Gemini AI’s new trick: turning photos into short video clips

Google is making it easier for people to create short videos with AI. Starting this week, subscribers to its Gemini AI Ultra and Pro plans can turn photos into eight-second video clips directly through the Gemini chat interface.

The feature, which is rolling out first on the web version of Gemini and then to the mobile app later this week, lets users upload a photo and add a text description to generate a short video with sound. The videos come out as MP4 files in 720p landscape format.

This tool isn’t brand new — Google showed it off back in May at its annual developer conference as part of Veo 3, its latest video-generation model. Until now, it was only available through Flow, Google’s standalone filmmaking tool, but now it’s baked into Gemini, making it easier for more subscribers to access.

The update comes as Google tries to keep up with competitors like OpenAI and Runway, along with Chinese tech giants Alibaba and Kuaishou, all racing to launch their own AI video tools.

Google claims it has strict guidelines in place to prevent misuse. For example, Google says you can’t create videos using images of celebrities, politicians, or other public figures, and it won’t produce content that promotes violence or bullying.

That said, the tool is far from perfect. Tests by Bloomberg found that when users tried to create talking videos from photos, the AI sometimes altered people’s faces or even changed their race. Simpler prompts — like making a plant sway in the wind or animating a cat to look like it’s talking — worked much better. But when asked to make someone breakdance, it ended up just making them wave to the camera.

A Google spokesperson told Bloomberg the AI isn’t programmed to change someone’s appearance and that the technology is still developing. Right now, it works best for animating nature scenes, drawings, and objects, but improvements to face animations are on the way in future updates.