Google has launched its most advanced video generator, Veo 2, which takes on the reigning champion, Sora Turbo. With improved understanding of real-world physics, Veo 2 produces videos with more detail and realism.
The AI model can generate videos up to 4K resolution and tackles common challenges, such as hallucinations like extra fingers. It also understands cinematography language, allowing users to specify genres, lenses, or angles for a desired effect.
Veo 2 is available to the public in Google Labs’ VideoFX tool, with an early access waitlist system that requires basic information from applicants. The model has been evaluated by human raters and emerged as the top-performing video generator against leading models, including Sora Turbo and Kiling v1.5.
Other recent updates include improved Imagen 3 image generation, which produces brighter and better-composed images with more diverse styles and textures. Additionally, Google unveiled Whisk, a tool that transforms input images into plushie-style stickers, leveraging its Imagen 3 and Gemini models to create detailed captions.
Source: https://www.zdnet.com/article/googles-veo-2-video-generator-takes-on-sora-turbo-how-to-try-it