FitFind

Paste a TikTok or Instagram link and get shoppable results matching the outfit and aesthetic.

• Next.js
• FastAPI
• Gemini Vision
• Python
• TypeScript

A fashion discovery app that turns a pasted TikTok/Instagram URL or uploaded screenshot into shoppable product results. yt-dlp and OpenCV extract key frames from the video server-side, then a single Gemini Vision call detects individual clothing items and classifies the overall aesthetic (e.g. 'gorpcore', 'quiet luxury'). Results are split into two sections — Exact Items, matched to the specific detected pieces, and Related Items, matched to the broader aesthetic — with sort and price filtering handled client-side after a single Serper API call per query to avoid redundant re-fetching. Built with a FastAPI backend and Next.js frontend, with retry logic for Gemini rate limits and gender-aware prompting for more accurate item classification.