IsItFake is a credit-based SaaS that tells you whether an image or video was generated by AI. Users upload media or paste a URL; the backend pulls the file, dispatches it to a stack of detection providers (Sightengine, Hive AI, and more), aggregates their scores into a single confidence verdict, and returns a Real / Fake call plus the per-provider breakdown. Mobile-first PWA so it installs cleanly on Android and iOS, and the same scan history is available in the dashboard for logged-in users.

Architecture: Next.js 16 App Router app with server actions in src/lib/actions/, Prisma 7 against Postgres, and NextAuth v5 (JWT, Google + Credentials) for accounts. sharp handles thumbnailing and pre-flight format normalization before media leaves the origin. Pricing is metered in credits — images cost one credit per scan, videos cost three — with credit packs and a subscription tier sold through the same checkout flow used elsewhere in the portfolio.

The interesting parts: a provider-adapter layer that lets new detection backends drop in behind a single interface, a queue-friendly scan model that can hold partial results until all providers report (or time out), and a graceful-degradation path so a single dead provider doesn't fail the whole verdict — the aggregator just lowers confidence and flags the missing signal. URL-input mode validates and re-fetches the media server-side so user-supplied links can't bypass content-size or MIME restrictions.