Inspiration
In today’s Horizon ecosystem, new users often feel lost, overwhelmed, or unsure where to go next. Social VR has endless possibilities — but no personal guide that adapts to each user.
I wanted to create an AI companion that feels alive, attentive, and truly helpful, not just a static NPC. This inspired Aura Verse — an AI companion built to understand your preferences, remember your interactions, and support you across Horizon experiences.
What it does
AuraVerse is your personal AI companion inside Meta Horizon.
✨ Key abilities:
Learns your preferences (interests, play style, social comfort level)
Remembers past interactions and adapts personality over time
Guides you inside Horizon worlds using voice or gesture
Suggests activities, communities, and quests based on your vibe
Holds natural conversations powered by Llama-based AI
Acts as your emotional support buddy — detects your tone and responds empathetically
Generates mini-missions to keep users engaged
Multi-modal inputs: speech, text, intent detection, object interaction
Aura Verse feels like a friend, a guide, and a smart assistant combined.
How we built it
How I built it
AuraVerse is built using a combination of AI, Horizon SDK, and vibe coding (AI-assisted development) techniques.
Meta Horizon SDK → for world interactions, objects, triggers, composable functions
Llama 3.1 API → for the companion's reasoning and dialog generation
GraphQL + Horizon scripts → for state management and event routing
Custom Memory Engine → stores user preferences and emotional patterns
Voice recognition module → converts player speech into intents
AI tooling → used to generate UI layouts, scripts, dialogues, and prototype logic
Fast prototyping pipeline → 70% of code prototyped using AI, refined manually
Every behavior from emotional responses to navigation hints is powered by a compact context-aware AI loop.
Challenges we ran into
Implementing long-term memory without increasing latency
Getting the AI to feel natural, not robotic
Blending Horizon interactions with AI-driven dialogue
Making the companion respond to gestures and world objects
Ensuring the AI does not speak too long or go off-topic
Creating a consistent personality without breaking immersion
Accomplishments that we're proud of
Created a fully personalized AI companion that actually remembers users
Built a stable Horizon interaction system linked with an LLM
Developed a dynamic emotional adaptation layer
Integrated voice + gesture inputs smoothly
Designed the entire UI/UX experience solo within the deadline
What we learned
How to integrate AI reasoning with VR world logic
How to use Horizon’s new scripting workflow effectively
How to build memory-driven AI personalities
How to rapidly prototype complex features using vibe coding
That AI companions, when designed right, can improve comfort and engagement in VR
What's next for Aura Verse
Adding multiworld cross-presence (companion follows between worlds)
Expanding personality types (calm, energetic, funny, mentor mode)
Launching a community-driven “persona marketplace”
Voice cloning for ultra-personalized companion voices
Emotional wellbeing tracking & reflective journaling in VR
Multiplayer Aura Verse: companions that talk to each other
Built With
- 3.1
- ai
- cloud
- engine
- graphql
- horizon
- javascript
- llama
- memory
- meta
- recognition
- scripts
- sdk
- voice
- vr
- webapi
Log in or sign up for Devpost to join the conversation.