🔒 Take Me to the Moon

This case study is password protected.

← Back to Projects

Take Me to the Moon — Designing an AI-Directed Story in Immersive VR

When extreme climate change forces humanity to flee Earth for the Moon, what choices get you there? An AI agent rewrites the story around every decision, while Visko renders each scene as live video inside a Meta Quest 3 experience.

Winner
Best Storytelling · Visko Sponsor Prize
51
Teams
182
Hackers
267
Attendees
30-second demo · captured in Meta Quest 3

Concept

Inspired by Immerse the Bay 2026's theme, "Take Me to the Moon," we imagined a journey from a Stanford dorm all the way to the Moon. Extreme heat has made Earth unlivable, and you have to escape your dorm and get to the Moon as fast as you can.

Most escape games ask whether you survive. This one asks how.

Never the same journey twice

AI writes the choices. An AI story director, guided by our prompts, writes each scene from your earlier decisions, so even at the same checkpoint the options you face are different every time.

Visko renders the result. The live video comes from Visko.ai and is generated from the option you pick, so what you see on screen changes with every choice you make.

Along the way, interactive moments put you physically in the action, like grabbing items before you flee and throwing them to fuel the rocket.

At each checkpoint, your choices and mission results raise or lower your wellness points, and your final score decides whether you're welcomed as a Moon Citizen. At the end, you get a report revealing what kind of traveler you turned out to be!

Award

Winner, Best Storytelling (Visko Prize) — Immerse the Bay 2026

Team

Yoomi Cha, Shreyaa Dani, Hyun-suk Bang, Cheyoung Ahn

Tools

Figma, Unity, Meta Quest 3, OpenAI API, Visko, Skybox, Claude

Role

Project Lead · UX / VR UI Designer

Timeline

Nov 2026 · Hackathon weekend

Award & Team

Best Storytelling, Visko Living Worlds Track


Team Take Me to the Moon on stage receiving the Visko Best Storytelling award
On stage at Immerse the Bay 2026: Visko Living Worlds Track, Best Storytelling
The four-person team at their hackathon table with a Meta Quest headset
Our team of four: Yoomi Cha, Shreyaa Dani, Hyun-suk Bang, Cheyoung Ahn

View the project on Devpost →

System

One real-time loop, from choice to live video


The headset calls OpenAI directly. The AI acts as a story director, returning structured data every turn — the scene, two choices, their wellness costs, and video prompts. Those prompts go to Visko, and the generated video streams back into Unity. Carrying this loop through an entire playthrough was the hardest part of the build.

Built for a standalone headset. Everything runs on Meta Quest 3, with no PC or server kept running during play.

One story controller holds the inventory, choices, and wellness score, so the UI, the physical objects, and the video always agree on where the player is.

Meta Quest 3 headset with Touch Plus controllers
01

Player choice

The player picks one of two options inside the VR panel.

02

AI story director

OpenAI writes the next scene, two choices, and wellness costs from the full history.

03

Visko generation

Video prompts are turned into a live, generated scene.

04

Unity playback

Video streams onto a curved screen wrapped in a 360° skybox.

Live generated video on a curved screen based on the player's decision
Displaying a live screen based on your decision

Fixed where it sets the tone

  • The opening emergency broadcast is prerecorded, so every player starts with the same urgent, cinematic hook.
  • The success and failure endings are prerecorded, so the verdict always lands cleanly.

Live where choices matter

  • Every story scene between the dorm and the Moon is written and rendered live from the player's history.
  • The lunar arrival is generated too, so the ending reflects the journey that led to it.

My Role

Game flow, VR UI & demo production


I started the project and led a team of four. I designed the game flow, the UI, and its components in Figma, and created the background and visual assets the team built from. Our developer translated those designs into readable VR panels in Unity. I also produced the showcase video, capturing in-game footage and editing it with VLLO and Figma.

Design

  • Mapping the six-checkpoint game flow
  • Designing VR UI panels and reusable components in Figma
  • Creating background and visual design assets
  • Defining panel size, distance, tilt & contrast for VR
  • Building layouts that survive AI text of any length
  • Designing the wellness bar and choice feedback

Story & Production

  • Storyboarding scenarios on a whiteboard with the team
  • Tuning the wellness score through live testing
  • Capturing in-game footage
  • Editing the demo video (VLLO, Figma)
Choice card component states in Figma

Choice cards

Green and blue variants with default, hover, and selected states.

Wellness change popup component

Wellness feedback

The +5 / −5 popup and the wellness bar's good, warning, and critical states.

Launch site fuel mini-game UI with target, gauge, and throws left

Mini-game UI

Target, fuel gauge, and throws-left counter for the launch pad.

Design Approach

Theme → Fixed Storyboard → Self-Writing Story


Starting point: this year's Immerse the Bay theme, "Take Me to the Moon." We imagined the journey beginning somewhere every player could picture — a Stanford dorm room — and focused on storytelling so anyone could pick it up and see themselves in it.

Theme

Turn a fixed destination into a question about character: not if you escape, but who you become on the way.

Fixed storyboard

Lock six locations and leave the events at each one open, so the AI has guardrails without losing variety.

Self-writing story

Let the AI write each scene in real time, remembering earlier choices so the journey feels connected.

Designing the AI's Rules

Content rules that keep live writing readable in a headset


The AI writes every line the player reads, so the copy had to be designed as a system, not as individual screens. We gave the story director a set of rules that keep each scene short enough to read at a glance over moving video, and every choice hard enough to matter.

Choices

Exactly two

Every decision is a real dilemma between two grabbable objects, never something small like left or right.

Labels

Verb first, ≤ 6 words

Labels address the player as "you" and start with an action, so they read instantly in VR.

Narration

≤ 20 + 30 words

One sentence for the outcome, then two for the next scene: what's in front of you, then the problem.

Pacing

One event per scene

Each scene has one thing happening and at most one new character, so the story never outruns the player.

The tempting choice

Every decision has one good and one bad option, and the player isn't told which is which. The bad option always has to be tempting: safer, faster, or more loyal to someone the player cares about. Helping someone costs the player something, and helping yourself costs someone else.

The Journey

Six checkpoints, endless playthroughs


You get 15 seconds to grab three of five items from your room and place them in your inventory tray. Those items come back at the launch pad, where you throw them to fuel the rocket. The locations are fixed; what happens at each one is written live, and people you help or abandon can return later.

Selecting three items in the dorm room
01

The dorm room

Wake your roommate, or slip out alone.

Selecting an option at Memorial Church
02

Memorial Church

On a burning campus: steal a motorbike, run the Caltrain tracks, or flag down an ambulance.

Throwing items at the target to fuel the rocket
03

The launch pad

Fuel the rocket by hand, throwing your items into the tank.

Low orbit decision
04

Low orbit

The rocket is too heavy. Cut loose the strangers on your ladder, or pull them in and risk everything.

Approach to the Moon
05

The approach

Your roommate returns to the story — beside you, or calling from Earth.

The game result on the Moon
06

The Moon

Your choices are judged at the colony gate.

Design Challenges & Solutions

Seven friction points, seven design decisions

Challenge 1

Designing spatial UI for the first time

The friction: this was my first VR hackathon, and I had never designed spatial UI. Flat-screen rules broke in the headset: text over constantly changing live video became hard to read.

Design strategy

Before the event, I studied Meta's Quest design guidelines and Apple's Vision Pro Human Interface Guidelines. During the build, I applied them to rethink panel size, viewing distance, tilt, and contrast for the headset.

Outcome

I could make spatial layout decisions with confidence from day one, and every scene kept a consistent look, feel, and navigation, readable no matter what video played behind it.

Intro warning screen in VR

Challenge 2

Keeping up with a story that kept changing

The friction: the storyline shifted constantly as the team iterated, and every change meant new components and assets, often within minutes, under tight time pressure.

Design strategy

I used AI throughout my workflow to generate and adapt components and assets quickly. When I found out late that the video storyboard had to be vertical, not the wide format I had started with, AI let me update every frame in a fraction of the time.

Outcome

Prototyping time dropped sharply, and design never became the bottleneck as the story evolved.

Vertical 9:16 storyboard frame, shot 01 cold open Vertical 9:16 storyboard frame, shot 08 end card
30-second trailer storyboard with eight shots, each with visual, camera, caption, and sound notes
The 30-second trailer storyboard: 8 shots, each with visual, camera, caption, and sound notes. It started wide, then moved to a 9:16 vertical format.

Challenge 3

Designing for unpredictable text

The friction: the AI writes every scene and option live, so we never knew how long any line of copy would be.

Design strategy

Designed flexible layouts that hold AI-generated text of any length without breaking, instead of designing around fixed copy.

Outcome

The UI stayed intact across playthroughs where no two scenes were ever the same.

Wellness point popups for positive and negative choices

Challenge 4

Interaction without a tutorial

The friction: we wanted players to do more than point and click, but there was no room for a tutorial in a fast-paced escape.

Design strategy

Built physical mini-games tied to the opening — grabbing items in 15 seconds, throwing them into the fuel tank — so actions explain themselves.

Outcome

The three items grabbed at the start come back at the launch pad, linking interaction to story.

Throwing items at the target

Challenge 5

Keeping a self-writing story coherent

The friction: an open-ended AI story can drift, and we had designed far more than one weekend could hold.

Design strategy

Mapped scenarios on a whiteboard before building, fixed six locations, and cut scope until one clear idea remained. A premise document set the world, tone, and rules for the AI, and the AI remembers earlier choices to keep the journey connected.

Outcome

The constraint of six checkpoints sparked more diverse ideas within those limits, while keeping the generated story on track.

Logo and starting point

Challenge 6

Making designs land in 3D space

The friction: a design that looks right in Figma can feel wrong once it's placed in a 3D scene, and a flat screen alone doesn't feel like a world.

Design strategy

I worked closely with the developers, communicating placement, scale, and positioning in detail as they built each scene in Unity. I also created 360° backgrounds that surround the player at every checkpoint.

Outcome

The designs translated into VR as intended, and the judges said the 360° environments made the experience feel more vivid and immersive.

360-degree environment on the approach to the Moon

Challenge 7

Hiding the wait for live video

The friction: generating, streaming, and decoding each scene takes time. A blank screen or a sudden cut would break immersion right after every choice.

Design strategy

The screen never goes empty: it holds the last frame until the next one is ready. After a choice, the current video morphs into the outcome while the next scene prepares in the background. Choices appear only after a few seconds of visible video, and a Continue button lets the player move on when they're ready.

Outcome

Transitions feel like one continuous shot. Testing in the headset caught what looked fine on a monitor: UI blocking the video, unclear interactions, and transitions that ran too long.

Live video on the curved screen

Moon Citizen — or Moon Reject

Everyone reaches the Moon. Not everyone is welcome there. Every player starts at 100 wellness, and each choice moves it up or down by 5, judged by how you treat others. Score 85 or more to become a Moon Citizen; anything less and you're turned away at the colony gate.

Feedback

Showing the score without explaining it


Choice feedback

A +5 or −5 popup appears after each decision, so players feel the weight of a choice without being told the rules in advance.

Wellness bar

A bar stays in view from the first decision onward and shifts from green to orange to red as the score drops.

The result screen

The end screen shows the verdict, final wellness, successful throws, the full decision history, and a traveler type.

Mission report end screen with traveler type, final wellness, throws used, and choice recap
The mission report: traveler type, final wellness, throws used, and every choice you made

Verdict

Five traveler types for Moon Citizens


Tuning the score took several rounds of live testing, so the verdict reflects what kind of person you are under pressure.

The Guardian The Redeemed The Loyal One The Ace Pilot The Survivor

The Guardian

Never put yourself first. Everyone you met made it because of you.

The Redeemed

Left your roommate in the first minute, then spent the journey making it right.

The Loyal One

Would do anything for your own people. Strangers had to earn it.

The Ace Pilot

Hit every throw at the launch pad to fuel the rocket.

The Survivor

Not perfect, and you knew it. You did enough good to be let in.

Not admitted

"You made it to the Moon. The people here know what you did to get a seat."

The game result screen
The game result

Learnings & What's Next

Constraints made the creativity possible


What I learned about design

  • Preparing early paid off. Studying the Meta Quest and Vision Pro guidelines before the event turned an unfamiliar medium into one I could design for confidently under time pressure.
  • AI is a real accelerator under pressure. It cut prototyping time dramatically and made late changes, like moving the storyboard from wide to vertical, quick instead of costly.
  • Constraints fuel creativity. A tight scope made ideas feasible, and a smaller, complete journey was worth more than a long list of unfinished mechanics.
  • Map before you build. Laying out scenarios on a whiteboard first exposed missing pieces and kept the storyline consistent.
  • Test in the headset early. It caught problems a monitor hides, from UI that blocked the video to transitions that ran too long.
  • Environment design is part of the experience. The 360° backgrounds weren't decoration; the judges called them out as what made the world feel alive.

What I learned beyond design

  • Detailed communication bridges design and engineering. Frequent progress checks kept us on schedule, and talking through exact placement and scale is what made the Figma designs work in VR.
  • Writing for the AI is a design task. I wrote a premise document that set the world, tone, rules, and limits for the AI story director, and saw how much a clear brief and length rules shape what the AI produces.
  • I got a hands-on look at the development side. I studied how the OpenAI API connects to the game, and how each player choice becomes the next scene, two choices, and video prompts.
  • Sharing through GitHub brought design and code closer. I shared my design files in the team's repo so developers could pull them straight into the build, and saw how my work moved from Figma into the game.

What's next

  • Classroom edition on climate — land planning and water conservation, with a hopeful tone for kids.
  • Clearer visual consequences, so each choice reads instantly on screen.
  • More journeys beyond Stanford and the Moon.
  • Shorter waits between scenes, and eventually a continuous world with no cuts as live video generation speeds up.