💡 What it does

FitCheck is a Chrome extension that lets users virtually try on clothes directly on e-commerce websites. It detects product images, adds a “Try On” button, and uses Google Gemini with a Firebase backend to generate realistic try-on previews.

Before processing, the user’s uploaded image is validated through the Prompt API, which checks whether it’s suitable and properly formatted for virtual try-on. The result appears instantly in a sleek modal where users can download or share their look — all without leaving the website.

🛠️ How we built it

FitCheck was built as a Manifest V3 Chrome extension with popup, content, and background scripts.

Popup: users upload avatars via file input or drag-and-drop. The Prompt API checks image suitability, then Gemini performs on-device validation (with a RAM check via chrome.system.memory). Valid avatars are stored locally.
Content script: auto-detects clothing pages, injects “Try On” buttons, and handles image clicks synced with the popup.
Background: routes messages and relays VTO/fetch requests through a proxy to bypass CORS.
Backend: a Firebase Cloud Function calls Gemini for virtual try-on, using detailed prompts for pose, lighting, and garment details. A Flask proxy on Cloud Run manages CORS and base64 conversions, with secrets in Google Secret Manager.
UI: designed with gradient pill buttons, fade/slide modals, and soft #EFBFFF accents.
Supports major sites like Nike, Zara, Amazon, H&M, and Adidas, with avatars and settings saved in chrome.storage.local.

⚙️ Challenges we ran into

Handling CORS restrictions and Manifest V3 limitations during extension development.
Realizing the Prompt API requires 16 GB RAM, forcing a redesign of on-device validation.
Deploying and debugging the proxy and VTO Cloud Functions on Cloud Run, including CORS and permission errors.
Coordinating team collaboration, syncing code and environments across multiple members.
Navigating through endless, cryptic Chrome and API errors, testing our patience and persistence.

🏆 Accomplishments that we're proud of

Building a working end-to-end virtual try-on inside a browser.
Seamlessly integrating Gemini AI, Firebase, and Chrome APIs.
Designing a clean, intuitive UX that feels native to e-commerce sites.
Achieving privacy-friendly, on-device image validation.
Overcoming CORS, memory, and deployment challenges under tight time pressure.

🧠 What we learned

How to bridge browser limitations with backend AI orchestration.
How to design around Manifest V3 constraints and CORS restrictions.
Effective collaboration across frontend, backend, and AI integration.
Building production-ready workflows with Gemini and Chrome extensions.

🚀 What's next for FitCheck

Expand beyond fashion to include accessories, glasses, and furniture.
Build an outfit builder for mixing and matching items from multiple retailers.
Add personalized recommendations using user style data and try-on history.
Enhance try-on realism with improved pose, lighting, and texture understanding.
Support more e-commerce platforms and optimize performance for lower-RAM devices.