Google, with the help of creative renaissance mastermind Donald Glover, demoed an AI video generator that competes with OpenAI's Sora. The model is called Veo, and while no firm release date or rollout plans have been announced, the demo appears to show a Sora-like product capable of producing high-quality, compelling videos. It seems that.
Google I/O 2024: What to expect after OpenAI's GPT-4o rollout
What's “cool” about VEO? “You can make mistakes faster,” Glover said in a video aired during Google's I/O 2024 livestream. “At the end of the day, all you really want, at least in art, is to make mistakes quickly.”
Credit: Google Promotions Mashable Screenshot
“Veo creates high-quality 1080p videos from text images and video prompts,” said Demis Hassabis, CEO of Google Deepmind, who spoke on stage at Google I/O in Hawaii. This makes Veo the same type of tool and has the same resolution as Sora at its highest settings. The slider shown in the demo shows that the length of the Veo video has been stretched to just over a minute, which is about the same length as the Sora video.
Veo and Sora are both unreleased products, so there's little point in comparing them in detail at this point. However, Hassabis says the interface will allow Veo users to “further edit their videos with additional prompts.” According to creators who have been granted access, this is a feature that Sora does not currently have.
mashable light speed
Tweet may have been deleted
What was Veo trained in? It's not clear at this time. About a month ago, YouTube CEO Neil Mohan told Bloomberg that if OpenAI were to use YouTube videos to train Sora, it would be a “clear violation” of YouTube's Terms of Service. Ta. But Alphabet, YouTube's parent company, also owns Google, which created Veo. In an interview with Bloomberg, Mohan strongly hinted that YouTube would feed content to Google's AI models, but only if users signed off.
Here's what we know about the creation of Veo: Hassabis says the model is based on Deepmind's Generative Query Network (GQN) research published in 2018, VideoPoet last year, Google's introductory video, and more. , which is the culmination of many similar projects from Google and Deepmind. The generators are Phenaki and Google's Lumiere, which was demoed earlier this year.
Glover's specific AI-based film production projects have not been announced. According to the I/O video, Glover says that “for a few years now he's been interested in AI” and reached out to Google, but it doesn't appear to be the other way around. “I got in touch with some people at Google and they were working on their own development, so we all got together,” Glover says in his video of Google's Veo Demo.
There is currently no way for the general public to try Veo, but there is a waitlist sign-up page.
topic
artificial intelligence
