Google has introduced significant upgrades to its video and image generation models, with Veo 2 and Imagen 3 achieving state-of-the-art results.
The latest versions of these models are now available in various tools, including VideoFX, ImageFX, and a new Labs experiment called Whisk.
Veo 2, the updated video model, can create high-quality videos in different styles and subjects. It has been shown to outperform leading models in head-to-head comparisons by human raters.
One of Veo 2’s key features is its ability to understand cinematography terminology, allowing users to specify genres, lenses, and cinematic effects. The model also produces more realistic outputs with fewer unwanted details.
To ensure safety and responsible development, Google has been carefully measuring the availability of Veo 2 and identifying areas for improvement.
The new capabilities are rolling out in VideoFX and YouTube Shorts next year. In addition, a SynthID watermark is included in all AI-generated outputs to help reduce misinformation and misattribution.
Imagen 3, the image generation model, has also been upgraded to generate brighter, better-composed images with improved art styles accuracy.
The latest Imagen 3 model is now available globally in ImageFX, while Whisk allows users to input or create images to visualize their ideas, then remix them to create unique content.
Source: https://blog.google/technology/google-labs/video-image-generation-update-december-2024