DeepMind, Google’s AI-focused division, has announced a new model for video generation, Veo 2.
Today, we’re announcing Veo 2: our state-of-the-art video generation model which produces realistic, high-quality clips from text or image prompts. ?
We’re also releasing an improved version of our text-to-image model, Imagen 3 — available to use in ImageFX through… pic.twitter.com/h6ejHaMUM4
— Google DeepMind (@GoogleDeepMind) December 16, 2024
The neural network is capable of:
- creating videos with resolutions up to 4K;
- understanding camera control requirements in prompts;
- accurately replicating the laws of physics and realistic human facial expressions.
“In direct comparisons of results conducted by experts, Veo 2 was preferred over other advanced video generation models,” noted DeepMind.
In the experimental Google VideoFX tool for video creation, clips are currently available with resolutions up to 720p and durations up to eight seconds. The full functionality of Veo 2 is provided to a limited number of users, with a broader launch planned for next year.
Google staff have also enhanced the capabilities of the Imagen 3 image generator. The neural network has gained the ability to apply various art styles such as realism, fantasy, and others.
We’ve also enhanced Imagen 3’s ability to:
▪️ Produce diverse art styles: realism, fantasy, portraiture and more
▪️ More faithfully turn prompts into accurate images
▪️ Generate brighter, more compositionally balanced visuals→ https://t.co/MmdiHyrugR pic.twitter.com/L1l2Qzs5up
— Google DeepMind (@GoogleDeepMind) December 16, 2024
It also more accurately creates images from prompts and generates brighter and more compositionally balanced visuals.
In December, Google released the new model Gemini 2.0, the Deep Research tool for reasoning, and the AI agent Project Mariner, which can perform actions on a computer on behalf of the user.