A virtual try on app that allows a prompt-based, gamified, modular, and social virtual try on experience to explore endless personal fashion possibilities.
Started on July 2nd, 2025, inspired by the challenge of visualizing fashion choices. Built the core MVVM architecture in Swift with basic features: avatar photo upload, closet management, and simple placeholder overlays for try-on visualization.
Improved rendering quality and refined the UI. Integrated OpenAI API for AI-powered style generation and replaced placeholder overlays with real clothing images for more realistic try-on visualization.
First major version with realistic try-on capabilities. Deployed image preprocessing models (CIHP, OpenPose, U2Net) and VTON-HD on Nebius virtual machine with GPU access. Integrated Amazon S3 for image storage and built custom API for real-time communication between the mobile app and cloud services.
Pivoted to Gemini Flash 2.5 API for simpler and more efficient try-on image generation. Enhanced the AI style explanation feature to provide clearer fashion insights.
Added multi-garment switching capability for more realistic outfit visualization. Secured first official collaboration with Duke University Store, marking an important step toward real-world application.
Leverages advanced AI models including VTON-HD and Gemini Flash 2.5 to generate realistic virtual try-on images, allowing you to visualize outfits before wearing them.
Prompt-based outfit creation using AI to suggest and generate fashion combinations. Get intelligent style explanations and fashion insights tailored to your preferences.
Organize and manage your wardrobe digitally with photo uploads. Easily browse through your clothing items and experiment with different combinations.
Seamlessly switch between multiple garments for comprehensive outfit visualization. Mix and match different pieces to create perfect combinations.
Built with Swift using MVVM architecture for iOS. Integrates OpenAI and Gemini Flash 2.5 APIs for AI-powered features. Utilizes computer vision models (CIHP, OpenPose, U2Net, VTON-HD) deployed on cloud infrastructure with GPU acceleration for older versions. Amazon S3 for scalable image storage and custom REST API for seamless client-server communication.