Detective Office
BedRoom
CrimeScene
Bedroom
Detective_Office
Living_Room
Living_Room
Husband's Office
Back_Yard(Garden)
QR

CHRONO-DETECTIVE: THE MOBILE TIMELINE SHOWDOWN

PROJECT DOCUMENTATION

================================================================================

INSPIRATION

================================================================================

The concept of Chrono-Detective was born from a fascination with classic detective mysteries and the innovative potential of Generative AI in immersive gaming. We wanted to create an experience that combines:

Classic Whodunit Mystery: Inspired by Agatha Christie's murder mysteries and games like Clue and LA Noire
Interactive Storytelling: Where players actively participate in uncovering the truth through exploration and interrogation
Cutting-Edge Technology: Leveraging Meta Horizon's GenAI NPC capabilities to create believable, conversational characters
Mobile-First Design: Built specifically for the Meta Quest platform with mobile performance in mind

The central question driving our development was: "Can we create a believable murder mystery where AI-powered NPCs can lie, deceive, and maintain their alibis while players hunt for the truth?"

================================================================================

WHAT IT DOES

================================================================================

Chrono-Detective is an immersive murder mystery game where players step into the role of a detective investigating the murder of a wealthy woman in her mansion. The victim was stabbed in the bedroom, and a valuable diamond was stolen from the locker. The player must interrogate suspects, collect physical evidence, and use photography to document the crime scene—all while uncovering which of the four suspects is the killer.

IMMERSIVE CRIME SCENE INVESTIGATION

Players explore five meticulously crafted rooms:

Detective Office: The hub where players start and solve the case
Living Room: Scene of suspicious activity with broken furniture
Garden: Outdoor area where the gardener witnessed key events
Bedroom: The primary crime scene with the victim's body
Husband's Office: Where the husband claims he was sleeping during the murder

CLUE COLLECTION SYSTEM

Players must discover and document seven critical pieces of evidence:

Broken Photo Frame - Signs of struggle
Bloody Cloth - DNA evidence
Broken Glass Table - Indicating violence
Broken Pot - From the garden altercation
Letter - Revealing hidden motives
Knife with Blood - The murder weapon (planted in husband's hand)
Dead Body - Must be photographed using the in-game camera

The game uses Meta Horizon's Camera API to allow players to photograph evidence, creating an authentic forensic investigation experience.

NPC INTERROGATION WITH GENAI

Four unique suspects inhabit the mansion, each with their own story:

SOFIA (HOUSE MAID) - THE KILLER

Claims: "I was tidying the Living Room and Kitchen. I never left the main floor."
Reality: She murdered the victim, stole the diamond, and framed the husband
Behavior: Confident, lying, maintains her alibi perfectly

JACK (HUSBAND)

Claims: "I fell asleep at my desk working late. I didn't hear anything."
Reality: He was genuinely asleep and is being framed
Behavior: Confused, defensive, worried about false accusations

JUSTIN (GARDENER)

Claims: "I was tending roses near the fence. I heard glass breaking from the living room."
Reality: He's telling the truth and is a key witness
Behavior: Nervous but honest, provides crucial testimony

COOK (KITCHEN STAFF)

Claims: "I was preparing meals in the kitchen all evening."
Reality: Working in kitchen, away from the crime scene
Behavior: Professional, cooperative, limited knowledge

Each NPC features:

Proximity-Based Dialogue: NPCs automatically greet players when they enter trigger zones
GenAI-Powered Conversations: Players can ask custom questions and receive contextual responses
Maintained Alibis: The killer (Sofia) consistently lies while maintaining her innocent act
Attention System: NPCs face and track players during conversations

CASE RESOLUTION: THE BOARD

Once all evidence is collected and the body is photographed, players return to the Detective Office where a case board awaits. The board presents:

10 Multiple Choice Questions testing the player's deductive reasoning
Evidence-Based Queries: Questions about motives, alibis, timeline, and physical evidence
Perfect Score Requirement: Players must answer all 10 questions correctly to solve the case
Feedback System: Wrong answers trigger "Sorry, try again" message with retry opportunity

Win Condition: 10/10 correct answers = Case Solved!

================================================================================

HOW WE BUILT IT

================================================================================

ENVIRONMENT SETUP

We created a multi-room mansion using Meta Horizon World's building tools:

Custom Room Design: Five interconnected rooms with distinct aesthetics
Interactive Props: Clue objects placed strategically throughout the environment
Navigation System: Button-based teleportation between rooms using Room_Spawn_Teleporter.ts
Trigger Zones: Proximity-based areas around each NPC for automatic conversation initiation

GENAI NPC SCRIPTING

The heart of our game lies in the AI-powered NPCs. Technical implementation:

Key Technical Features:

Npc vs NpcPlayer Separation: Npc entity handles conversation, NpcPlayer manages attention/behavior
Conversation API: Used npcEntity.conversation.speak() for dialogue delivery
Attention Target System: addAttentionTarget() makes NPCs face and track players naturally
Event-Driven Architecture: CodeBlockEvents for OnPlayerEnterTrigger/ OnPlayerExitTrigger
Error Handling: Comprehensive validation for missing entities and disabled conversation AI

CUSTOM UI AND INTERACTIVITY

Camera System (CameraFlashEffect.ts):

Grabbable camera object with flash mechanic
Light gizmo activation on trigger press
Shot counter and ammo display system
Audio feedback for camera shots

Item Collection (ItemManager.ts):

Dynamic spawning system for collectibles
Battery pickup mechanics for camera recharging
Particle effects and visual feedback

Navigation (Room_Spawn_Teleporter.ts):

Button-based teleportation system
Uses this.world.findEntity() to locate spawn points
Instant transport between rooms for mobile optimization

CASE BOARD & QUESTION SYSTEM

Built using Meta Horizon's UI components:

Conditional Activation: Board only unlocks after all clues collected + body photographed
MCQ Interface: 10 questions with multiple choice options
Score Validation: Logic to check for perfect 10/10 score
Retry Mechanism: Failed attempts allow players to try again

================================================================================

CHALLENGES WE RAN INTO

================================================================================

1. MOBILE PERFORMANCE CONSTRAINTS

CHALLENGE: Meta Quest has limited processing power compared to PC VR. Initial builds with high-poly models and complex lighting caused frame rate drops.

SOLUTION:

Optimized 3D assets with lower polygon counts
Used baked lighting instead of real-time shadows where possible
Implemented teleportation instead of smooth locomotion to reduce motion sickness
Limited particle effects and visual complexity in each scene
Strategic use of occlusion culling in room designs

2. NARRATIVE RELIABILITY VS. LLM FREEDOM

CHALLENGE: While GenAI NPCs provide engaging conversations, they sometimes deviated from their scripted alibis or revealed information too early, breaking the mystery.

SOLUTION:

Hybrid Approach: Scripted initial messages for critical alibi information
Proximity-Based Auto-Speak: NPCs deliver their main statement automatically when players enter trigger zones
Conversation AI for Follow-up: Players can ask custom questions after the scripted intro
One-Time Delivery: Used hasSpoken Map to ensure each player hears the message only once
Session Management: Tracked player interactions per session to maintain consistency

3. TYPESCRIPT API INTEGRATION

CHALLENGE: Horizon Worlds uses specific API patterns that differed from standard TypeScript:

Confusion between Npc and NpcPlayer classes
ReadableHorizonProperty requiring .get() and .set() methods
Module import paths (horizon/core, horizon/npc) not well-documented

SOLUTION:

Studied .d.ts type definition files extensively
Created reusable patterns across all NPC scripts
Implemented proper error checking with isConversationEnabled()
Used as() casting for entity type conversion

4. CAMERA API INTEGRATION

CHALLENGE: Integrating Meta's Camera API for evidence photography while maintaining game state.

SOLUTION:

Created custom CameraFlashEffect component with shot counter
Linked camera usage to clue validation system
Used network events to broadcast camera flash across clients
Implemented visual and audio feedback for successful photos

================================================================================

ACCOMPLISHMENTS THAT WE'RE PROUD OF

================================================================================

EFFECTIVE GENAI INTEGRATION

We successfully created a murder mystery where:

AI NPCs maintain complex alibis throughout multiple player interactions
The killer (Sofia) lies convincingly while other NPCs tell the truth
Natural conversation flow emerges from hybrid scripted + GenAI approach
Players can't easily "break" NPCs with unexpected questions due to carefully crafted conversation contexts

This represents one of the first fully playable VR detective games powered by Generative AI on the Meta Quest platform.

SEAMLESS MOBILE VR EXPERIENCE

Despite Quest's hardware limitations, we achieved:

Stable 72fps performance across all five rooms
Intuitive VR interactions with grabbable objects and camera
Zero motion sickness reports from playtesters (thanks to teleportation)
Rich narrative experience without compromising performance

COMPLETE GAME LOOP

We built a fully functional end-to-end detective game:

Immersive crime scene with multiple rooms
Interactive NPC interrogations with AI
Physical clue collection system
Camera-based evidence documentation
Logic puzzle resolution with MCQ board
Win/lose conditions with feedback

Many Mobile game prototypes remain incomplete—we're proud to have shipped a polished, playable experience from start to finish.

================================================================================

WHAT WE LEARNED

================================================================================

TECHNICAL LEARNINGS

Meta Horizon Worlds Scripting: Deep dive into TypeScript-based world building, component architecture, and event systems
GenAI NPC Development: Understanding the balance between AI freedom and narrative control
VR UX Design: How spatial interaction differs from traditional gaming (proximity triggers > click buttons)
Performance Optimization: Mobile VR requires aggressive optimization at every level

GAME DESIGN LEARNINGS

Player Agency vs. Narrative Structure: Too much freedom confuses players; we need clear objectives and progression gates
Environmental Storytelling: Physical clues placed in logical locations tell a story without explicit exposition
Pacing: The 10/10 perfect score requirement creates tension and encourages thorough investigation

TEAM COLLABORATION

Documentation Importance: Creating setup guides (SOFIA_NPC_SETUP_GUIDE.md, etc.) was crucial for team members to configure entities correctly
Iterative Playtesting: Early feedback revealed that players needed more guidance on where to find clues
Version Control for VR Assets: Managing Horizon World states required careful coordination

================================================================================

WHAT'S NEXT FOR CHRONO-DETECTIVE

================================================================================

EXPANDING THE STORY

Multiple Cases: Add 3-5 additional murder mysteries with different suspects, motives, and locations
Dynamic Killer Selection: Randomize which NPC is the killer each playthrough for replay value
Branching Narratives: Player choices during interrogations affect which endings are available
Backstory Layers: Deeper character development with hidden relationships and motives

CLUE VALIDATION SYSTEM

Currently, the board unlocks based on: if (allCluesCollected && bodyPhotographed) { enableBoard(); }

Future Enhancement:

Photo Quality Checks: Validate that evidence photos actually capture the clue clearly
OCR Integration: Read text from photographed letters/documents
Clue Notebook: Visual inventory showing collected evidence with photos
Timeline Reconstruction: Players arrange events in chronological order

ADVANCED GENAI FEATURES

Dynamic Alibi Generation: NPCs generate unique alibis each playthrough using LLM
Relationship System: NPCs reference each other's testimonies dynamically
Emotional States: NPC behavior changes if confronted with contradictory evidence
Voice Synthesis: Integrate Meta's text-to-speech for more natural NPC dialogue

MULTIPLAYER EXPANSION

Co-op Mode: 2-4 detectives work together to solve cases faster
Competitive Mode: Race against other detectives to solve first
Asymmetric Gameplay: One player is the killer trying to deceive AI detectives

ACCESSIBILITY & LOCALIZATION

Subtitles: Full text support for NPC dialogue
Colorblind Modes: Alternative visual markers for clues
Multiple Languages: Translate story content and UI to Spanish, French, Japanese
Difficulty Settings: Adjustable clue requirements and question count

================================================================================

CONCLUSION

================================================================================

Chrono-Detective: The Mobile Timeline Showdown demonstrates the exciting potential of Generative AI in immersive storytelling. By combining classic detective game mechanics with cutting-edge NPC technology, we've created an experience that feels fresh yet familiar—a VR murder mystery where you truly interrogate intelligent suspects who can lie, deceive, and maintain complex alibis.

Our journey from concept to playable game taught us that the future of VR gaming isn't just about graphics or performance—it's about creating believable worlds populated by intelligent characters that react and adapt to player actions in meaningful ways.

As VR hardware improves and AI becomes more sophisticated, we envision detective games where: