Google has introduced Veo 2, a new video AI tool to compete with OpenAI’s Sora. This comes shortly after OpenAI made Sora accessible to users. Google’s Deepmind team aims to offer an alternative with Veo 2, which can generate minute-long clips in 4K resolution. This AI is designed to better depict movements, whether of people, animals, or objects, and even the “camera” movements that record the scene. This functionality provides filmmakers with new possibilities to create scenes exactly as they envision.
Google conducted its own comparison tests between Veo 2 and other video AI tools. While these results should be taken with caution, as companies tend to highlight favorable outcomes, Veo 2 was reportedly preferred by participants for the quality of its video output and the accuracy of prompt execution. Competitors in the test included Meta Movie Gen, Kling v1.5, Minimax AI, and Sora Turbo. Google claims that Veo 2 performed particularly well against OpenAI’s Sora, with 67% of participants favoring Veo 2’s output and similar numbers praising its prompt accuracy.
Despite its strengths, Veo 2 has limitations. Users have noted that Veo 2 can generate better physics representations compared to Sora and other video AIs, such as cutting a tomato or eating spaghetti without mishap. However, Google acknowledges that Veo 2 is not without errors. Consistency issues arise in longer scenes, where background figures might suddenly disappear, merge, or appear out of nowhere. Physics challenges remain, as demonstrated by a clip where a fishing rod behaves unrealistically. Google plans to improve these aspects based on user feedback.
Access to Veo 2 is currently managed via a waitlist, but it is not yet available in all regions. Those attempting to join the waitlist may find that the service is unavailable in their country, with a message stating, “VideoFX is not yet available in your country.” Google intends to expand access to more users in the future.
According to TechCrunch, even users with access face restrictions. Currently, Veo 2 can only generate clips with a maximum resolution of 720p and a length of eight seconds, whereas Sora can produce clips with 1080p resolution and up to 20 seconds long. Despite these limitations, Veo 2 represents a significant step in video AI technology, offering new tools and possibilities for content creators.
In summary, Veo 2 is Google’s latest venture into video AI, competing directly with OpenAI’s Sora. While it shows promise in generating high-quality video content and accurately following prompts, it still faces challenges in scene consistency and physics representation. As Google continues to refine Veo 2, the tool is expected to become more widely available and capable, offering exciting opportunities for filmmakers and content creators worldwide.