Building Multimodal Apps with Gemini 2.0 and Gemini SDK
summary
Workshop shows hands-on labs using the Gemini SDK and Vertex AI via Colab to call multimodal APIs (image, video, audio, text), deploy models, and get API keys; demo uses Gemini 2.5 with a unified SDK, an auto-selecting model optimizer endpoint, and a thinking budget to cap internal “thinking” tokens. Claims include million-plus context windows, video extension/animation tools, and up-to-98% accuracy on some J-style problems; attendees run code, enable APIs, and test grounding with Google Search.
AI-written summary · about
chapters (24)
- 0:00Introduction to talk
- 1:08Live demo of Project Astra
- 5:51Gemini 2.5 model overview
- 7:42Gemini 2.5 features and capabilities
- 12:02Gemini Code Assist demo
- 13:23Multimodal media generation
- 14:34Safety and legal aspects
- 17:49Video generation tools
- 22:52Thinking models explained
- 26:24Model benchmarks and accuracy
- 28:49Hands-on lab setup
- 32:55Claiming credits and project creation
- 41:02Lab walkthrough start
- 43:32Google Colab environment
- 46:02Gemini API SDK setup
- 47:28First API call and response
- 51:40Multimodal input demo
- 52:47Long context window benefits
- 53:35System instructions and persona
- 55:10Preventing hallucinations with tools
- 57:39Function calling demo
- 58:02Gemini platforms overview
- 59:25AI Studio and resources
- 1:00:07Closing and innovators program
related talks
01h 45mAI Engineering with the Google Gemini 2.5 Model Family
18mIntroduction to Gemini APIs and AI Studio
18mLet's test Gemini, Google's AI in an Android application!
37mIntroduction to the Gemini API for Web Apps
54mGemini, Google's Large Language Model
51mGemini, Google's Large Language Model
01h 34mHands-on AI workshop: Multimodal agents
16mOpen models in the Gemini era