| .github/ISSUE_TEMPLATE | ||
| Android | ||
| .gitignore | ||
| CONTRIBUTING.md | ||
| LICENSE | ||
| model_allowlist.json | ||
| README.md | ||
Google AI Edge Gallery ✨
-lightgrey){kind=icon}
Dive into a world of creative and practical AI: chat with advanced models, get insights from your images, experiment with prompts, and see real-time performance. All this, without needing an internet connection once models are downloaded.
🚀 What Can You Do?
The Gallery showcases a variety of on-device AI capabilities:
| Overview & Model Selection | Visual Q&A (Ask Image) |
|---|---|
| Explore available models and dive into different AI experiences. | Upload an image and ask questions about it, get descriptions, or identify objects. |
| Prompt Lab (Single-Turn LLM) | AI Chat (Multi-Turn Conversation) |
|---|---|
| Summarize text, rewrite content, generate code, or use freeform prompts. | Engage in dynamic, multi-turn conversations with an AI. |
Suggestion: Consider creating a short GIF showcasing these features in action and embedding it here for a more dynamic preview!
✨ Core Features
- 📱 Run Locally, Fully Offline: Experience the magic of Generative AI without an internet connection. All processing happens directly on your device, ensuring privacy and speed.
- 🤖 Diverse Model Zoo: Easily switch between different foundation models (e.g., from Hugging Face) and compare their performance and capabilities.
- 🖼️ Visual Question Answering (Ask Image): Go beyond simple object detection. Upload an image and ask complex questions, get detailed descriptions, or solve visual puzzles.
- ✍️ Creative Prompting (Prompt Lab): A sandbox for your ideas! Summarize articles, rewrite paragraphs, generate code snippets, or explore other single-turn LLM use cases.
- 💬 Conversational AI (AI Chat): Engage in fluid, multi-turn conversations. Explore the model's reasoning, memory, and conversational abilities.
- 📊 Real-time Performance Insights: Monitor key metrics like Time To First Token (TTFT), decode speed (tokens/sec), and overall latency to understand model efficiency on your device.
- 🧩 Bring Your Own Model (BYOM): Advanced users can easily test their own local LiteRT
.taskmodels. - 🔗 Developer-Focused Resources: Quick access to model cards, source code, and relevant documentation.
🏁 Get Started in Minutes!
- Download the App: Grab the latest
ai-edge-gallery.apkfrom our Releases Page. - Install & Explore:
- Enable installation from unknown sources on your Android device.
- Install the APK.
- Launch the app and download your first model!
- Need Help? For detailed installation instructions (including for corporate devices) and a full user guide, head over to our Project Wiki.
🛠️ Technology Powering the Gallery
This app leverages Google's cutting-edge on-device ML technologies:
- Google AI Edge: The core suite of APIs and tools enabling powerful on-device machine learning.
- LiteRT (MediaPipe GenAI): A lightweight, high-performance runtime optimized for executing Generative AI models directly on mobile devices.
- LLM Inference API: Specifically designed for running Large Language Models efficiently on-device, powering features like chat and prompt lab.
- Hugging Face Integration: Seamlessly discover, download, and utilize a wide range of pre-trained models from the Hugging Face Hub.
🤝 We Value Your Feedback!
This is an experimental Alpha release, and your input is crucial for shaping its future.
- 🐞 Found a bug? Help us improve by reporting it here.
- 💡 Have an idea or feature request? We'd love to hear it! Suggest it here.
- 🤔 Questions or discussions? Join the conversation on our Discussions Tab (if enabled) or raise an issue.
📄 License
Licensed under the Apache License, Version 2.0. See the LICENSE file for details.
🔗 Useful Links
- Project Wiki (User Guides & FAQs)
- LLM Inference guide for Android (Google AI Edge Docs)
- Google AI Edge Documentation
- Hugging Face LiteRT Community (for discovering compatible models)
Made with ❤️ by the Google AI Edge Team.