Click the image above to watch the demo on YouTube
Ever stood in front of your closet for 30 minutes, trying on outfit after outfit, only to end up wearing the same thing? We've all been there. We wanted to create a magic mirror that shows you what you could look like — a personal AI stylist available 24/7.
Silhouette is your AI-powered fashion advisor:
- 📸 Snap a selfie using the live camera preview
- 🎯 Click on your body to specify what you want to change
- 🎨 Pick your vibe — Streetwear, Formal, or Sporty
- 🤖 Let AI work its magic — Overshoot analyzes your photo, reads your preferences, and Gemini generates a stunning visualization
- 🛒 Shop the look — Get clickable links to actual products
- Frontend: Next.js 16 + React 19 with Radix UI and Tailwind CSS
- Backend: Express.js + Node.js + MongoDB Atlas
- AI:
- Google Gemini for image generation and recommendations
- Overshoot SDK for real-time video analysis
- Custom prompt engineering to make AI understand fashion
- Teaching AI the difference between "streetwear" and "trying too hard"
- Real-time video processing without melting laptops
- Making AI-generated fashion photos look realistic
- Rate limits when generating multiple outfits
- Cross-browser compatibility issues
- ✅ Fully functional AI fashion advisor built in hackathon timeframe
- ✅ Beautiful, intuitive UI for body annotation
- ✅ Seamless integration of multiple AI services (Overshoot + Gemini)
- ✅ AI-generated fashion photos people actually want to wear
- ✅ Real shopping links integrated
- ✅ Polished "Find Yourself" landing animation
- AI image generation has come incredibly far
- Prompt engineering is an art form
- Real-time video processing is powerful but challenging
- Building with multiple AI APIs requires serious orchestration
- Good UX is universal, even when fashion is subjective
- 👥 Social features — Share looks and get votes
- 🎨 More style vibes — Cottagecore, Y2K, Dark Academia
- 📱 Mobile app
- 💳 Direct checkout integration
- 🎥 Video try-on
- 🧵 Wardrobe integration
- 🤝 Brand partnerships
Because everyone deserves to feel like the main character. 🌟
We utilized Generative AI to build this project:
- ChatGPT
- GitHub Copilot
- Claude
- Cursor
We also utilized the Overshoot API and its documentation.
Our main IDE was DevSwarm! Using their parallel AI agent workflows, we were able to work on multiple branches (frontend, backend) simultaneously with different AI agents. We also used DevSwarm's multiple workflows to create different reasoning versions of our AI agent and tested to find the best version. DevSwarm was vital to our project and workflow.