gallery  by google-ai-edge

This application showcases on-device Machine Learning and Generative AI use cases, allowing users to explore, experience, and evaluate cutting-edge models locally on their mobile devices. It targets users interested in experiencing AI capabilities offline, developers looking to test on-device models, and researchers interested in performance benchmarks. The primary benefit is enabling powerful AI interactions without an internet connection.

How It Works

The gallery leverages Google AI Edge technologies, including the LiteRT runtime for optimized model execution and an LLM Inference API for on-device Large Language Models. It integrates with Hugging Face for model discovery and download, enabling users to switch between various models. The app supports offline functionality, local model testing, and provides real-time performance insights like time-to-first-token and decode speed.

Quick Start & Requirements

• Install: Download the latest APK from the project.
• Prerequisites: Android device (iOS coming soon).
• Resources: Requires model downloads; performance depends on device hardware.
• Links: Project Wiki, Google AI Edge Documentation

Highlighted Details

• Fully offline, on-device Generative AI execution.
• Supports model switching from Hugging Face.
• Features include "Ask Image" (image-based Q&A), "Prompt Lab" (single-turn LLM tasks), and "AI Chat" (multi-turn conversations).
• Provides real-time performance benchmarks (TTFT, decode speed, latency).

Maintenance & Community

This is an experimental Alpha release. Feedback and feature suggestions are encouraged via provided links.

Licensing & Compatibility

• License: Apache License, Version 2.0.
• Compatibility: Permissive license suitable for commercial use and integration with closed-source applications.

Limitations & Caveats

The application is an experimental Alpha release, indicating potential instability and ongoing development. iOS support is listed as "coming soon."