Google has unveiled its new generative artificial intelligence model Veo. The company demonstrated it at the Google I/O conference and stated that it is capable of creating 1080p videos lasting about a minute from text queries.
Veo has an "in-depth understanding of natural language" that allows the model to understand cinematic terms such as "time-lapse" or "aerial photography of the landscape."
Google claims that the resulting videos are "more consistent and coherent," depicting more realistic movement of people, animals, and objects in the frame. The company did not specify what exactly is less consistent than Veo video, perhaps referring to Google's previous models, or perhaps to OpenAI's Sora.
Users will be able to make edits to already generated videos. Demis Hassabis, the head of Google Deep Mind, said at the conference that the company is already working on creating storyboards for videos and extending the timing.
Veo is available for users within the VideoFX service, which, unfortunately, is not yet available in Ukraine. Google also plans to add some of its features to YouTube Shorts "in the future."