Google Has Unveiled Its New AI Video Generator, Veo 2, and It Looks Impressive.
The rapid progress in AI-powered video creation doesn’t seem to be slowing down anytime soon, and Google has announced a major update to its in-house video maker in the form of Veo 2. It follows the release of Veo in May and the launch of OpenAI. Earlier this month, the company introduced its Sora AI video generator to the public.
Google says that Veo 2 beats Sora and other competitors in terms of user preferences and fast compliance (you can see the corresponding charts for yourself on the Veo 2 website ) and that the new model brings with it “increased realism and accuracy”, as well as better motion processing and improved camera control.
What Veo 2 can do
According to Google, Veo 2 has enhanced its video creation capabilities with an “improved understanding of real-world physics and the nuances of human movement and expression,” so it looks like you’ll see less weirdness in your clips.
The new model also has a better command of film and film language. This means that clues might include a certain movie genre, a certain type of lens, or a certain cinematic effect. Terms like “low-angle tracking shot,” “close-up shot,” or “shallow depth of field” are now supported, which can help generated videos more closely match the cue.
Videos can be created in resolutions up to 4K (upgraded from 1080p) and can now be up to several minutes in length. Some of the examples of videos Google shows look impressive: drinks are poured, dogs ride on inflatables in pools, cartoon characters talk to the camera, and scientists look through microscopes.
It’s not flawless, however, and Google admits that Veo 2 still has consistency issues in complex scenes. This is the problem with AI video generators in general—they don’t understand real-world physics and space, even though they can make a rough approximation of it based on huge amounts of training data. You’ll notice that none of these videos show anything leaving the frame and then coming back, most likely because the AI will forget about it or display it differently.
How to try Veo 2
The process for getting the Veo 2 is the same as for the first Veo model: you need to go to Google’s VideoFX web tool , which is part of Google Labs, and sign up for the waitlist. For now, you need to be in the US to try Veo 2 and be at least 18 years old.
Google promises to expand the availability of Veo 2 over the next year, including including it in YouTube Shorts for creators, but for now it’s not available in other tools such as Google Gemini (which offers access to the latest version of Imagen 3). art generator model).
Since I don’t currently have access to Veo 2, I can’t go into detail about the limitations. Google says that “each user is limited to a certain number of daily generations,” but it’s unclear how many videos you can create each day or how long they can be. OpenAI allows ChatGPT Plus users to create up to 50 videos per month, which are limited to five seconds in length and limited to 720p resolution.
Any videos you create will be accompanied by an “invisible SynthID watermark” that identifies them as being created by artificial intelligence – this is part of Google’s attempts to minimize the risk of Veo 2 being used to deceive viewers and spread misinformation (security concerns are also part of the reason for the slow rollout , says Google).