Brief IA

Google I/O 2026: AI Gemini Revolutionizes the Event

🔬 Research·Tom Levy·

Google I/O 2026: AI Gemini Revolutionizes the Event

Google I/O 2026: AI Gemini Revolutionizes the Event
Key Takeaways
1Google I/O 2026 utilized AI to innovate and create, with tools like Gemini Omni and Nano Banana.
2The short film "TPU Training Day" was produced by combining puppetry and AI, illustrating the fusion of art and technology.
3The visual identity of the event was developed through a collaboration between the Google team and the Gemini models, resulting in dynamic icons.
💡Why it mattersThe integration of AI in the organization of Google I/O 2026 demonstrates its potential to transform creativity and event efficiency.
Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

Google I/O 2026: AI at the Heart of Innovation

At the Google I/O 2026 event, the focus was on using artificial intelligence to make AI accessible and useful in unprecedented ways. Not only did Google showcase its latest innovations in AI, but the company also employed these technologies to orchestrate the event itself, marking a significant milestone in how AI can transform creation and innovation.

This year, Google took on the ambitious challenge of using the same AI tools it presented on stage to innovate and create more efficiently. By combining human artistry with experimental technology, Google was able to move faster than ever, prototyping in real-time. One of the most striking examples of this approach is the film "Timmy TPU," which illustrates how these tools unleash creativity and alleviate repetitive tasks, allowing the team to focus on their core skills. The goal is to create an event so captivating that the audience forgets about the use of AI, thereby demonstrating the possibilities offered by these technologies.

AI and Cinema: "TPU Training Day"

To illustrate this fusion of technology and creativity, Google produced a short film titled "TPU Training Day." This project was created in collaboration with director Laurie Rowan and Nexus Studios, using simple materials like cardboard and markers, enhanced by AI. Tools like Nano Banana were used to stylize the images, while Gemini Omni helped integrate basic animation with stylized images, all while preserving the authenticity and human imperfections that make puppet films so charming.

The process began with capturing the performances of characters using puppets and simple 3D animation, offering complete control over framing and camera movement. Nano Banana was then used to generate stylized images from these raw sequences. To ensure image consistency, a custom tool was developed in Google AI Studio, allowing for large-scale image testing and ensuring perfect matches before generating the final sequences.

The fusion of basic animation and stylized images was achieved through Gemini Omni and other experimental models, elevating the film to a cinematic level while preserving the original human intent. The preservation of small human imperfections is essential for adding charm to puppet films, and the AI pipelines were designed to protect these details.

Visual Identity and AI

The visual identity of Google I/O 2026 was crafted through a collaboration between the Google team and the Gemini AI models. By using previous brand data and summaries from past events, the team experimented with Nano Banana to create dynamic icons that transform from 2D to 3D, ensuring visual consistency across all event platforms.

The process began by feeding the Gemini models with old brand guidelines and five years of I/O summaries. The initial outputs were not entirely satisfactory, leading to the realization of micro-experiences. By generating new images and iteratively feeding them into Nano Banana with feedback, the team was able to explore different icon styles. The final result was flat 2D icons that dynamically transformed into hyper-textured 3D icons, creating a cohesive brand expression across keynotes, physical signage, and digital applications.

Immersive Experiences: Jellectronica and Infinite Scaler

The pre-show of I/O introduced "Jellectronica," a generative musical experience that transformed the movements of jellyfish from the Monterey Bay Aquarium into music. Using tools like Google Flow Music and Lyria 3 Pro, the team created an immersive sound experience. Meanwhile, "Infinite Scaler" allowed players to generate game levels in real-time, illustrating the application of AI in video game development.

For "Jellectronica," a YOLO8 model was trained in Google Colab and run on Google's Coral NPU to track the movement of jellyfish, which controlled the music created using Google Flow Music and the Lyria API. For example, the more jellyfish there were in the low section, the stronger and more energetic the bass became. Mass stems were coded in Google Antigravity to automate the production of musical stems, such as bass, chords, melody, and percussion.

"Infinite Scaler," another part of the pre-show, was a video game where players competed and generated levels while playing. To achieve this, Nano Banana was used to generate sprite sheets from user prompts and reference images via the Gemini API. Foreground elements were sent back to Nano Banana to generate normal, roughness, and emission maps, allowing for depth inference and texture mapping on a 3D cardboard box rendered in WebGL before adding them to the global world stack. Google AI Studio was used for rapid prototyping before moving to Google Antigravity for development, and the game's music was entirely generated with Lyria 3.

Creativity and AI: Title Cards and Stickers

Every speaker at I/O received personalized title cards, generated by AI models like Gemini Omni. For instance, an animation of Josh Woodward, VP of Google Labs, was created using elements generated by Nano Banana Pro. Additionally, custom stickers were printed for attendees, thanks to an interactive web application using AI to merge creative choices into unique designs.

For the title cards, basic elements like ingredient reference sheets were generated with Nano Banana Pro. These ingredients were used to create storyboards, experimenting with variations and adding personal details. In Google Flow, Veo was used to help prototype actions and generate animations like a dunk. Animations were also generated with Gemini Omni in Google Flow, which was particularly useful for handling complex sports movements. Detailed prompts ensured that the AI outputs remained consistent with the reference sheets. Finally, the raw movements from the generated videos were composed and temporally remapped into polished titles.

For the stickers, an interactive game was built on a custom web application. Players had 20 seconds to catch falling prompts using an Android bot, with over 100 categories of prompts available. Players selected two prompts or pressed "I'm feeling lucky" for a random mix. The backend, using Nano Banana for Gemini and Android, merged these choices to create a highly personalized I/O sticker design, such as a 3D "I/O" made of solid gold waffles or a motherboard made of gummy bears. The designs were printed immediately for participants to collect.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.