How we used Gemini to build Google I/O 2026

Google I/O 2026 served as a definitive showcase for how artificial intelligence is becoming a helpful, integrated partner in human creativity. While the event focused on the latest breakthroughs in AI, the production team took a meta-approach: they utilized the very tools being announced on stage to conceptualize, design, and execute the event itself.

This year, the goal was to out-innovate and out-create by blending human artistry with experimental technology. By offloading mundane technical tasks to AI, the team reclaimed valuable time for the creative work that only humans can provide. The result was a seamless experience where the technology faded into the background, leaving the audience to focus on the innovation rather than the machinery behind it.

AI x Film: Bringing "Timmy TPU" to Life

One of the most ambitious projects at I/O 2026 was the short film "TPU Training Day," featuring a character affectionately known as "Timmy TPU." The project sought to answer a fundamental question: Could simple, tactile materials like cardboard and markers be transformed into a cinematic experience using AI?

Working alongside director Laurie Rowan and Nexus Studios, the team utilized a hybrid pipeline:

• Puppetry and Animation: The team began with traditional physical puppetry and basic 3D animation to maintain control over framing and camera movement.
• Nano Banana Integration: Raw footage was processed through Nano Banana, which generated stylized frames. To ensure consistency, the team developed a custom tool within Google AI Studio to test frames at scale.
• Cinematic Polish: The final sequences were merged using Gemini Omni and various experimental models.

The goal was never to replace the human touch; rather, it was to enhance the charm of the puppets while maintaining the "human imperfections" that give such films their unique character.

Defining the Visual Identity

The visual brand identity for I/O 2026—a sophisticated four-color gradient with interlocking icons—was the result of a deep, iterative collaboration between designers and Gemini models.

The process involved feeding the AI five years of historical I/O brand guidelines and event recaps. While initial outputs required refinement, the team utilized micro-experiments to steer the model toward the desired aesthetic. By feeding outputs back into Nano Banana, they were able to evolve flat 2D icons into hyper-textured 3D assets that maintained a cohesive look across keynotes, signage, and digital applications.

"The output image must perfectly preserve the exact pose, camera angle, scale, and lighting of the input icon. The final result should be the icon from the reference, but with the new, detailed texture applied." — Internal prompt strategy for icon styling.

Immersive Experiences and Generative Music

The I/O pre-show pushed the boundaries of real-time interaction, most notably with "Jellectronica," a generative music experiment created in partnership with the Monterey Bay Aquarium.

• Jellectronica: By training a YOLO8 model in Google Colab and running it on the Coral NPU, the team tracked the movement of moon jellies. These movements were translated into sound via Lyria 3 Pro and Google Flow Music, where the density of jellyfish in the tank dynamically adjusted the bass and melody of the performance.
• Infinite Scaler: This interactive game allowed attendees to generate 3D worlds on the fly. Using Nano Banana and the Gemini API, the system converted 2D user prompts into sprite sheets, which were then mapped onto 3D geometry in WebGL.
• Code the Countdown: A global challenge invited developers to design numbers using Canvas or AI Studio, which were then stitched together into a collaborative, code-powered countdown.

The Antigravity Coffee Co. Pop-up

The I/O experience extended to the physical world through the Antigravity Coffee Co., an app that allowed attendees to design custom latte art. This project served as a masterclass in Generative UI.

Using Flutter and the A2UI protocol, the team built adaptive interfaces that shifted in real-time based on user interaction. Instead of static forms, the app utilized Firebase to bridge the frontend with Nano Banana, allowing for complex reasoning and content generation. This architecture ensured a high-quality, zero-latency experience across various hardware, demonstrating the power of modern cloud-native AI development.

Creative Delight: Speaker Cards and Sticker Swag

Even the smallest details, such as speaker title cards, were treated as opportunities for AI-driven storytelling. For instance, the title card for Josh Woodward, VP of Google Labs & Google Gemini, featured a digital version of himself riding a Chrome Dino and performing a slam dunk.

The production pipeline for these assets was rigorous: 1. Reference Sheets: Nano Banana Pro generated ingredient reference sheets to maintain character consistency. 2. Prototyping: Google Flow and Veo were used to storyboard actions and generate complex movements. 3. Compositing: Raw motion from generated videos was time-remapped and composited into the final, polished title cards.

Attendees also engaged with a custom sticker-generation game. By selecting two prompts—ranging from "disco balls" to "gummy bear motherboards"—users triggered a backend process where Nano Banana fused the concepts into a unique, printable 3D design.

Key Takeaways: The Future of Production

The production of Google I/O 2026 serves as a blueprint for how organizations can integrate generative AI into their workflows. By focusing on the following principles, the team successfully navigated the shift:

• Human-Centric Design: AI was used to handle the "mundane" tasks, allowing human creators to focus on the artistic intent and emotional resonance of the content.
• Iterative Prototyping: The use of Google AI Studio and Google Flow allowed for real-time adjustments, significantly shortening the feedback loop.
• Consistency at Scale: By building custom tools to enforce constraints on lighting, pose, and texture, the team ensured that AI-generated assets remained professional and on-brand.

As we move forward, the lesson from I/O 2026 is clear: the most effective use of AI is not to replace the creative process, but to act as a force multiplier for human imagination. When the technology is used correctly, the audience doesn't just see the AI—they see the magic it helps create.