Google I/O 2026 had a clear theme: making AI helpful for everyone in new ways. The event showcased new tools, models, and developer experiences. More importantly, it also did something different — it used those same technologies to run the event itself.
Takeaways 🚀
✅Gemini helped build I/O experiences
✅Crawler impact notes on products.
✅Google showcased practical AI applications
✅Human guidance shaped final outcomes
✅I/O highlighted helpful AI innovation
In other words, Google didn’t just announce AI innovations on stage. Teams used Gemini and other AI tools to design, produce, and support the event from the ground up. That included short films, visual branding, interactive games, live music experiences, a coffee pop-up, speaker graphics, and custom stickers.
With that in mind, this article breaks down how we used Gemini to build Google I/O 2026 — what was built, which tools were used, and why it matters to everyday people.
Here’s what you’ll learn:
- How AI helped create a short animated film about TPUs
- How Gemini shaped the visual identity of Google I/O 2026
- How music and games powered the pre-show
- How attendees interacted with AI in person
- How AI generated personalized title cards and stickers
Google I/O 2026 Used AI as Both Topic and Tool
Google I/O 2026 centered on making AI helpful for everyone. That idea ran through every product announcement and developer session. But beyond the stage, the event itself also put that idea to work.
AI didn’t just appear as a concept being discussed. Instead, Google teams used it as a practical production tool. They relied on Gemini and related systems to test ideas, explore design directions, build media assets, and speed up development.
That said, AI didn’t replace human teams. People still made the key creative decisions, guided the work, and handled quality control. AI, on the other hand, helped with exploration, speed, and repetitive tasks.
As a result, the event became something where the tools discussed on stage also helped shape what attendees saw, heard, played, and experienced.
Bringing Google I/O 2026 to Life Through Film
Perhaps the clearest example of AI in production is the short film “TPU Training Day.” It follows a group of TPUs getting ready for the computing demands of Google I/O 2026.
A TPU, or Tensor Processing Unit, is a custom chip Google built for AI and machine learning work. Rather than keeping that technical idea abstract, the film turned it into a fun animated story using puppetry, cardboard, and AI tools.
TPU Training Day: Where Human Craft Meets AI
The project started with a simple creative question: could a team use basic materials — cardboard and markers — and then use AI to animate them?
Working with director Laurie Rowan and Nexus Studios, the team combined several techniques:
- Puppetry
- Traditional animation
- Simple 3D animation
- AI-generated visual styling
- AI-assisted frame and sequence generation
Throughout the process, human craft stayed at the center. Puppets and handmade materials gave the film texture, movement, and personality. After that, AI tools helped transform that raw material into a polished animated short.
The production used Gemini Omni, Google AI Studio, experimental DeepMind models, and Nano Banana. Specifically, Nano Banana generated stylized first frames from raw footage. Meanwhile, a custom tool inside Google AI Studio helped the team test and compare frames at scale, keeping the visuals consistent throughout.
Why This Example Matters
Animation involves countless small decisions across thousands of frames. Keeping characters consistent is hard — especially with a new visual style. That’s exactly where AI made a difference.
AI helped the team move faster and stay consistent. Even so, people still drove the story, tone, and final look. Ultimately, the film shows one practical role for AI in creative production: helping teams go from rough material to finished output without losing the human touch.
In short, Gemini and related models acted as creative assistants — not replacements for the filmmakers.
Building the Google I/O 2026 Visual Brand Identity
Beyond film, AI also shaped the visual identity for Google I/O 2026. This covered everything from keynote graphics and signage to apps and digital materials.
The final identity used a four-color gradient, overlapping transparencies, and interlocking icons. Together, these elements created a consistent look across many different formats and surfaces.
How Gemini Supported the Design Process
To get started, the design team fed Gemini previous brand guidelines and five years of I/O recaps. This gave the AI context about the conference’s history, tone, and visual style.
As expected, early outputs weren’t ready to use. So the team ran small experiments, reviewed results, gave feedback, and used Nano Banana to explore icon styles. Over time, they developed a visual system where flat 2D icons could shift into textured 3D icons — connecting simple interface graphics with richer event visuals.
AI as a Design Exploration Tool
Think of this process as a faster way to sketch. Normally, designers create many drafts before picking a direction. With AI, the team could test far more options in far less time.
However, more options don’t automatically mean better design. People still chose what fit the brand, what worked across surfaces, and what felt clear to viewers. So while Gemini helped generate and refine the options, humans made the final calls.
Jellectronica: Turning Jellyfish Into Music
Moving to the pre-show, the I/O opening featured Jellectronica — a live music experiment created with the Monterey Bay Aquarium. It used moon jellyfish movement to shape generative music in real time.
Generative music changes based on live data or rules — not a fixed recording. In this case, jellyfish movement directly influenced how the music sounded.
How the Jellectronica Pre-Show Worked
The team used Google Colab, Google Coral NPU, Google Flow Music, Lyria 3 Pro, and Google Antigravity. Here’s how the system worked, step by step:
- A machine learning model learned to track jellyfish movement.
- That model ran on Google hardware built for AI tasks.
- The movement data then translated into musical signals.
- Google Flow Music and the Lyria API generated the sound.
- Finally, different movements affected different parts of the music — bass, chords, melody, or drums.
For example, more jellyfish moving in the “bass zone” made the bass louder and more active.
What This Shows About AI and Live Events
Jellectronica wasn’t just a visual display. Rather, it showed how AI can link live movement to generated sound — turning nature into a music controller.
This kind of experience makes events feel more alive and responsive. It also demonstrates how AI can bridge different types of information: motion becomes sound; data becomes art.
That connects directly back to the core idea behind Google I/O 2026 — making AI helpful for everyone by showing it through familiar, accessible experiences.
Infinite Scaler: AI-Generated Game Worlds
Also part of the pre-show was Infinite Scaler, a video game where players built levels as they played. Rather than using pre-made content, AI turned player descriptions into playable 3D worlds in real time.
The game used Google AI Studio, the Gemini API, Gemini Canvas, Google Antigravity, Lyria 3, and Nano Banana.
How Infinite Scaler Turned Prompts Into Levels
Players started by typing a prompt or uploading a reference image. From there, Nano Banana generated sprite sheets through the Gemini API.
A sprite sheet is one image that holds multiple visual elements for a game — characters, backgrounds, props, and more.
In addition, AI generated different types of maps to add depth and texture:
- Normal maps — simulate surface depth and shape
- Roughness maps — control how shiny or matte a surface looks
- Emission maps — add glowing effects
These maps helped the game turn 2D images into objects that looked three-dimensional inside a WebGL browser environment. On top of that, Lyria 3 generated all the in-game music.
Why Infinite Scaler Is Worth Noting
Normally, building game assets takes time and technical skill. Infinite Scaler showed how AI can compress that process significantly. A player could describe an idea, and the system would help turn it into something playable.
Still, AI didn’t do everything. Human-built systems organized the prompts, rendered the worlds, and kept the game running. In other words, AI generated the content — but people built the structure around it.
Antigravity Coffee Co.: AI You Could Taste
Shifting from digital to physical, Google I/O 2026 also included a real-world pop-up called the Antigravity Coffee Co. Attendees could design a latte with custom art and order it on the spot. They could even build their own version of the coffee-ordering app.
The experience used Flutter, Gemini Enterprise Agent Platform, Google Antigravity, and Nano Banana.
How the Coffee Experience Used AI
Rather than showing a fixed form, the app used generative UI — meaning the interface adapted in real time to each user’s interaction. Flutter delivered the app consistently across different devices, while Firebase connected the frontend to AI models and backend systems.
Nano Banana handled image and content generation. Meanwhile, Google Cloud and Firebase managed the app’s data, functions, and monitoring.
For attendees who wanted to go further, Google Antigravity’s agentic coding features let them build their own order apps. Agentic coding means AI can help plan, write, and revise code with more independence than a basic autocomplete tool.
Why This Experience Stood Out
Most importantly, the coffee pop-up made AI tangible. Attendees didn’t just read about AI or watch a demo — they interacted with it directly to order a drink, receive custom latte art, and even build an app.
That kind of hands-on experience makes abstract technology feel useful and approachable to a wide audience.
Speaker Title Cards: Personalized AI Visuals
In addition to physical experiences, Google also used AI to build custom title cards for each speaker at the event. These short visual clips introduced presenters on stage in a personalized way.
The tools included Gemini Omni, Google Flow, and Nano Banana Pro.
How the Title Cards Came Together
Take one example: a speaker appeared as a digital character riding Chrome Dino and dunking a basketball. To build this, the team first generated reference sheets using Nano Banana Pro.
Reference sheets keep characters and objects visually consistent across scenes. With those in place, the team used storyboarding to try different variations and add personal details.
Next, Google Flow and Veo helped prototype movement and animation. Gemini Omni in Google Flow then handled complex motion like sports sequences. Finally, editors trimmed, timed, and polished the generated footage into finished title cards.
AI-Generated Video Still Needs Human Editing
This example is a useful reminder: AI-generated video often still needs finishing work. AI creates motion and visual options, but editors still adjust timing, framing, and consistency.
For events like I/O, though, this approach saves significant time. Teams can build personalized visuals for many speakers without starting from scratch each time. Human review then keeps the results accurate and on-brand.
Custom Sticker Swag: Personalization at Scale
Finally, one of the most playful examples from the event was custom I/O stickers — generated and printed on the spot for attendees.
The tools used were Gemini and Nano Banana.
How the Sticker Experience Worked
Google built an interactive sticker game in a custom web app. Players had 20 seconds to catch falling prompts using an Android bot. Prompts came from over 100 categories — blueberries, disco balls, lasers, wood, and more.
Players then picked two prompts, or hit “I’m feeling lucky” for a random mix. The backend combined the two choices to generate a unique sticker design, which was printed immediately for attendees to take home.
Results could be anything from:
- A 3D “I/O” made of solid gold waffles
- A gummy bear motherboard
- Lasers fused with wood grain
- Two completely unrelated ideas merged into one design
What This Example Shows
Custom stickers may seem simple, but they reveal a broader principle. When the design rules are clear and fixed, AI can generate many unique variations very quickly.
Importantly, the system didn’t produce random images. It followed a defined layout, kept the I/O shape intact, and applied different materials or textures within that structure. That balance between variety and constraint is, in fact, a key strength of practical AI workflows.
What These Examples Reveal About Gemini and Event Production
Looking across Google I/O 2026, Gemini and related AI tools took on several roles: generating ideas, creating assets, prototyping experiences, and powering interactive systems. Though the use cases varied widely, they shared a few important patterns.
AI Enabled Faster Iteration
Creative work involves a lot of trial and error. Fortunately, AI made it faster to test styles, generate options, and compare results. This showed up clearly in the brand identity work, speaker title cards, and sticker designs.
AI Bridged Different Media Types
In several cases, AI transformed information from one format into another. For instance, Jellectronica turned movement into music. Infinite Scaler turned text into game worlds. The coffee app turned user input into latte art. In each case, AI acted as a bridge between text, images, motion, sound, and code.
AI Worked Best With Clear Human Direction
Across all examples, the most effective results came when AI worked inside a structured process. People wrote the prompts, set the constraints, provided reference materials, defined creative goals, and reviewed the outputs.
Put simply, AI didn’t remove the need for planning, design, or editing. Instead, it changed how some of that work happened.
AI Made Experiences More Personal
Finally, custom stickers, speaker title cards, player-generated worlds, and latte art all relied on AI to create something unique for each person or moment. That kind of personalization makes large events feel more human — and helps people understand AI through direct experience, not just explanation.
Conclusion: Google I/O 2026 Put AI to Work
Google I/O 2026 carried a clear message: make AI helpful for everyone in new ways. As this article shows, the production behind the event lived up to that idea in very concrete ways.
Teams used Gemini and related AI tools not just to fill an announcements agenda, but to actually build the event. TPU Training Day showed AI supporting animation. The visual brand identity showed AI as a design partner. Jellectronica connected jellyfish movement to live music. Infinite Scaler turned player prompts into game worlds. Antigravity Coffee Co. made AI something you could order and drink. Speaker title cards and custom stickers, meanwhile, showed AI working at a personal scale.
Taken together, these examples offer a grounded view of what AI can do when people guide it well. The tools expanded what teams could create and test — and human judgment shaped the final result every step of the way.
