Gemini and Internal Models Power Film, Branding and Pre-show at I/O 2026: what developers gain

At I/O 2026, engineering and creative teams used Gemini together with several internal generative models to produce the event’s short film, visual identity and AI-driven pre-show, demonstrating a multi — model pipeline that accelerated real-time prototyping while keeping human creative control. The work combined human craft — puppetry, simple 3D animation and director — led decisions — with model — driven tooling to speed repetitive tasks and let humans focus on higher — value creative choices.

The short film “TPU Training Day” (aka “Timmy TPU”) was produced through a hybrid pipeline that began with puppetry and simple 3D animation for base performances. Nano Banana generated stylized first frames, and a custom tool inside Google AI Studio tested frames at scale for pixel — level consistency. Those outputs were merged with Gemini Omni and experimental DeepMind models to enhance the footage while intentionally preserving the small, handcrafted imperfections that give puppet — based media their charm. Director Laurie Rowan and Nexus Studios collaborated on the project.

For the I/O 2026 visual identity, teams fed Gemini models past brand guidelines plus five years of I/O recaps, then ran micro — experiments with iterative feedback into Nano Banana to explore iconography and texture. The resulting system centers on a four-color gradient with overlapping transparencies and interlocking icons; flat 2D icons can dynamically transform into hyper — textured 3D versions for use across keynotes, physical signage and digital apps, enabling a single generative pipeline to produce assets at multiple scales and formats.

The event pre-show, billed as “Jellectronica,” used a separate stack of AI and runtime tools to generate immersive elements. The blog lists Google Antigravity, Google Colab, Google CoralNPU, Google Flow Music and Lyria 3 Pro as part of the experiential pipeline. The producers framed these tools as components of a coordinated production flow rather than standalone consumer features, linking model outputs to real-time runtimes and stage presentation systems.

Beyond individual models and tools, the production workflow emphasized rapid iteration, custom orchestration and tooling that let teams test model outputs at scale. That orchestration was designed to offload repetitive technical tasks to AI while retaining human oversight on creative decisions and preserving the deliberate imperfections that make handcrafted media distinctive. The approach showcases how multi — model systems and studio tooling can shorten production cycles and shift creative teams toward higher‑value work.