The main page
Results generated
Image uploading and giving the prompt

EventSense: AI-Powered Visual Event Planning with Google Gemini

Inspiration

We’ve all been there: you see a photo on Pinterest or Instagram that perfectly captures the vibe of an event you want to host — but you have no idea how to translate that static image into a real party.

The gap between visual inspiration and logistical execution is huge. You know what it looks like, but not how to schedule it, budget for it, or even what food matches that aesthetic.

We built EventSense to close that gap.

Our goal was to create an AI planner that doesn’t just read text — it actually sees your vision and instantly handles the boring math and planning for you.

What It Does

EventSense is a multimodal AI system that transforms a single inspiration image into a complete, actionable event plan.

Upload an image. Describe your event. Get a full plan.

Core Features

Visual Analysis Upload an image (e.g., a cyberpunk neon party or a cottage-core wedding).
Constraint Integration Provide basic details like:
- Budget: $5000
- Guests: 50
Generative Planning (Powered by Gemini)

The system analyzes the image and generates:

Thematic Analysis → Color palettes & aesthetic mood
Logistics → A full Run of Show agenda + budget breakdown
Decor Strategy → Exact items needed to recreate the look
Menu & Catering → Theme-matching cuisine (with cost estimates)
Marketing → AI-generated headlines & poster prompts

How We Built It

We used a modern full-stack architecture powered by Google Gemini’s multimodal capabilities.

The Brain (AI)

Google Gemini (gemini-1.5-flash + Gemma models)
Supports image + text in a single API call
Maintains full contextual understanding

Backend

**Python
Pillow for image processing
python-dotenv for secure environment variables
Enforces strict JSON-only responses from the AI

🎨 Frontend

React
Modern Glassmorphism UI
Real-time image preview
Dynamic dashboard that converts JSON → clean visual cards

Deployment

Layer	Platform
Backend	Render
Frontend	Vercel

This allowed us to deploy a fully live, accessible web application.

Challenges We Ran Into

Structured Output from Chaos Getting an LLM to analyze an abstract image and return strictly formatted JSON was hard. We heavily refined system prompts to prevent rambling.
Model Availability Some Gemini models were region-locked or unavailable on free tiers. We built fallback logic so the demo never breaks.

Accomplishments We’re Proud Of

True Multimodal AI — visual input directly drives planning logic
End-to-End Automation — hours of planning reduced to one click
UI/UX Excellence — dynamic rendering of complex schedules & menus
Cloud Deployment — split frontend/backend deployed within hackathon time

What We Learned

The Power of Multimodal AI Giving AI eyes drastically improves contextual understanding.
Prompt Engineering = Coding The prompt is just as important as the Python logic.
State Management in React Handling async AI responses sharpened our understanding of hooks.

What’s Next for EventSense

Vendor Recommendations Integrate real APIs (Amazon, Yelp) for actual product suggestions
Multi-Image Mood Boards Upload multiple images and extract the common aesthetic
Cost Optimization Modes Budget / Standard / Luxury plans for the same event
Mobile App Convert the React app into a React Native experience