Google has unveiled Veo 2, its latest AI video generation model, and early comparisons suggest it’s significantly outperforming OpenAI’s recently launched Sora video generator. While OpenAI made Sora available for general use earlier this month as part of its “12 days of Shipmas” campaign, Google’s Veo 2 remains in early preview—yet it’s already generating buzz for its superior performance.
Early testers and AI enthusiasts have been sharing side-by-side comparisons of both models using identical prompts, with Veo 2 consistently producing more accurate and physically realistic results. TED host and former Google employee Bilawal Sidhu demonstrated this with the prompt “Eating soup like they do in Europe, the old fashioned way,” which generated what he described as a “terrifying result” in Sora compared to a more impressive output from Veo 2.
Another striking comparison involved a prompt from YouTube star Marques Brownlee: “A side scrolling shot of a rhinoceros walking through a dry field of low grass plants.” Sidhu noted that while Sora “nails the look,” Veo 2’s prompt adherence and physics modeling were “leagues apart.” Perhaps most tellingly, EasyGen founder Ruben Hassid tested both models with a prompt for someone cutting a tomato with a knife. Veo 2 showed the knife cleanly slicing through the tomato while avoiding fingers, whereas Sora’s version disturbingly showed the knife cutting through the person’s hand.
Andreessen Horowitz partner Justine Moore spent several hours testing both models and concluded that Sora biases toward more motion, while Veo focuses on accuracy and physics, with a larger percentage of Veo clips being usable. The secret to Google’s apparent advantage may lie in YouTube, which Google owns and has used extensively to train its AI models. The company has been transparent about leveraging YouTube’s vast content library for AI development, though it prohibits others from doing the same.
Interestingly, The New York Times previously reported that OpenAI had trained its models using YouTube data despite this restriction, which YouTube CEO Neal Mohan stated would violate Google’s policies. Google’s DeepMind division has also previously tapped YouTube’s content to build an impressive music generator that was never publicly released. While these comparisons represent cherry-picked examples, the consensus among AI enthusiasts is clear: Google has delivered a superior AI video generation product that sets a new benchmark in the rapidly evolving generative AI space.
Key Quotes
Prompt adherence & physics modeling? Leagues apart. Sora nails the look, but…
Former Google employee and TED host Bilawal Sidhu made this observation after testing both models with identical prompts, highlighting that while Sora produces visually appealing results, Veo 2 excels in accuracy and realistic physics simulation.
I feel like comparing a bike vs. a starship
EasyGen founder Ruben Hassid used this vivid metaphor to describe the performance gap between Sora and Veo 2, particularly after testing a tomato-cutting prompt that showed Veo 2’s superior understanding of object interactions and safety.
Sora biases towards more motion, whereas Veo focuses more on accuracy / physics. And a larger % of clips from Veo are usable.
Andreessen Horowitz partner Justine Moore provided this analysis after spending several hours testing both models, identifying the fundamental philosophical difference in how each system approaches video generation and practical usability.
Our Take
Google’s apparent victory with Veo 2 reveals a crucial insight about AI development: proprietary data access may be the ultimate competitive moat. While OpenAI has pioneered many AI breakthroughs, Google’s ownership of YouTube—the world’s largest video platform—provides an unmatched training dataset that translates directly into superior model performance. This raises uncomfortable questions about market concentration and whether a handful of tech giants with massive proprietary datasets will dominate AI development. The irony that OpenAI allegedly violated Google’s policies by training on YouTube data, only to be outperformed by Google’s legitimate access to that same data, underscores how data ownership is becoming the defining factor in AI supremacy. As we move forward, regulatory scrutiny around data access and AI training will likely intensify, potentially reshaping the competitive landscape.
Why This Matters
This development represents a significant milestone in the AI video generation race, demonstrating that Google remains a formidable competitor to OpenAI despite the latter’s recent momentum. The superior performance of Veo 2 highlights the critical importance of training data quality and quantity in AI model development—Google’s access to YouTube’s massive video library appears to provide a decisive competitive advantage.
For businesses and content creators, this signals that AI video generation technology is rapidly maturing, with practical applications becoming increasingly viable. The focus on physics accuracy and prompt adherence suggests these tools are moving beyond novelty toward genuine utility. The competitive dynamics between tech giants also raise important questions about data access and AI development—Google’s prohibition on others using YouTube data while leveraging it themselves creates potential antitrust and fairness concerns.
This story also underscores the accelerating pace of AI innovation, with major breakthroughs occurring within weeks of each other. As these tools become more sophisticated, they’ll likely disrupt video production, advertising, entertainment, and education industries, making this a pivotal moment for anyone involved in visual content creation.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- Google’s Gemini: A Potential Game-Changer in the AI Race
- OpenAI’s Valuation Soars as AI Race Heats Up
- Sam Altman’s Bold AI Predictions: AGI, Jobs, and the Future by 2025
- The Artificial Intelligence Race: Rivalry Bathing the World in Data
Source: https://www.businessinsider.com/google-ai-video-veo-openai-sora-comaprison-2024-12