Google Launches Veo 2 With Improved Capabilities
In a blog post, the tech large detailed the brand new video-generation AI mannequin. Currently, the older Veo AI mannequin is out there to customers through Vertex AI and VideoFX, and it’s also getting used within the video background generator characteristic for YouTube Shorts. Now, the corporate launched Veo 2, which is alleged to have achieved state-of-the-art outcomes throughout inside testing, beating video fashions resembling Sora.
Introducing Veo 2, our new, state-of-the-art video mannequin (with higher understanding of real-world physics & motion, as much as 4K decision). You can be part of the waitlist on VideoFX. Our new and improved Imagen 3 mannequin additionally achieves SOTA outcomes, and is coming at the moment to 100+ nations… pic.twitter.com/FtwNKyicEc
— Sundar Pichai (@sundarpichai) December 16, 2024
The tech large claimed that Veo 2 has made enhancements within the element and realism of the generated movies. Additionally, the AI mannequin can now perceive the language of cinematography. This means customers can specify a style, lens, cinematic results, and digital camera actions and Veo 2 can generate movies to the precise specs.
Highlighting an instance, the publish stated Veo 2 can perceive advanced prompts resembling “a low-angle monitoring shot with 18mm lens that glides via the center of a scene”. The point out of the lens will let the AI know that the consumer needs to generate a wide-angle shot which the lens is understood for, the corporate added.
Additionally, Veo 2 can generate movies as much as 4K decision and may lengthen video period to “minutes” in size. The firm didn’t specify the utmost period of movies that may be generated utilizing the AI software. The tech large additionally claimed that the video mannequin reduces the situations of hallucination and undesirable particulars resembling additional fingers, morphed limbs, or sudden objects will likely be seen much less incessantly.
On security, Google acknowledged that the corporate has added the native SynthID watermark on generated movies to assist establish AI movies from actual movies, lowering its misuse to create deepfakes. The firm has additionally deliberate a gradual rollout of the AI mannequin to manage its entry and enhance its high quality and security earlier than international availability.
Currently, Veo 2 is out there to pick out customers in VideoFX through Google Labs. Those taken with attempting out the software are requested to join the waitlist. Further, the tech large plans to increase Veo 2 to YouTube Shorts and different merchandise subsequent yr.
Imagen 3 Also Rolling Out Globally in ImageFX
Alongside the introduction of Veo 2, the tech large additionally rolled out the AI picture technology mannequin Imagen 3 in ImageFX to greater than 100 nations. The newest iteration of the mannequin has acquired a number of upgrades, together with the flexibility to render extra various artwork types, better accuracy, and better-composed pictures.
Imagen 3 additionally shows higher immediate adherence and may render extra particulars and textures in pictures, the corporate claimed. It is alleged to additionally excel particularly types resembling photorealism, impressionism, summary, and anime.
