VYBE | Devpost

Inspiration

We realized that everyone has moments where they’re unsure what to wear whether it’s a job interview, a date or a casual hangout. Finding a good outfit that fits the occasion and style isn’t always easy and professional stylists aren’t accessible to everyone. That’s why we built VYBE, which is an AI stylist that helps users upload their outfit, select the scenario or category (like formal, casual, street, vintage or sporty) and receive personalized analysis and recommendations to elevate their look.

What it does

VYBE is an AI-powered fashion assistant that helps users refine their style through scenario-based outfit analysis and smart AI recommendations. Here’s what users can do:

Upload a photo of their outfit and select a category or scenario such as formal, casual, street or vintage.
Receive an AI-driven evaluation tailored to that scenario including outfit scores, color harmony and trend fit.
Get personalized improvement tips like what to keep, adjust or replace.
Browse AI-curated clothing recommendations that match the chosen style, complete with direct purchase links.
Select items they like, and VYBE will generate front and back virtual try-on images, showing how they would actually look wearing the new pieces.
VYBE makes fashion guidance both personalized and interactive, helping users plan outfits for any occasion while visually exploring new styles.

How we built it

We built VYBE using a TypeScript-based React frontend and a JavaScript (Express.js) backend. The frontend was developed with Vite + React UI, focusing on clean design and responsive interaction. The backend handles image uploads and routes requests to the Gemini API, which analyzes the outfit, generates improvement suggestions and recommends matching fashion styles. To provide real purchase options, we integrated the Google Search API to fetch real-world clothing items that align with each AI recommendation. We used Multer for image upload handling and prompt-engineered Gemini queries to ensure consistent, context-aware responses across all fashion categories.

Challenges we ran into

Integrating multimodal AI responses: Getting the Gemini API to consistently analyze both images was tricky especially when users uploaded outfits with complex lighting, poses or mixed styles.
Prompt tuning for fashion feedback: We had to experiment with prompt engineering to make Gemini’s responses more fashion-specific and structured, ensuring it returned useful advice like “what to keep” and “what to improve” instead of generic comments.
Balancing accuracy and performance: Generating results quickly while maintaining high-quality image analysis and realistic outfit feedback required optimizing API calls and reducing unnecessary re-renders in React.
Recommendation relevance: Using the Google Search API to fetch clothing links sometimes returned unrelated items, so we refined our filtering logic and ranking system to match the selected style or scenario (e.g., formal vs. sporty).

Accomplishments that we're proud of

Built a working end-to-end AI outfit analysis system that’s interactive and intuitive.
Generated realistic front and back-view outfit visualizations using Gemini’s image model.
Designed a modern, fashion-forward interface that feels premium yet easy to use.
Seamlessly connected AI insights with real shopping links, bridging analysis and action.

What we learned

Throughout this project, we learned how to:

Build a complete React app using TypeScript with modular components and state management.
Connect and interact with AI APIs like Gemini for multimodal (text + image) analysis.
Integrate the Google Search API to dynamically fetch relevant clothing products.
Apply prompt engineering techniques to improve the accuracy and tone of AI-generated feedback and recommendations.
Coordinate between frontend and backend efficiently to deliver real-time visual feedback and outfit generation.

What's next for VYBE

Introduce 3D virtual try-on and real-time camera previews for more lifelike outfit visualization.
Expand into virtual reality (VR) which allows users to step into a virtual fitting room and view their AI-generated try-ons through VR goggles, making the experience fully immersive.
Build a social fashion feed where users can share outfits, exchange feedback, and discover community trends.
Partner with clothing brands and online retailers to make VYBE a complete fashion discovery and shopping hub.