About Us

Who We Are

Google Flow AI is Google's dedicated creative studio for AI-powered filmmaking and visual storytelling. We sit within Google Labs — the part of Google that builds early-stage products and experiments that push the boundary of what is technically possible. Our team is made up of researchers, filmmakers, designers, and engineers who share one conviction: that the ability to tell a story through moving images should not be limited to those with access to expensive equipment, large production budgets, or years of technical training.

We are based primarily in Mountain View, California, with a growing creative research team in New York and a safety and policy group in London. Between us, we have shipped products used by hundreds of millions of people, directed short films screened at international festivals, and published research on generative video that is cited across the academic community. What brought us together under one roof was a shared frustration: the tools available to independent creators had not kept pace with the ambition of the stories those creators wanted to tell.

Google Flow AI launched publicly in 2025. In its first year, creators on the platform produced over forty million video sequences — ranging from personal passion projects and documentary shorts to commercial campaigns and experimental art installations. That number tells us we are solving a real problem. The feedback we receive every day from those creators tells us we still have a long way to go, and we would not have it any other way.

Where It Started

The idea for Google Flow AI grew out of a research project that began in late 2022 inside Google DeepMind. A small group of researchers was exploring whether a language model could be taught to reason about time — not just sequences of words, but sequences of visual events: the way light changes as a cloud passes over a building, the micro-expressions that cross a person's face in the half-second before they speak, the rhythm of a chase scene cut together from six different camera angles.

The early results were rough. The first internal demo took eleven hours to render twenty seconds of footage, and the outputs were far more useful as conversation starters than as finished work. But the underlying capability was genuinely new. For the first time, it was possible to describe a visual idea in plain language and receive something back that captured not just the content of the description but its emotional register — whether the light felt hopeful or foreboding, whether the pacing felt urgent or contemplative.

Over the following eighteen months, that research prototype was rebuilt from the ground up as a product. A dedicated team formed around it, bringing in expertise in video compression, real-time rendering, user experience design, and content safety. By the time Google Flow AI entered closed beta in early 2024, generation times had dropped from hours to seconds. The tool that had started as a research curiosity had become something creators could actually use in a working day.

What We Build

Google Flow AI is a filmmaking platform, not a single feature. That distinction matters to us. We are not building a button you press to get a five-second clip. We are building a full creative environment where you can develop a narrative from a one-line concept, generate and iterate on individual scenes, adjust pacing and visual style, stitch sequences together into a coherent edit, and export a finished piece ready for distribution.

At the core of the platform is our video generation model, Veo, which Google DeepMind developed and which Google Flow AI integrates directly into the creative workflow. Veo understands cinematic language — it responds to prompts that reference shot types, lighting moods, lens characteristics, and editing conventions, not just surface-level descriptions of what should appear on screen. Around Veo, we have built tools for scene planning, storyboarding, style consistency across a project, and collaborative editing so that teams can work on a production together in real time.

We also build infrastructure that most users never see: the content safety systems that run on every generation, the feedback mechanisms that route creator reports to human reviewers, the evaluation pipelines that let our research team measure whether a model update makes the platform better or worse for real creative use. None of that is glamorous, but all of it is necessary if a platform like this is going to operate responsibly at scale.

How We Think About Responsibility

Generative video is a genuinely powerful technology, and powerful technologies require clear thinking about how they can be misused. We do not treat safety as a compliance checkbox or a legal requirement to be satisfied at minimum cost. We treat it as a design constraint that shapes every decision we make about the platform — what it can generate, how outputs are labeled, what information we share with users about the nature of the content they are producing.

Every video generated on Google Flow AI carries a SynthID digital watermark embedded at the model level, making it detectable as synthetically generated content even after common post-processing steps like compression, cropping, or color grading. We believe that synthetic media should be identifiable as such, and we are committed to maintaining that standard even as detection becomes technically harder.

We prohibit the creation of non-consensual intimate imagery, content that sexualizes minors, deepfakes designed to deceive or defame real individuals, and content that spreads deliberate political misinformation. These are not edge cases we handle reluctantly — they are lines we draw clearly and enforce actively, with both automated systems and human review. When our systems make mistakes, which they do, we want to hear about it. Our content reporting channel is staffed by a dedicated team, not routed to a generic support queue.

We also think carefully about the creative community whose work trained the models we build on. We are active participants in ongoing industry conversations about compensation, attribution, and the rights of human artists in a world where generative models exist. We do not claim to have solved these questions, but we are not waiting for the industry to solve them for us.

The People Behind the Platform

Google Flow AI has around 200 people working on it at any given time, drawing on engineering, research, design, policy, legal, and operations disciplines. We also work closely with a community of creators — filmmakers, animators, documentary producers, music video directors, and visual artists — who participate in our extended beta programs, attend our quarterly creator summits, and give us the kind of honest feedback that internal testing rarely surfaces.

We run a Creator Fund that provides direct grants to independent filmmakers using the platform for projects with no commercial backing. In 2025, the fund supported 34 projects across 19 countries, with completed films screened at Sundance, SXSW, and the International Documentary Film Festival Amsterdam. Supporting that kind of work is not a marketing exercise. It is how we stay connected to the creative problems the platform should be solving.

Internally, we operate as a small team inside a large company — which means we move faster than most Google products, make decisions closer to the ground, and feel the weight of each choice more directly. The people who build Google Flow AI use it themselves, argue about it in the same Slack channels where they ship code, and care about whether it is genuinely good in a way that is difficult to manufacture.

What Comes Next

The roadmap for Google Flow AI is longer than we can fully describe here, but a few directions are worth naming. We are working on native audio generation — not just matching music to a clip, but generating dialogue, ambient sound, and original score that is compositionally tied to what is happening on screen. We are building more granular character consistency tools so that a face or a costume or a location can be held stable across an entire multi-scene project, not just within a single generation.

We are also expanding the ways creators can bring their own reference material into the platform — whether that is footage shot on a phone, a set of photographs that define a visual world, or an existing edit that a creator wants to reimagine. The goal is a platform where the boundary between what you shot and what you generated becomes a creative decision rather than a technical limitation.

None of what we are building is finished. Probably none of it ever will be in the way a physical object is finished. Filmmaking is a living practice, and a tool that serves it well has to keep moving. We are committed to moving alongside the people who use Google Flow AI, listening to what they need, and building toward a version of the platform that makes the story they are trying to tell genuinely easier to bring into the world.

If you want to be part of that, you are already in the right place.

Get in Touch

We are always open to hearing from creators, researchers, press, and partners.

General enquiries: [email protected]

Press & media: [email protected]

Creator Fund applications: [email protected]

Address: Google Flow AI, 1600 Amphitheatre Parkway, Mountain View, CA 94043, United States