Inspiration
We've all stood in front of a full closet feeling like we have nothing to wear. The problem isn't the clothes — it's the mental effort of putting them together. We were inspired by the idea of giving everyone access to a personal stylist in their pocket, one that actually knows what you own, understands the weather, and builds outfits with purpose, and sense of fashion.
What it does
FORME is an Gemini AI-powered personal styling platform with six core features. The Today tab generates three fresh outfit suggestions every morning based on real-time local weather and different occasions, letting users like, dislike, or save looks. My Closet is a digital wardrobe where users can add items by searching any product online or uploading a photo. The Style Builder lets users pick any anchor piece and have Gemini AI construct a full outfit around it, pulling from their closet and suggesting items to buy. Saved Looks stores every outfit a user saves in one place. Outfit Preview shows saved looks in an editorial moodboard layout with a mannequin view that applies outfit colours to each body zone. The Shop tab surfaces real fashion products filtered by style aesthetic, gender, and budget.
How we built it
We built FORME using Next.js 15 with the App Router for page routing and server-side API routes. All styling was done with Tailwind CSS. Google Gemini 2.5 Flash powers the AI features — daily outfit generation, the Style Builder, and styling tips. SerpAPI's Google Shopping integration helps implement both the closet item search and the shop product discovery. We used React Context with localStorage for persistent state across the closet and saved looks. We built the project was using GitHub with a feature branch workflow, with the two of us owning specific pages and components.
Challenges we ran into
Getting Gemini to consistently return structured JSON was one of our biggest early challenges — we had to carefully engineer prompts and add response cleaning to handle markdown code fences. Gemini model availability was a recurring issue as several models returned 404 errors before we landed on gemini-2.5-flash. We also ran into quota limits on the free tier during development which required us to enable billing. On the frontend, sharing state across pages without a database required building a custom React Context system backed by localStorage, which took several iterations to get right. Merge conflicts from collaborative development also tested our Git workflow.
Accomplishments that we're proud of
We're proud of building a fully functional AI styling product in a single hackathon. The Style Builder genuinely produces thoughtful outfit recommendations that pull from a user's closet. The Today tab integrates real weather data seamlessly with Gemini to produce contextually relevant suggestions. The Shop page fetches live products filtered by style, gender, and budget. We're also proud of the overall design.
What we learned
We learned a lot about working with large language model APIs in production — specifically around prompt engineering, structured output parsing, and handling quota and model availability issues gracefully. We deepened our understanding of Next.js App Router patterns, particularly server-side API routes and how to keep API keys secure. On the collaboration side, we learned the importance of clear file ownership and consistent branching to avoid conflicts. Most importantly we learned that AI features require careful UX thinking - the technology is only as good as the interface around it. We feel we both collaborated well when planning what we wanted to do but neither of us expected the challenge ahead, still it was a very fun and learning experience.
What's next for StyleAI
The most immediate next step is a real virtual try-on feature using a dedicated model like Fashn.ai, where users upload a selfie and see clothing composited directly onto their body. We also want to add user accounts and a cloud database so closets and saved looks persist across devices. A community feed where users can share and discover outfits from others would add a social dimension. On the AI side, we want to use Gemini's vision capabilities to automatically identify and categorise clothing from a photo the moment it's uploaded to the closet, making onboarding effortless.
Log in or sign up for Devpost to join the conversation.