Add Generative AI capabilities to your Android app
summary
Concrete roadmap and runnable tools: Gemini Nano runs on Pixel 8/9, Samsung S24/Flip6/Fold6 and select Motorola phones and powers the Pixel Screenshots feature that finds package numbers via natural‑language search, proving on‑device multimodal search works. MediaPipe LLM task and Light RT let apps run open models (Gemini family, Falcon) on device after conversion plus hardware acceleration. Cloud option uses the Gemini API (Gemini 1.5 Flash 1 million token context, 1.5 Pro 2 million token context), integrates via Firebase in ~10 lines and supports multimodal inputs and function calling.
AI-written summary · about
chapters (12)
- 0:00GenAI on Android intro
- 1:22Use cases and GenAI dev cycle
- 3:51On-device GenAI benefits
- 7:30Gemini Nano and AI Core
- 9:17MediaPipe LM and Light RT
- 11:03On-device GenAI limits
- 11:25Cloud GenAI with Gemini API
- 14:16Gemini API features
- 16:00Function calling and integrations
- 17:06Gemini in Android Studio
- 19:15GenAI app debugging
- 20:20Early access programs and resources
related talks
39mAdd Generative AI capabilities to your Android app
41mBuild intelligent Android apps with Google's AI
30mGenAI on Android with Gemini Nano and Firebase
18mLet's test Gemini, Google's AI in an Android application!
41mGen AI for Android Developers
10mAI in your pocket: Building intelligent Android apps
56mBuilding Multimodal Apps with Gemini 2 0 and Gemini SDK
39mGetting Started with GenAI for Android Apps