PASSPORT.AI.

An ID photo can be turned down for small things: uneven light, a tilted head, the wrong background, a size that is slightly off. Many people end up in a photo booth or in an image editor. The brief was to take the photo anywhere, catch the common photo problems, such as no face, poor light, blur, a tilted head or closed eyes, before any credits are spent, and return a file cropped to the exact pixel size of the chosen document.

Flutter for iOS and Android, with the face check on the phone: Apple Vision on iOS and Google ML Kit on Android behind one platform channel, plus light and sharpness measured in Dart. A photo can be turned down for any of eight reasons, from no face to closed eyes, and the app explains the main problem in plain words with a tip for fixing it. Nothing is uploaded until the user taps Generate with AI. Then a reduced copy is uploaded and a Cloud Function charges the credits, asks a Gemini image model to apply the chosen background, outfit and retouching, and crops the result to the document's size. Credits sit in a server-side ledger and come back if generation fails. Every document has a size in millimetres or pixels, with the 35 × 45 mm ICAO size as the default, and custom sizes are allowed. The interface follows a paper-and-ink design language with viewfinder brackets, monospaced data and a stamp for approved photos.

Passport.AI covers the whole path in code, from the four-step flow (document, photo, style and export) to the Cloud Functions behind it. It covers 250 country and territory entries, six social-profile sizes and custom sizes, 10 outfits and 10 retouching options, and exports a JPEG, a PNG or an A4 PDF with cut lines that holds one, two, four or six photos per page. The app runs in English and Turkish, in light and dark themes, and its weekly, monthly and yearly credit plans are handled through the RevenueCat SDK. Deploying the backend and submitting to the stores come next.