n Monday, DeepMind, Google’s AI research lab, introduced the Veo 2, the successor to Veo which represents an advanced video-generating model based
on artificial intelligence technology. This new video AI model represents an important step in DeepMind’s AI journey because is able to create videos with more than two minutes long and with a resolution of up to 4K frames per second.
It should be mentioned that these capabilities of the Veo 2 can easily outshine the Sora AI video model due to the fact that OpenAI’s new video model can generate only a 20-second clip with a resolution of up to 1080p.
But, for now, it’s just a theoretical advantage because Google’s experimental video creation tool, called VideoFX limits these capabilities. Veo 2 is exclusively available in VideoFX and this creation tool limits the video creation to 720p resolution and a maximum video length of eight seconds, lower than OpenAI’s Sora.
Currently, Google’s VideoFX can be accessed just by a limited number of users, but the company is planning to expand the users’s number later this week.
So, the new video model Veo 2 is able to generate videos based on a text prompt or based on the given image. According to Google DeepMind, this new video AI model features different advanced capabilities compared with the first model and is able to generate a clearer video image.
Also, DeepMind stated that the Veo 2 video model can generate some more realistic motions and more fluid actions along with more realistic shadows and reflections.
Subscribe to our newsletter
It’s important to know that the Veo 2 video model has been trained on a vast number of videos because this is a common method in the AI journey of many artificial intelligence-based models. By analyzing various examples of data, they are able to learn patterns from the provided data in order to generate new content.
“Veo has been trained on high-quality video-description pairings. Video-description pairs are a video and associated description of what happens in that video.”, Eli Collins the VP of product at Google DeepMind stated.
In addition to Veo 2, DeepMind has also introduced a new upgrade for its commercial image generation AI model called Imagen 3. So, now all ImageFX users are able to create images with better quality and also in different styles such as anime, photorealism, or impressionism.
But the quality upgrade is not the only release, as Google made some changes to the UI interface as well, and now ImageFX users will see that their type prompts will be transformed into “chiplets” together with drop-down menus suggesting related words. These new “chiplets” allow users to improve their prompts or even choose from different auto-generated descriptions.
“This upgrade [to Imagen 3] also follows prompts more faithfully, and renders richer details and textures,”, DeepMind stated.
By
George King
•
December 17, 2024 9:15 AM
.webp)
.webp)
