Data_Bleed

A psychological horror game that teaches scam awareness through experience, because fear teaches what lectures forget.

Comment

Inspiration

Every year, $10 billion is lost to online scams. Traditional cybersecurity training is boring, forgettable, and ineffective. People sit through mandatory videos, click "I understand," and then fall for the exact same scams weeks later.

I realized that the most effective way to learn about manipulation isn't through PowerPoint slides, it's through experiencing it firsthand in a safe environment. Horror games excel at creating visceral, memorable experiences that stick with you long after you stop playing. What if we could harness that psychological power for education?

That's how DATA_BLEED was born: a psychological horror game that teaches scam awareness through interactive storytelling, where every decision has consequences and every mistake teaches a lesson you won't forget. Because fear teaches what lectures forget.

What it does

DATA_BLEED is an interactive psychological horror experience that teaches scam awareness through three character-driven stories:

Eli: A young gamer facing peer pressure, gambling scams, and toxic online communities Maya: A cybersecurity investigator dealing with romance scams and catfishing Stanley: A businessman targeted by identity theft and financial fraud Players make real-time decisions that affect their Trust Score (0-100), which measures their vulnerability to scams. Bad decisions lower trust and trigger visual corruption effects, glitches, distortions, and unsettling audio that intensify as vulnerability increases. Good decisions build resilience and unlock protective knowledge.

ChromaBot, an AI assistant, provides context-aware guidance, answering questions about scam tactics and offering scene-specific advice. The game teaches recognition of 15+ common scam tactics including phishing, social engineering, urgency manipulation, FOMO exploitation, and authority bias.

How we built it

Development Stack: Frontend: JavaScript (ES6+), HTML5, CSS3, Web Audio API, Three.js (for future 3D integration) Backend: Node.js, Express.js, OpenAI GPT-4o-mini API Deployment: Vercel (frontend), Railway (backend), GitHub (version control) Creative Tools: LTX Studio & To Moviee (AI video generation for character scenes) ElevenLabs (AI music generation for atmospheric soundscapes) ChatGPT/Claude (script writing, dialogue generation, scam research) Google Nano Banana (concept art and visual assets) Adobe Express (video editing and post-production)

6-Week Development Process: Week 1-2: Research real scam cases, FBI reports, and victim testimonials. Designed three character archetypes and built the Trust Score System with branching narrative engine.

Week 3-4: Created ChromaBot AI with scene-specific training data. Implemented visual corruption system using CSS animations and WebGL effects that scale with trust score. Generated character videos using LTX Studio & To Moviee, w/voice narration and audio scoring with ElevenLabs.

Week 5: Integrated OpenAI API for intelligent chatbot responses. Built Node.js backend with character-specific training data and context-aware response system. Created atmospheric soundscapes with ElevenLabs.

Week 6: Deployed to Vercel and Railway, optimized mobile responsiveness, fixed CORS issues, and added email signup system. Final polish with Adobe Express for video transitions and timing.

Challenges we ran into

  1. Balancing Horror with Education: Making the game scary enough to be memorable without overwhelming players. Solution: Used psychological horror (visual corruption, unsettling audio) that escalates gradually based on choices instead of jump scares.

  2. AI Chatbot Context Awareness: ChromaBot was giving generic responses to direct questions. Solution: Implemented 4-tier response system: (1) scene-specific training data, (2) direct character/story answers, (3) OpenAI API for contextual intelligence, (4) generic fallbacks only for weird inputs.

  3. Trust Score Calibration: Finding the right balance between too punishing and too lenient. Solution: Extensive playtesting settled on -15 to -25 for bad decisions, +10 to +15 for good decisions, with critical threshold at 30.

  4. State Management: Maintaining player progress across 6 scenes without a database. Solution: Used browser sessionStorage with JSON serialization and state manager that syncs with pause menu dashboard.

  5. Video Generation Consistency: Getting LTX Studio & To Moviee to maintain character consistency across scenes. Solution: Created detailed prompt templates with consistent character descriptions, lighting conditions, and camera angles. Used reference images and iterative refinement.

  6. Audio Synchronization: Syncing ElevenLabs narration with video timing and decision points. Solution: Built custom audio manager with precise timing controls, pause/resume functionality, and scene-aware playback.

  7. Mobile Performance: Horror effects caused lag on mobile devices. Solution: Implemented CSS will-change for GPU acceleration, reduced animation complexity on smaller screens, and disabled heavy effects on low-end devices.

Accomplishments that we're proud of

✅ Created a genuinely innovative educational format that combines psychological horror with cybersecurity training, something that doesn't exist in the market

✅ Built a fully functional AI chatbot that adapts to player choices and provides contextual, scene-specific advice using OpenAI integration

✅ Designed a Trust Score system that quantifies vulnerability in real-time and creates meaningful consequences for player decisions

✅ Generated 18+ AI videos using LTX Studio & To Moviee with consistent character designs and compelling narratives

✅ Produced professional voice narration with ElevenLabs that sounds natural and emotionally resonant across 6 scenes per character

✅ Implemented visual corruption effects that scale dynamically with trust score, creating escalating psychological tension

✅ Deployed a production-ready application with separate frontend/backend architecture, CORS configuration, and mobile optimization

✅ Created 15+ easter eggs rewarding curious players who explore the pause menu chat system

✅ Achieved seamless audio-visual synchronization between AI-generated videos, voice narration, and atmospheric music

What we learned

Technical Skills:

Full-stack development with Node.js/Express backend and vanilla JavaScript frontend AI API integration (OpenAI API) with context-aware response systems State management across complex branching narratives Horror game design principles: visual glitch effects, corruption animations, psychological tension Deployment architecture with Vercel and Railway, including CORS configuration Mobile-responsive design and performance optimization

AI Creative Tools: LTX Studio & To Moviee: Prompt engineering for consistent character generation, understanding motion controls, and iterative refinement techniques

ElevenLabs & To Moviee: Emotional tone control, natural speech synthesis & Atmospheric music generation, genre blending, and mood creation

Prompt Engineering: Crafting detailed, consistent prompts across multiple AI tools for cohesive storytelling

Game Design: Narrative branching with meaningful player choices Trust mechanics that quantify and visualize vulnerability Balancing entertainment with educational outcomes Creating psychological horror without jump scares

Psychology & Scam Tactics: Deep understanding of social engineering, phishing, romance scams, gambling addiction, and identity theft How cognitive biases (urgency, FOMO, authority) make people vulnerable Why experiential learning is 10x more effective than passive training

Content Production: Video editing and post-production workflows Audio synchronization and sound design Maintaining narrative consistency across AI-generated content Iterative creative refinement with AI tools

What's next for Data_Bleed

Immediate Goals (3-6 months): 3D Character Integration: Using NeRF (Neural Radiance Fields) to create photorealistic, interactive AI characters that emerge from the screen Complete Maya & Stanley Stories: Finish all 18 scenes (6 per character) with full video, narration, and branching paths Advanced Analytics Dashboard: Track player vulnerability patterns and provide personalized scam awareness reports Mobile App: Native iOS/Android versions with offline play and push notifications for daily scam awareness tips Medium-term Vision (6-12 months):

Multiplayer Mode: Collaborative scam investigation scenarios where players work together to identify and stop scams Corporate Training Platform: Custom scenarios for businesses with admin dashboards, progress tracking, and compliance reporting (potential $50K+ per enterprise client vs. millions in scam losses) Educational Partnerships: Integration with schools and universities as part of the digital literacy curriculum Localization: Translate to 10+ languages to reach global audiences Long-term Impact (1-2 years):

VR Version: Full immersion using Meta Quest/Apple Vision Pro for maximum psychological impact and retention AI-Powered Personalization: Adaptive difficulty that adjusts scam complexity based on player performance Real-World Scam Database: Integration with FBI/FTC databases to simulate actual scam tactics being used today Certification Program: Partner with cybersecurity organizations to offer official scam awareness certification

Built With

Share this project:

Updates