Google Ups Pressure On OpenAI With Next-Gen Video Model Veo 2

Veo 2

Google claims audiences prefer its next-generation AI video model, Veo 2, to OpenAI’s recently released video generation tool, Sora Turbo.

Google stepped up the competition against OpenAI Monday with the launch of the newest version of its AI video generator, Veo 2. The release cements Google’s position as a frontrunner in the race to develop sophisticated AI tools that could revolutionize film and media as lawmakers and creative industries grapple with the implications of increasingly sophisticated artificial content. 

Veo 2 builds on Google’s earlier AI video model, Veo, which the company made available on its cloud platform Vertex AI last week. Compared to its predecessor, Google said Veo 2 can craft longer, higher-quality videos with better transitions, motion, and cinematic effects. 

The next-gen model can generate clips in resolutions up to 4K, Google said, and while 4K clips are limited to 8 seconds in length by default, but Google said they can be extended to two minutes or more. By comparison, the earlier Veo model can generate videos in resolutions up to 1080p and “can go beyond a minute.”

Google said Veo 2 also “brings an improved understanding of real-world physics and the nuances of human movement and expression” and “understands the unique language of cinematography.” This means users can be confident “Veo 2 will deliver” if they specify lenses, cinematic effects and genres.  

Leading the Pack

Veo 2 will compete in a growing cohort of video generators vying to attract content creators, filmmakers and marketers. This includes products from other US firms like OpenAI and Meta, the parent company of Instagram and Facebook, and Chinese giants like Alibaba and Kuaishou. 

Microsoft-backed OpenAI launched the latest version of Sora, Sora Turbo, last week after months of hype and a turbulent testing period. OpenAI said Sora Turbo can generate clips up to 1080p resolution–four times less than Veo 2’s 4k resolution–up to 20 seconds long, in a variety of aspect ratios. Runway, which makes one of the most popular AI video generators on the market, only produces clips at a resolution of 720p.

In a blog post announcing the release, Google described Veo 2 as “state-of-the-art.” It claimed the model bested industry leading rivals like OpenAI’s Sora Turbo, Meta’s Movie Gen, Alibaba’s Minimax and Kuaishou’s Kling in human-judged tests. In a head-to-head fight against Sora Turbo, for example, 58% of more than 1,000 human judges preferred Veo 2 overall. A similar proportion of testers thought Google’s model followed prompts better too. 

While an indicator of quality, the tests–and particularly internal tests–are not sufficient to determine which AI video generator is best. The metrics do, however, give a good, albeit rough indicator of overall quality and of Google’s leading position in the field. The closeness of the results also give a good indication of how close leading video tools are to one another and suggest there are likely to be few noticeable differences for casual users. 

Limited Access

Veo 2 is available through VideoFX, Google Labs’ video generation tool. Access to VideoFx is currently locked behind a waitlist, however, though Google says it is “expanding the number of users who can access it.” It is not clear when waitlisters will be granted access or under what conditions. 

The launch comes days after OpenAI resumed signups for Sora. The US firm temporarily limited availability and disabled signups for Sora mere hours after launch citing overwhelming demand. OpenAI said it plans to widen Sora availability next year. 

Google said it has plans to expand access to Veo 2 next year, naming YouTube Shorts and “other products” as beneficiaries. Google did not indicate a timeline for the expansion nor did it specify what other products it intends to include in the rollout. 

Creative Concerns and Fears Over Misuse

Concerns over the potential abuse of AI tools, as well as the impact on creative industries, have become one of the leading and most visible policy debates surrounding the use of generative AI. 

The ability to create convincing and realistic video, audio and images have sparked particular fears the tools will be misused to spread misinformation and facilitate abuse, such as creating child abuse material and sexual deepfakes. 

To help “reduce the chances of misinformation and misattribution,” Google said Veo 2 outputs would include the same invisible watermarks identifying them as AI-generated as its other video and image generation models. Google’s generative AI use policy prohibits using its tools to create child sex abuse or exploitation content, “non-consensual intimate imagery” or promote misinformation.  

Upping the Stakes

Veo 2 not only cements Google’s prime position as a leader in generative AI but also amplifies the urgency of discussions around its responsible deployment. The launch maintains Google’s momentum in AI, following last week’s launch Gemini 2.0, the latest version of its flagship model set to power products across its ecosystem. Google heralded the release as the start of the “agentic era” of AI, where AI agents can act on users’ behalf.

Subscribe to our newsletter for updates

Join thousands of media and marketing professionals by signing up for our newsletter.

"*" indicates required fields

Share

Related Posts

Popular Articles

Featured Posts

Menu