source: google ai: how we used gemini to build google i/o 2026
level: technical
google's i/o 2026 conference showcased ai not just in announcements but in production. the marketing team used gemini models, nano banana, and other tools to speed up creative work. they made a short film with cardboard puppets and ai-generated frames. they designed the event's visual identity by feeding past brand guidelines into gemini and iterating with nano banana. the goal was to blend human artistry with experimental technology, freeing up time for creative decisions.
the pre-show included a generative music experiment that turned jellyfish movements into sound. a yolo8 model tracked the jellies, and lyria 3 pro generated music based on their motion. another pre-show piece was a video game where players built 3d worlds from 2d image prompts. nano banana created sprite sheets, and gemini api handled level planning. attendees also coded a countdown sequence using canvas or ai studio. a pop-up coffee shop let people design latte art and build their own ordering apps with flutter and gemini enterprise agent platform.
speaker title cards were custom-generated with nano banana pro and gemini omni. the team created reference sheets, then used veo and gemini omni in google flow to prototype animations. they composited the raw motion into polished titles. an on-site sticker game let attendees catch falling prompts and fuse them into custom designs. the backend used nano banana to combine two random concepts, like gold waffles or gummy bear motherboards, and printed the results instantly. these experiments showed how ai can handle repetitive tasks while keeping human craft central.
why it matters: the project demonstrates practical ai integration for creative and event production, offering a template for data scientists and developers to automate asset generation and rapid prototyping.
source: google ai: how we used gemini to build google i/o 2026