TryOn AI — virtual outfit try-on
Built a full-stack AI platform that generates a realistic photo of the user wearing any garment, on a credit-based Stripe billing model with daily reward spins to drive retention.
Shoppers had no way to visualize a garment on their own body before buying, driving high return rates and cart abandonment — and any AI try-on feature needed to be believable and monetizable, not a one-off demo, while keeping generation time reasonable despite the heavy compute involved.
A credit-based AI platform where users upload a photo of themselves and a photo of a garment; an AI model generates a realistic composite image in under 1.5 minutes. Credits are purchased via Stripe, and a 24-hour spin-wheel rewards loop nudges users to return daily and keeps engagement — and repeat purchases — high.
How it was built.
Built the core generation pipeline in Python — user photo + garment photo in, AI-composited try-on image out — served via FastAPI, orchestrating calls to a third-party AI image-generation API and tuned to return results in roughly 1–1.5 minutes.
Built the React frontend for photo upload, garment selection, a live 'generating...' status state, and a result gallery.
Designed a credit-based usage system: each generation deducts credits, with tiered credit packs for different user needs.
Integrated Stripe for credit-pack purchases, handling checkout, webhooks, and real-time balance updates.
Built a 24-hour spin-wheel reward feature — one spin per day for bonus credits or discounts, designed to drive daily return visits and reduce churn.
Added usage tracking (credits spent, generations run, spin history) so the team can see what's actually driving retention and revenue.
Users get a realistic try-on image back in under 1.5 minutes despite the heavy AI compute involved, and the daily spin-wheel keeps users returning — turning a one-time novelty into a recurring habit.