Google dropped Veo 3.1 into Flow on February 19, and the update is less about making prettier clips than about what you can do to them after they exist. For the first time, Flow users can add objects to generated scenes or (soon) erase things from them entirely, a pair of features that nudge the tool closer to something resembling an actual editing suite.
275 million videos and counting
The numbers are big: over 275 million videos generated in Flow since the tool launched about five months ago, according to Google's blog post. That figure spans both Veo 2 and Veo 3 generations, though, so it's tough to know how many of those represent serious creative use versus people typing "a cat riding a skateboard" and moving on. Google counts generations, not finished projects, which is the kind of metric that sounds impressive until you think about it for ten seconds.
Still, the adoption pace is worth noting. Flow launched at Google I/O as a paid tool for Google AI Pro ($19.99/month) and Ultra subscribers. Getting to nine figures in generation volume that fast suggests Google's distribution advantage is doing exactly what you'd expect it to.
Insert, Remove, and the editing question
The real story is the new editing layer. "Insert" lets you describe an element and place it into an existing scene. Google says Flow handles shadows and lighting automatically, making additions look natural. "Remove," which strips objects or characters and reconstructs the background, is listed as coming soon.
Thomas Iljic, Director of Product Management at Google Labs, framed these capabilities as bringing Flow closer to how filmmakers plan and shoot. That's a generous comparison. Dropping a fantastical creature into a generated 8-second clip isn't exactly the Avid workflow, but the direction is clear: Google wants Flow to be where you iterate, not just where you generate.
This matters competitionally. OpenAI's Sora 2, which launched late last month, still operates primarily as a generation tool. You prompt, you get a video, you prompt again if you don't like it. Flow's Insert and Remove features, rough as they may be in V1, give users a reason to stay inside the app longer and do actual creative work there. That's a different kind of lock-in than just having a better model.
Audio across everything
Veo 3.1 also brings generated audio to features that previously only produced silent clips. "Ingredients to Video" (which composites scenes from reference images), "Frames to Video" (transition generation between start and end frames), and "Extend" (which appends footage based on the last second of an existing clip) all now include synchronized sound effects, ambient noise, and sometimes dialogue.
The audio quality is experimental and Google says as much. Speech generation works better with longer text prompts, and clips depicting minors have speech muted entirely. Generated audio can fail quality checks, in which case your credits get refunded and you're left re-prompting. Not exactly the polished experience the marketing suggests.
Extend deserves a mention: it now supports pushing clips past two minutes by chaining extensions together, with each new segment built from the final second of the previous one. For establishing shots and ambient sequences, it works. For anything requiring narrative coherence across a 148-second video, you're still going to hit the limits of what prompt-based generation can manage.
Where this fits
Veo 3.1 outputs at 720p or 1080p, 24 frames per second, with 4K upscaling available via the Gemini API and Vertex AI. Pricing matches the previous Veo 3 tier for API users. Flow itself requires a Google AI Pro or Ultra subscription, and generations burn credits, with top-ups now available to Pro subscribers.
Google is also pushing Veo 3.1 across its wider ecosystem: the Gemini app, YouTube Shorts via Dream Track, and Google Vids for enterprise presentations. Everything generated carries a SynthID watermark, Google's approach to labeling AI content.
The February 19 update landed alongside a separate Google Labs announcement for Pomelli Photoshoot, a free tool turning product snapshots into marketing-grade images for small businesses. Between video editing, image generation, and music creation via Lyria 3, Google Labs is assembling a creative tools portfolio at a pace that might worry standalone AI media startups more than it worries Adobe.
Google I/O 2026 runs May 19-20 in Mountain View. Expect Veo and Flow to feature prominently, likely with whatever "Remove" looks like by then.
