Veo 3, Google's new video generation model that makes waves across the internet, is now available to everyone in the public preview, the company announced Thursday.
The tool was initially available only to Gemini Ultra subscribers and Flow, Google's AI-driven filmmaking platform revealed in the latest I/O. As of Thursday, it will be accessible as a public preview from all Google Cloud customers and partners in Vertex AI Media Studio.
Also: Best AI Image Generators of 2025: Gemini, ChatGpt, Midjourney, etc.
Announced last month at Google's annual developer conference, I/O, the VEO 3, can generate video with synchronous audio, a long-standing technical challenge in this field. For example, imagine urging your system to generate a video set inside a busy subway car. VEO 3 creates videos with ambient background noise generated by AI to add a sense of realism. According to Google, it can also encourage people to generate voices of their voices.
The model also specializes in realistically simulating real-world physics, such as the fluid dynamics of water and shadow movements, and promotes Google's broader mission to bring usable AI to the creative industry.
Users can create videos in VEO 3 via natural language text prompts and tweak instructions to change subtle creative details.
Use Cases – and Disadvantages
In a blog post, Google pointed out that various companies are actively experimenting with VEO 3 to generate customer-facing content, including internal material such as social media ads and product demos, such as training videos. One CEO described it as “the only biggest leap in AI that has been practically useful in advertising since Gen AI first invaded the mainstream in 2023.”
Also: When AI knocks, open source skills can save your career
Google and other leading AI developers have invested heavily in tools designed to generate videos from natural language prompts, and bet this will become a major practical case of generator AI. For example, AI Avatar Company Synthesia offers technology as a way to make enterprise content faster and less resources by allowing users like CEOs to replicate portraits and create company video addresses.
Get the top stories of the morning in your inbox every day Tech Today newsletter.
The reactions between creative experts are mixed. Some see positive possibilities for AI-supported filmmaking in the future. Acclaimed Director Darren Aronofsky has formed a creative partnership with Google Deepmind. Similar contracts have been fierce between Lionsgate and the AI startup runway.
However, others were critical of the increased intrusions of AI-generated videos across the creative industry. For example, a video ad for Toys R'Us, created using Openai's SORA last year, received a wide range of online ridiculous laughs. Unions of entertainment workers are organized to protect their jobs as technology evolves rapidly.
This does not prevent tech companies from building and releasing new video generation tools for marketers. Earlier this month, Amazon Ads announced the general release of its video generation tool in the US. Meta reportedly aims to automate every step of the advertising production process.
Major technical challenges
VEO 3 is one of the first models of leading high-tech developers that can synchronize video and audio generated by AI. The meta film gen, released in October, is another one. Several other tools, like Runway's Gen-3 Alpha, come with the ability to enable audio generated by AI to video in the post-production process, but simultaneous generation of two requires major power calculations and resources like Google.
Also: I chatted with 5 AI bots – these had the best conversation
Building AI models that can generate synchronized video and audio is a troubling technical challenge and an active research area across the AI industry. Both AI-generated video and AI-generated audio are clear technical challenges, and when they are combined, a whole new complexity is introduced. This is the Veo 3 demo.
https://www.youtube.com/watch?v=94kmlfyiao8
For one, the video is still a series of frames, while the audio is a continuous wave. Therefore, two synchronization requires a model that can work with these two modalities, explaining the very different timescales in which they work.
Also: Google Flow is a new AI video generator for filmmakers – how to try it today
AI models that combine video and sound must be able to dynamically explain variables such as material, distance, and velocity. A car driving at 100 mph is very different from one that travels at 10 mph. A horse walking on cobblestones is different from one walking on grass.
