GOOGLE LAUNCHES GEMINI OMNI: THE AI THAT TURNS ANYTHING INTO VIDEO, AND VICE VERSA Google announced Gemini Omni at I/O this month and it is genuinely unlike anything the company has shipped before. This is not another text model with image understanding bolted on. Omni is a natively multimodal system — one neural backbone trained simultaneously on text, images, audio, and video — that can accept any combination of those inputs and generate high-quality video output synchronized with realistic audio. Paste in a photo, a script, and a sound file. Get a short video out. Edit an existing clip by typing what you want changed. Swap characters. Alter the background. Change the camera angle. Add sound effects tied to on-screen events. Early tests showed the system generating a short video of a stuffed animal rafting down a river and creating a photorealistic deepfake of a journalist standing in front of the Eiffel Tower. That range gives a sense of both the creative power and the potential for misuse baked into a single tool with no friction at all. Google calls this unified native multimodal architecture, and it is a direct answer to OpenAI’s Sora — which OpenAI famously announced to enormous fanfare in 2024, then abruptly shut down in March of this year, leaving studios and developers stranded. Gemini Omni Flash is already rolling out to Gemini Ultra subscribers, YouTube Shorts creators, and Google Flow users. The AI video race just became very real, very fast. Keywords: Google Gemini Omni, AI video generation, multimodal AI model, Google IO 2026 AI