A bigger garage for everyone: How we built a music video with AI and Weezer to kick off Next

Sarah Kennedy
Vice President, Marketing, Google Cloud
When the lights came down at the start of Google Cloud Next this morning, we kicked things off with a reminder of where we started — and a vision of where we’re going.
As the iconic opening chords of Weezer’s "In the Garage" echoed through the arena, we were transported back to a cluttered, unmistakable Silicon Valley garage. Over the next few minutes, attendees took a visual journey from Google’s humble beginnings to the invention of the Transformer, all the way to today's revolution of agentic AI workflows.
When Google was just getting started in that garage, we, along with so many other innovators, had to build a lot of the tools we needed ourselves. If you follow the journey from that garage, just look at all that we’ve built together and how far it’s taken us. Today, we’re providing the cloud and AI platforms that not only help the next generation of innovators build the future but build it bigger, faster, and more securely thanks to everything the cloud offers.
The same ethos that built Google, and later Google Cloud, is also what built this video. To create this keynote opening, we embraced our favorite philosophy: Google is customer zero. The exact same infrastructure and tools that power Google are the ones we offer to our customers in Google Cloud.


Storyboarding our idea using Gemini Enterprise and Nano Banana 2.
To put that to the test, we used our own generative AI tools to concept, storyboard, and produce the entire opening. And we did it in record time, which was only possible thanks to the latest advances in Google’s generative AI, infrastructure, and cloud offerings.
The agent platform for creativity
To usher in a new generation of agentic workflows, we knew we had to build this video using them. Every model we used to generate this music video was served through our Gemini Enterprise Agent Platform.
Instead of jumping between disconnected software, our creative teams and visual artist partners used the Agent Platform to seamlessly chain together tasks. Furthermore, every single prompt, image, and video render was powered by our state-of-the-art TPU infrastructure — the exact same high-performance, scalable compute that we make available to Google Cloud customers everywhere.
The creative process started with a blank page and a lot of ambition. We partnered with visual artists at Magnopus (our previous collaborators on The Wizard of Oz at Sphere) to dial in the creative direction, leveraging Google DeepMind’s (GDM) best generative models to help us brainstorm concepts, outline the narrative arc, and prototype the visual style.


Some of the storyboard sketches that Gemini Enterprise created, helping us quickly iterate on our ideas.
When it came time to storyboard, we turned to Nano Banana 2 for image generation. Instead of spending weeks sketching out scenes, the team could generate hyper-realistic concept art in seconds.
This allowed a large team of artists to ideate, prototype, and deliver finished assets across skill sets, film scenes, and time zones. The visual transition from a 1998 garage to a glowing, futuristic AI data center presented a particular creative opportunity to experiment with the latest in Veo technology and deliver a dynamic “wow” moment to open the show that wouldn’t just bring Google’s story to life but also make the technology itself come alive.
Here is an example of a prompt our team used to lock in the opening shot:
A medium-wide shot of an open-bay, light beige garage with a flat trimmed roof. The garage interior, illuminated by warm golden hour sunlight from the right, reveals a creative home studio featuring a ping-pong table and an old CRT monitor on the left, and walls covered in framed art, tech posters, and a detailed whiteboard. The right side has an acoustic guitar, a white PC tower, an audio amp, a huge LEGO tower, and a tall white cabinet.
Nano Banana 2 also helped the team establish a consistent visual language that would guide the rest of the production.
Generating the final cut with Veo 3.1
Once the storyboards were locked, we moved into production using Veo 3.1, our most advanced video generation model.
Traditional production for a CGI-heavy, era-spanning music video of this caliber would typically take months of rendering, animating, and editing. With Veo 3.1, we flattened the production cycle significantly. Then, using Nodey — Magnopus’s unified production canvas built on Gemini Enterprise Agent Platform — the team was able to guide camera movements, transitions, and the evolution of the environment.


With the help of AI, we could quickly identify images to represent dozens of our customers to feature.
Here is the prompt we used to create the climactic transition into the AI era:
Stop-motion animation with high-detail cardboard textures. The sedan enters from screen left and travels exclusively toward screen right. The car performs an abrupt, high-inertia stop. The rear wheels lift into the air briefly due to momentum before slamming back down. The trunk hatch opens naturally. The colorful server tower hops out from the trunk and lands perfectly upright.
From the initial concept in the Agent Platform to the final, high-definition render, we delivered this opening video in record time.
Putting Google’s research into action
To push the creative boundaries of the keynote video, we leaned into our close partnership with Google DeepMind.
We utilized GDM's world model, Genie 3, to generate dynamic, interactive environments inspired by our customers' ideas, as explored in concepts like the "Genie Carpet" and "Awaken Dreams" experiences for Google Cloud Next.




A place to rev up our dreams...
Furthermore, to achieve the highest visual fidelity, we successfully leveraged an experimental super-resolution (dubbed "superes") model from GDM's internal research, pushing the final video outputs all the way to a stunning 11k resolution.
What Will You Build Next?
The opening film is a celebration of the journey we’ve all been on. We went from a single server rack built out of Lego bricks to a globally distributed cloud architecture that empowers millions of developers, creators, and enterprises.
The barrier to entry for world-changing innovation has never been lower. By giving customers access to the same TPUs, agentic platforms, and generative models that power Google, we aren't just giving you a set of tools — we're hoping everyone feels they have a bigger garage to build from. We can’t wait to see what you build in it.



