The progress in generative AI is absolutely mind-blowing! Seeing models like Veo 2 create such high-fidelity videos, understanding complex prompts, and even film terminology, really highlights how far we've come.
It's a testament to the power of the underlying architectures, from diffusion models to transformers.
For anyone looking to understand these advancements, diving into the foundational concepts is key. Understanding how these models are trained, the role of attention mechanisms, and the nuances of different architectures is crucial. It's an exciting time to be learning about AI!
Today, we’re announcing Veo 2: our state-of-the-art AI video generation model which produces realistic, high-quality clips from scratch. 🎥. ✨
It’s able to:
▪️ Create videos at resolutions up to 4k
▪️ Understand camera controls in prompts, such as wide shot, POV and drone shots
▪️ Better recreate real-world physics and realistic human expression
Veo 2 can also understand film terminology including: genres, lens types, timelapse and more, meaning users can be more granular with the tool when prompting.
In head-to-head comparisons of outputs by human raters, it was preferred over other top video generation models.
Veo 2 is now available on VideoFX and we’ll be expanding access to this tool from next year. → https://goo.gle/veo-2