Tori | Devpost

Inspiration

Experts say we are in a loneliness epidemic. Many want to meet together with others, but they don't know where to go!

AI has the power to generate engaging experiences anywhere.

Tori was inspired by the idea of using AI and VR to bring people together rather than isolate them from one another, creating engaging experiences quickly, anywhere.

What it does

Tori is a communal AI tour companion that lives in the real world through your camera.

With Tori anyone can immediately spin up an engaging tour of a city and broadcast it to others. Anyone can see in progress tours in their area and join the party!

There's no planning needed, and anyone seeking a good time can join in the fun!

Users open the app, hold up their phone, and see Tori, a friendly AR guide who can talk in real time using ElevenLabs voice.
Tori responds using Gemini AI to create warm, short, inclusive conversations.
Instead of being tied to one campus or event, Tori can appear anywhere — museums, city centers, hackathons, or classrooms — sparking curiosity and connection.

Multiple people can interact with the same AI at once, creating a sense of shared discovery!

How we built it

Framework: React Native + Expo for a cross-platform mobile app
AR Layer: Expo Camera for live, immersive background view
AI Engine: Gemini 2.5 Flash for conversational intelligence
Voice: ElevenLabs API for natural speech synthesis
UI: Minimal, translucent design focused on shared presence and accessibility
-AR/VR: Unity
Fallbacks: Web Speech API for browser speech input and Expo Speech for offline demo

All environment variables are managed via .env and deployed to Expo, allowing quick iteration and multi-device testing.

Challenges we ran into

Getting speech-to-text working reliably across web and mobile without paid APIs
Integrating real-time voice output (ElevenLabs) smoothly inside Expo’s audio system
Handling multi-user interaction logic while keeping the experience lightweight
Ensuring camera permissions and AR rendering worked consistently across iOS, Android, and web

Accomplishments that we're proud of

Built a fully functional prototype in less than two days — complete with AI chat, voice, and AR background
Created a distinct personality and visual identity for Tori that feels friendly and approachable
Achieved real voice output and natural conversation without backend servers
Designed a concept that redefines what AI assistants can be: shared, local, and community-oriented

What we learned

How to bridge Gemini’s generative models with real-time voice synthesis in a mobile environment
How design and personality can dramatically shape user trust and emotional connection with AI
The importance of balancing technical realism and social imagination — sometimes less is more when crafting believable digital beings

What's next for Tori

True multi-user mode: persistent shared state so everyone near Tori sees the same conversation
Geo-anchored experiences: use GPS and landmarks to let Tori guide tours dynamically
Memory & personalization (opt-in): remember names, past interactions, and community events
Custom voices and personas: local accents, seasonal themes, and sponsor tie-ins
Event-based challenges: scavenger hunts, “find Tori” quests, and team-building games
Security and Privacy Implementations