Flow packages Veo access around scenes, not single clips

The shot

On 20 May 2025, Google introduced Flow, describing it in its own announcement as a tool "custom-designed for Veo, Google's state-of-the-art generative video model," aimed at scene- and shot-level filmmaking control rather than single-clip generation. The description here reflects Flow as a living product, retrieved 16 September 2026; Google's own materials date its public availability but continue to describe an evolving tool rather than a fixed, one-time release.

What the documents show

The announcement states Flow gives "direct control over camera motion, angles and perspectives," lets a user "seamlessly edit and extend your existing shots," and allows a generated character or object, called an "ingredient," to be reused with "consistency" across different clips and scenes. A separate DeepMind Veo page, archived 31 May 2025, corroborates the framing directly: "Flow. Built with creatives, for creatives. Flow enables you to create seamless cinematic clips, scenes, and stories using our most capable generative AI models." Both are the company's own descriptions of intended use, not confirmed film credits; Google names no specific finished production made with Flow in either document.

The workflow

The announcement describes a three-stage process: a user first creates or uploads "ingredients," reusable characters or elements, then composes them into individual clips, then references those same elements across multiple scenes using plain-language prompts. Google states Flow launched "available to subscribers of our Google AI Pro and Google AI Ultra plans in the U.S., with more countries coming soon," with the Pro tier including "100 generations per month" and the Ultra tier adding "highest usage limits and early access to Veo 3 with native audio generation." This tiered structure means the scene-persistence workflow described is gated by subscription level, not universally available at the same capacity.

What the tool does not change

Google's own announcement calls the product's current state "early days," language that does not claim Flow has closed the gap between assembling consistent ingredients and completing a finished narrative production. Continuity decisions, such as whether a reused character or shot genuinely serves a scene's story purpose, remain a filmmaker's judgment call that the cited documents describe supporting with tools, not replacing with automation.

• Does the "ingredient" consistency Google describes hold up across the specific number of scenes a project requires, beyond what the announcement demonstrates?
• What generation or usage limits apply at the subscription tier a production can actually access?
• Since Google calls this "early days," what has the product's documentation since added or changed as of the date it is checked?

Flow is documented as a packaging change as much as a technical one: Veo's existing generation is wrapped in a workflow aimed at scenes and reused elements, with the vendor's own materials naming the plan tiers and stated intent rather than an independently confirmed production pipeline.