Google Opens Veo 3 to Everyone: How to Try It

The AI video model that dominated tech conversations for weeks is no longer locked behind a premium tier. Google has moved Veo 3 into public preview, putting it within reach of every Google Cloud customer.

From Ultra-only to widely available

When Veo 3 first appeared, the only ways in were a Gemini Ultra subscription or Flow, the AI filmmaking platform Google showed off at its most recent I/O. That changed when the company announced the model could now be accessed as a public preview by all Google Cloud customers and partners through the Vertex AI Media Studio.

Veo 3 was unveiled at I/O, Google's annual developer conference. Its headline trick is generating video together with synchronized audio — something the field has struggled with for years. Ask it for a clip set inside a crowded subway car, for instance, and Veo 3 returns the footage along with AI-generated ambient noise that sells the realism. Google says you can even prompt it to produce the sound of human voices.

The model is also built to mimic real-world physics convincingly — the way water flows, the way shadows shift — which is part of why it appeals to filmmakers and fits Google's larger push to bring practical AI into creative work. Creators drive Veo 3 with plain-language text prompts and can dial in fine details, "from the shade of the sky to the precise way the sun hits the water in the afternoon light," as Google described it in a blog post.

Use cases — and the friction

Google pointed out that companies are already putting Veo 3 to work on customer-facing material like social ads and product demos, as well as internal content such as training videos. One CEO went as far as calling it "the single greatest leap forward in practically useful AI for advertising since gen AI first broke into the mainstream in 2023."

Google and its rivals have been pouring money into text-to-video tools, convinced this is one of generative AI's most valuable real-world applications. Synthesia, an AI avatar company, already sells the idea as a faster, cheaper way to produce enterprise content — letting executives clone their own likeness to deliver company-wide video messages, for example.

Among creative professionals, the reaction has been split. Some are optimistic about AI-assisted filmmaking: director Darren Aronofsky has entered a creative partnership with Google DeepMind, and a comparable arrangement exists between Lionsgate and the startup Runway. Others are pushing back hard against AI video creeping into creative industries. A Toys R Us ad made with OpenAI's Sora last year was mocked widely online, and entertainment-worker unions are organizing to defend jobs as the tools advance.

That backlash hasn't slowed the release pipeline. Amazon Ads recently rolled out its Video Generation tool across the US, and Meta is reportedly aiming even higher — automating every stage of ad production.

The technical hurdle Veo 3 clears

Veo 3 is one of the first models from a major developer that can synchronize generated video and audio at the same time. Meta's Movie Gen, released in October, is another. Tools like Runway's Gen-3 Alpha can attach AI audio to video, but only as a separate post-production step; producing both simultaneously demands the compute and resources of a player on Google's scale.

Here's a demo of Veo 3:

Why is this so hard? Video is a sequence of still frames, while audio is a continuous wave. Synchronizing them means building a model that works across both modalities and reconciles the very different timescales they live on.

A model fusing picture and sound also has to reason about variables like material, distance, and speed on the fly. A car at 100 mph sounds nothing like one at 10 mph; a horse on cobblestones sounds nothing like a horse on grass. Getting all of that right, together, in a single generation pass is the real achievement.