"Flutter Gemini Image Analyzer: Camera Photos to AI"
"Build a Flutter Gemini image analyzer that captures camera photos, compresses them to control token costs, and sends them to the Gemini API with firebase_ai. Complete runnable code, a decision table for picking your SDK, and the production mistakes to avoid."
Point your phone at a plant, a receipt, or a broken error message on a monitor, and get an AI description in about two seconds — inside your own Flutter app, with no API key baked into the client. A Flutter Gemini image analyzer is one of the highest-value AI features you can ship: camera in, understanding out, one model call. This guide builds it end to end — camera capture, smart compression that keeps token costs sane, the Gemini call with firebase_ai , a complete analyzer screen, and the production mistakes that break everyone the first time. What you'll build A camera-first photo capture flow with image_picker (gallery as a fallback) Automatic compression (max 1024px, 85% quality) that cuts image token costs by ~70% A clean ImageAnalyzer service using firebase_ai 4.0.0 with the gemini-2.5-flash model A complete analyzer screen: preview, prompt field, loading state, result card, error banner Production error handling that maps Firebase exceptions to human messages The cost and security knowledge to ship this without leaking your API key Step 1 — Wire up firebase_ai and image_picker Add the dependencies. firebase_ai is the Firebase AI Logic SDK — it calls Gemini through your Firebase project, so there is no raw API key sitting in your Dart code (more on why that matters in Step 5). # pubspec.yaml dependencies: flutter: sdk: flutter firebase_core: ^4.0.0 firebase_ai: ^4.0.0 # Firebase AI Logic SDK (Gemini backend) image_picker: ^1.2.1 # camera + gallery capture Initialize Firebase before your app runs, then create the analyzer service: import 'package:firebase_ai/firebase_ai.dart'; import 'package:firebase_core/firebase_core.dart'; Future<void> main() async { WidgetsFlutterBinding.ensureInitialized(); await Firebase.initializeApp(); // add…