Google unveiled Veo, its “most capable” video generating model to date, promising to produce high-quality, 1080p resolution videos in multiple styles that exceed 60 seconds.
Although Veo is not yet available to the general public, a select group of producers will be able to test it out using the experimental VideoFX tool, and other interested parties can sign up for a waitlist.
Google released a series of videos that it said were created by Veo. These featured both whimsical and hyper-realistic aerial nature pictures, like alpacas dancing and a crocheted elephant strolling.
There were few clips with clear images of people or their faces and bodies. Certain clips felt a little too heavily edited, which was indicative of their AI source.
Google acknowledged that it would be difficult to maintain visual consistency in these situations, but it also stated that filmmakers and other producers will be providing feedback on Veo.
“It gives an unmatched degree of creative control—understanding prompts for all kinds of cinematic effects, like time lapses or aerial shots of a landscape—and accurately captures the nuance and tone of a prompt,” Google wrote in a post.
OpenAI, the company behind ChatGPT, gave a small group of testers access to their Sora text-to-video generator earlier this year.
A sample clip that OpenAI said Sora had produced showed a woman walking down a neon-lit metropolitan street. But problems like twisted body parts and movements made it obvious that AI was used to create the clips.
Regarding the subject of whether the company had used YouTube videos to train its video generator, OpenAI CTO Mira Murati was similarly unable to respond.
According to Google, Veo was in charge by design.
“Videos produced by Veo are subjected to safety filters and memorization checking processes that help mitigate privacy, copyright, and bias risks. We watermark videos using SynthID, our state-of-the-art tool for watermarking and identifying AI-generated content,” the tech behemoth stated.
