A simple Flutter application that demonstrates the power of on-device, offline, multimodal AI using the flutter_gemma package. This app can translate a Japanese menu from an image, even without an internet connection.
This project was built as a proof-of-concept for my Medium article, which details the entire journey and implementation. Using Gemma for Flutter apps: First steps with on-device AI in Flutter using flutter_gemma package
- Offline First: The entire AI interaction happens on-device, with no internet connection required.
- Multimodal Input: Understands both images and text prompts in a single query.
- Powered by Gemma 3N: Utilizes Google's lightweight and powerful Gemma 3 Nano model.
- Simple, Single-Screen UI: A minimal, easy-to-understand chat interface.
To run this project locally, follow these steps:
Add Your Hugging Face Token:
You will need a Hugging Face access token to download the Gemma model. Open lib/data/gemma_downloader_datasource.dart and replace the placeholder with your own token:
const String accessToken = 'YOUR_HUGGING_FACE_TOKEN_HERE';Flutter: The UI toolkit for building the application.
flutter_gemma: The core package for running Gemma models on-device.image_picker: For selecting images from the camera or gallery.
Any contributions is welcomed! 🙏