firm, we are leveraging generative AI to build the next generation text-to-openworld engine, enabling players to live an adventure with a new form of online interactive experiences.
Text to Open World
Existing open-world products are abundant, whether they are metaverse social products or pixelated games like Roblox. However, these products face a challenge: they struggle to evolve into universal UGC (User-Generated Content) platforms. This is because their world guide book and asset styles are official, but not open, diverse, or conducive to UGC, limiting them to vertical user bases. While these bases are highly engaged, this engagement doesnโt easily generalize across different user groups. The advent of generative AI offers a viable solution to these traditional shortcomings in open-world games. It fills the gaps in world-building and creative freedom for assets.
From a functional standpoint, generative AI brings transformative features:
Text to Story: Real-time logical reasoning for text generation allows AI to create world guide books.
Text to Assets: Real-time graphic rendering translates text into visuals, personalized according to user-defined styles, scenes, and elements. While 3D generation still has a long way to go, 2D is already feasible.
Behind these features, the most significant change may be in the production relationships,
shifting creative threshold downward.
This enables an open world based on UGC, personalized construction, and decentralized governance.
Looking ahead to the next generation of open worlds, without getting into implementation details, they should at least include the following elements:
World Guide Book, including storylines, characters, world objectives, and core conflicts.
Personalized NPCs, including customized voices, behaviors, language styles, appearances, and clothing.
Visual Rendering, including 2D and 3D forms of characters, scenes, props, and materials.
Consumer-Level AI Infrastructure, reducing the daily cost per user to under $1 for scalability.
Decentralized Finance, including payments, asset trading, and financial systems.
These elements #1๏ผ #2๏ผ #3๏ผ #4 form the basic building blocks of Text-to-Open World.
So, whatโs different in user behavior and consumer mindset in this new generation of open worlds?
Imagine a game experience tailored just for you, where the entire world unfolds before you, and your actions in the game influence the storyline in real-time, NPCs remember your choices and react accordingly
Assets are customized to your imagination, not limited by official constraints. The assets you create can be traded or sold, not just within one game but across an entire universe of interconnected experiences.
The gaming experience is infinite and non-repetitive, with AI balancing individual and collective harmonization.
The game itself becomes consumable UGC content, shareable and open to secondary creations.
Immersive integration and autonomous NPC dialogues transport you into a โReady Player Oneโ type of alternate world for chatting and interaction.โ
This is the future we envision, a future where the lines between gaming, social interaction, and content creation are blurred, thanks to the power and potential of generative AI.
In this new paradigm, the game is not just a game; itโs a living, breathing ecosystem. Itโs a platform for creativity, social interaction, and even economic activity. Weโre not just talking about a technological shift; We believe it is a cultural revolution.
What we are trying to do
In our startup, we aim to address these gaps by leveraging advanced generative AI to build the next Text-To-OpenWorld Engine and power the ability of UGC to create and play in his/her open world. We believe that the future of gaming and interactive products lies in the seamless integration of generative AI, and weโre excited to lead the way.
For the Zagii engine, the core features include but are not limited to:
Fine-tuning based on open-source models with annotated RPG data to teach the AI how to create RPG games.
Multi-LLM-model ensemble architecture, or what we call โllm-augmented autonomous agents architecture.โ Different large models take on different tasks and roles based on their unique capabilities during the generation process.
RAG (Retrieval Augmented Generation), which utilizes a search framework to employ an external knowledge graph and meta-documents as search index, building prompts by adding the relevant retrieved data in context.
Fast visual rendering: customized stable diffusion needs to generate an image within 2 seconds. *Among these, the annotation of RPG data and fine-tuning of open-source models will require significant resource and computational power investments.
This is the technological foundation weโre laying, and itโs not just ambitious; itโs feasible. Weโre not just dreaming big; weโre executing a scalable plan to make those dreams a reality. Achieving the ultimate form all at once is unrealistic. There will be an iterative path. We will start by supporting the creation of one-dimensional, small-scale open worldsโspecifically, โmurder mysteryโ scenariosโas a stepping stone to gradually build our AI-OpenWorld capabilities.
But please note that we are NOT building one RPG game, but building an AI-Engine to generate RPG games, to generate mini-worlds.
Creation Tooling
Ecosystem
In order to better utilize the AI Engine capability with 3rd party partners, we open the public API to selected partners for real-time game scenarios rendering.
And, Welcome to the Public API! ๐ผ๐๐ป
What's in the API
Last updated
