Google pushes – Google says Nano Banana users have generated more than 50 billion images so far, but the company is trying to stop creators from treating its tools as one-off experiments. At Google I/O, it unveiled an updated Flow that uses Gemini Omni to support chatting wit
Summer 2025 was when Nano Banana first appeared online—and it spread fast. Google says users have since generated more than 50 billion images with Nano Banana.
The tool’s appeal was immediate: it can edit existing photos, turning a familiar kind of media workflow into something an AI system can do in seconds. Google also has one of the industry’s leading video models and has gained significant traction in AI media generation.
But behind the numbers. Google is staring at a problem the rest of the AI media business also knows well: a lot of use is quick and casual. People ask Google’s Gemini app to generate an image or a short video clip and then move on. “These tools started as something you put a prompt into and then get an output out of. like a coin-operated [machine]. ” says Elias Roman. VP at Google Labs.
Now Google is trying to change what those tools are for. Roman says the company is building “a new Google product line that’s entirely dedicated to creativity. ” aiming at artists. filmmakers. and other professionals who would return to the product repeatedly across the creative process—not just to test an idea and scroll away.
Flow sits at the center of that push. Flow is an online video-generation tool built by Google Labs that Google unveiled at its 2025 developer conference. Earlier versions of Flow could generate images and 8-second video clips from text prompts.
At this week’s Google I/O, the company unveiled an update to Flow designed to go beyond standalone asset generation. The new Flow lets users chat with an AI agent to brainstorm and storyboard projects, develop scenes and character art, and ultimately generate videos.
For video generation, Flow uses Google’s new Gemini Omni model. The update is also where Nano Banana-style editing capabilities come into play for video, and where Google is trying to make the output feel less like a one-off clip and more like a coherent production.
One concrete example is character and look consistency. Flow aims to keep characters consistent and follow other stylistic guidelines. Roman points to a specific feature: Flow can preserve the same camera lens look across every shot without users having to specify it in every prompt.
“Flow is evolving from this prompt-in, content-out tool to an agent that’s a copilot at every step of the creative process,” Roman says.
That “copilot” vision is reinforced by what Google is now willing to let creators do with the platform itself. Roman says the company is allowing users to customize Flow—letting “creators basically vibe-code any tool or workflow they want.”
With the platform’s AI agent, creators can build tools that add video filters and compare two generated versions of a clip to spot differences. Roman adds that once a tool is made, it can be shareable—and even remixable by anyone else on the platform if the creator chooses to make it public.
This is where the story shifts from demo to strategy. Flow is still a Google Labs project. and turning it into a full-fledged product capable of competing with established creative-industry leaders like Adobe won’t be easy. Yet the way Google is repositioning Flow suggests the company is serious about moving past “playground” status.
Flow launched last year with promotional help from Donald Glover, leaning into a Hollywood-internal-experiment vibe. Roman says Google later realized that approach was too narrow. “We had an overly limited view of that,” he admits. “A year ago, we really thought it was for filmmakers.”
After Flow was released. Google saw marketers. architects. and even kindergarten teachers using it—people who may never have seen themselves as potential users of traditional creative suites built for media professionals. Roman says the company “didn’t appreciate” how wide the audience could be when tools are built to be AI-native.
“What we didn’t appreciate,” Roman says, “was that with truly AI-native tools, you’re able to serve an incredibly wide audience.”
The question now for Google is whether the updated Flow can turn that broad curiosity into something stickier—an environment creatives actually return to, day after day, until making images and videos looks less like a machine you feed once, and more like a workflow you live inside.
Google Nano Banana Flow Gemini Omni AI video generation AI image editing creativity tools Google I/O Google Labs Elias Roman Donald Glover