Latitude Moves Beyond AI Dungeon with Voyage, a Generative Engine for Persistent RPG Worlds
Table of Contents
From Infinite Text to Persistent Worlds
Latitude, the startup that first captured the internet’s imagination in 2019 with the open-ended chaos of AI Dungeon, is attempting to move from a novelty-driven text adventure to a sophisticated game design ecosystem. The company has unveiled Voyage, a platform that shifts the focus from simply playing an AI-generated story to actively architecting AI-powered role-playing games (RPGs).
While AI Dungeon proved that large language models could simulate a believable, if sometimes erratic, narrative, it often struggled with “hallucinations” and a lack of long-term memory. Voyage is designed to solve these structural failures. Rather than relying on a single prompt, the platform utilizes Latitude’s proprietary World Engine—a system five years in the making—to manage the tension between generative freedom and the rigid rules required for a functioning game.
The World Engine manages the heavy lifting of state tracking: remembering that a player burned down a specific village, tracking the inventory of a hidden chest, and maintaining the complex relationship dynamics of non-player characters (NPCs). This persistence means the world reacts logically to player agency, moving away from the generic, repetitive dialogue loops common in traditional RPGs.
The Architecture of Unscripted Agency
For creators, Voyage functions as a low-code environment where AI acts as the lead developer. Users can define the geography of their world, specify the primary antagonists, and establish core mechanics like leveling systems and combat constraints. Once the parameters are set, the AI generates the underlying logic to bring those concepts to life.
The result is a gameplay experience where the “correct” path is entirely subjective. In a traditional game, a combat encounter with a goblin might be limited to a few dialogue choices or a fight-or-flight mechanic. In Voyage, a player could theoretically attempt to negotiate a peace treaty or offer psychological counseling to the goblin, with the AI narrating the outcome based on the character’s established skills and the NPC’s personality profile.
During early testing, the depth of this personality simulation became apparent. One tester reported an encounter with a troll that, instead of attacking, spent the interaction detailing its marital frustrations—a level of emergent behavior that goes beyond simple text generation and into the realm of simulated character depth.
Strategic Backing and Technical Integration
The launch comes alongside a strategic partnership with Google’s AI Futures Fund. To achieve the multimodal experience Voyage promises—which includes audio narration and visual generation—Latitude is integrating a mix of proprietary models and third-party tech. Specifically, the platform leverages Google’s Gemini Flash for image generation and Gemma for text, audio, and video processing.
The company is also strengthening its industry ties with the addition of former Roblox Chief Business Officer Craig Donato to its board. This move suggests a clear ambition: Latitude isn’t just building a game, but a platform for user-generated content (UGC) that could rival the scale of Roblox, but with the creative flexibility of generative AI.
Market Positioning and Access
Voyage is currently in expanded beta, with data showing that early testers have already interacted with over 160,000 unique AI characters. The platform is free to play for now, but Latitude intends to introduce a tiered subscription model. Plans will be priced at $15, $30, and $50 per month, providing users with advanced AI capabilities and removing limits on the number of actions they can take within their worlds.
CEO and co-founder Nick Walton noted that while AI Dungeon was the “initial promise” of generative AI in gaming, Voyage is the execution of that promise at scale. By adding deterministic systems—rules that the AI cannot ignore—Latitude is attempting to bridge the gap between the unpredictability of a chatbot and the reliability of a commercial video game.
An open beta is scheduled for later this year, with safety filters and parental controls implemented to manage the mature content that naturally emerges from unscripted, user-driven narratives.
