Google may be working on a new AI video model called Gemini Omni, according to a recent report. The leaked feature reportedly lets users generate, remix, and edit videos directly in Gemini. Early demos showed improved handling of text, motion, and object interactions. Metadata suggests Omni may build on Google's Veo model. Two sample generations reportedly used 86 percent of a daily AI Pro quota. Google has not confirmed the feature, but it could be announced at Google I/O 2026.