FitCheck

FitCheck is your context-aware AI stylist. It builds outfits from your closet based on real-time weather & schedule, and suggests items to buy to complete the look

Comment

Inspiration

We’ve all experienced "Wardrobe Paralysis"—staring at a closet full of clothes for 15 minutes, yet feeling like we have nothing to wear. The average person spends months of their life just deciding what to put on. We realized that existing "digital closet" apps are just static inventories; they don't actually help you decide.

We wanted to build an engine that doesn't just list your clothes, but understands your life. We were inspired to create a "second brain" for your style—one that considers the rain outside, the meeting on your calendar, and the clothes you actually own, to eliminate decision fatigue instantly.

What it does

FitCheck is a context-aware AI stylist that lives in your browser.

  1. Smart Ingestion: Users upload photos of their clothes. The app uses Gemini 1.5 Pro Vision to automatically tag them with metadata (Color, Type, Formality, Seasonality)—no manual data entry required.
  2. Context Engine: It pulls real-time Weather data and checks your Google Calendar.
  3. The Recommendation: You can ask via Voice or Text: "What should I wear today?" FitCheck analyzes your schedule (e.g., "Client Meeting"), the weather (e.g., "Humid"), and your inventory to generate the optimal outfit.
  4. The Completer: If you have a great shirt but no matching trousers, FitCheck identifies the "Gap" in your wardrobe and suggests exactly what item you should buy to complete the look.

How we built it

We stuck to a high-performance, 100% JavaScript stack to move fast:

  • Frontend & Backend: Built with Next.js 14 (App Router) for a seamless full-stack experience.
  • AI Logic: We utilized Google Gemini 1.5 Pro. We used its multimodal capabilities to analyze clothing images and its reasoning capabilities to match specific inventory IDs to complex user prompts.
  • Database: We used SQLite managed by Prisma. This local-first approach ensures the app is incredibly fast and responsive, with zero cloud latency for the demo.
  • Voice Integration: We implemented the native Web Speech API for a lightweight, privacy-focused voice interface.
  • Styling: Tailwind CSS for a clean, mobile-first aesthetic.

Challenges we ran into

  • Structured AI Outputs: Getting a Large Language Model to consistently output strict JSON for database queries (e.g., selecting specific Clothing IDs) was difficult. We had to refine our system prompts heavily to ensure Gemini didn't "hallucinate" items the user didn't own.
  • Context Merging: figuring out how to weight conflicting data points was a fun logic puzzle. (e.g., If it's hot outside but the user has a formal boardroom meeting, which factor wins? The answer: Breathable fabrics, but formal cuts).
  • Latency: Processing images for auto-tagging can be slow. We solved this by using optimistic UI updates so the user didn't feel the wait.

Accomplishments that we're proud of

  • "The Completer" Logic: We didn't just build a closet organizer; we built a recommendation engine that identifies what you don't have. This turns the app from a utility into a genuine shopping assistant.
  • Zero-Friction Onboarding: The auto-tagging feature works seamlessly. You just snap a picture, and Gemini does the rest. It feels like magic.
  • Local Performance: Using SQLite and running the logic locally makes the app feel instant, which is crucial for a daily utility tool.

What we learned

  • Multimodal is King: Text-only AI is limited. Being able to pass an image of a shirt and ask "What goes with this?" unlocks a level of utility that text prompts can't match.
  • The Power of Context: A recommendation is only as good as the context behind it. Integrating Weather and Calendar turned generic advice ("Wear a suit") into actionable intelligence ("Wear the grey suit, but take the trench coat because it will rain at 4 PM").

What's next for FitCheck

  • AR "Virtual Mirror": We want to implement a generative AI feature that shows the user wearing the suggested outfit before they even put it on.
  • Social "Vibe Check": Adding a feature to share outfit options with friends for a "Tinder-style" vote before leaving the house.
  • E-commerce Integration: Turning "The Completer" suggestions into real, one-click purchase links to monetize the platform.

Built With

Share this project:

Updates