Wisp World - Interactive AR RPG for Spectacles

Tools: Parabrain, Lens Studio, Snap Spectacles, LLM Integration

Wisp World is an interactive RPG that transforms your real environment into a magical story space using Snap Spectacles. Players discover three lost wisps scattered throughout their home and help reunite them by solving conversation and object-finding challenges. The entire narrative unfolds in AR, blending digital characters with physical spaces to create an entirely new form of storytelling.

Wisp World AR experience overview

I designed the personalities and conversation systems for three distinct wisp characters using Parabrain. Starting with research into character dynamics and narrative tropes, I initially developed four characters before narrowing to the optimal trio. Each wisp required extensive character profiling - background lore, sample dialogue, thought processes, and personality quirks.


The iterative process was intense: Tictac became the charming guide, Grindle the grumpy one wary of humans, and Gliss the hyperactive optimist. Conversation design proved especially challenging since most AI defaults to servile, flowery language. I had to build intuition for prompt engineering - testing different LLMs (Gemini vs ChatGPT) for specific tasks and discovering that a single misplaced word could break an entire character's personality. Each character went through dozens of prompt iterations to achieve authentic, consistent voices.

Character development and personality design

I built complex state management systems that tracked player progress through conversation and object interactions. Each wisp had multiple conversation states with specific triggers - for example, Tictac couldn't reveal their backstory until players proved they weren't harmful. This required mapping dozens of game states and defining what elements were active/inactive at each stage.

Game state progression diagram

Since most users had never tried AR experiences, I designed minimal onboarding that taught through play rather than lengthy tutorials. User testing revealed critical issues - like gesture recognition failing when fingers were too close together - that required iterating on both the technical systems and instructional design. The final experience guides users naturally through Tictac's dialogue rather than overwhelming them with upfront instruction.

AR UX testing and gesture design