Fitcheck

Inspiration

We personally find it hard to find clothes to wear in our wardrobe when shopping don't know what we have hard to find a fit that's good.

What it does

FitAI is an AI-powered personal style assistant that: Lets you upload a photo of yourself with any clothing item and get an instant fit check with a compatibility score, color analysis, and styling tips Maintains a digital wardrobe catalogue of clothes you already own Suggests matching pieces from your wardrobe or online stores (with budget, mid, and premium price tiers) Adapts suggestions based on occasion (date night, office, gym), weather (auto-detected from your location), and your personal style profile Offers a "Style Inspo" mode where you type any celebrity or aesthetic and get matching items from your wardrobe plus shopping links Includes an outfit planner calendar so you can plan your week and avoid repeating looks

How we built it

Frontend: React with Vite and TailwindCSS for a fast, responsive UI Backend: Python FastAPI handling image processing and API orchestration AI: OpenAI GPT-4o Vision for outfit analysis and styling advice, DALL-E 3 for virtual try-on visualization Weather: Open-Meteo API with browser geolocation for real-time weather-aware suggestions Storage: Local JSON persistence for wardrobe data, localStorage for style profiles and saved outfits Development: Built entirely using Kiro, an AI-powered development environment

Challenges we ran into

The virtual try-on visualization was our biggest hurdle, generating a realistic image of the user wearing specific clothes required detailed prompt engineering. DALL-E creates a new image rather than editing the original photo, so getting it to resemble the actual user took extensive iteration on how we describe the person to the model. We also hit Python dependency conflicts between OpenAI and httpx versions, and had to work around zsh shell issues with special characters in our project path.

Accomplishments that we're proud of

Clean, intuitive UI that anyone can use without instructions The style profile onboarding that personalizes every recommendation Weather-aware suggestions that auto-detect your location Price tier comparison so users can shop at any budget level The inspo mode with pre-built aesthetic presets makes it fun to explore

What we learned

How to use Kiro to rapidly build a full-stack app from idea to working product. We also learned how to orchestrate multiple AI models (GPT-4o for analysis, DALL-E for generation) in a single workflow, and how prompt engineering dramatically affects output quality.

What's next for Fitcheck

Faster and highly accurate visualization using dedicated virtual try-on models (like IDM-VTON) so users can upload a full-body photo as a persistent avatar and see clothes rendered directly onto their body in real-time User accounts with cloud sync so your wardrobe travels with you Social features — share outfits with friends and get votes Integration with shopping platforms to auto-import items from purchase history A "Do I need this?" feature that checks if you already own something similar before you buy