Google is introducing another AI model to Gemini, specifically the Veo 2 video generator. It's rolling out to the Gemini app and website for paying customers. Veo 2 works like other video generators, taking text prompts and generating animations. Google claims it has a good grasp of real-world physics, especially human movement. It will be available in the model dropdown but its location could change. The rollout started today and it may take weeks for all Gemini Advanced subscribers to get access. Currently, it's limited to 8 seconds of 720p video that can be downloaded as an MP4 file with a monthly limit (not confirmed).
There's also an early way to play with Veo 2 through Whisk, a Google Labs experiment added last year. Whisk allows generating images with text prompts and example images and now has an "animate" option using Veo 2 with a 100-video monthly limit. However, the author wasn't overly impressed with Veo 2, as seen in an example where the Martian moon Phobos didn't have the expected effect. Google says Veo 2 is safe and marked with a SynthID digital watermark.