Google has introduced Gemini Omni, a new range of generative AI models that can create and edit videos using a combination of text, images, audio, and existing clips.
Gemini Omni represents a significant advancement in Google’s consumer AI tools, expanding beyond last year’s image-centric Nano Banana to offer complete multimodal video generation. Users have the flexibility to input various forms of content such as photos, drawings, voice recordings, and written prompts, allowing Omni to seamlessly blend them into a cohesive video. The editing process is interactive, enabling creators to request changes like altering backgrounds, adding objects, or adjusting actions while maintaining consistent characters and realistic scenes throughout multiple iterations.
Google emphasized that Omni is rooted in Gemini’s practical knowledge, enabling it to consider historical, scientific, and cultural contexts rather than producing visually appealing yet nonsensical footage. The model showcases an enhanced intuitive grasp of concepts like gravity and fluid dynamics, resulting in more lifelike movements.
With a digital avatar feature, users can create a personalized avatar that resembles and sounds like them. While Google is still refining safeguards for editing speech and audio content, all generated output includes an undetectable SynthID watermark. Verification tools can be accessed through the Gemini app, Chrome, and Google Search.
Gemini Omni Flash is currently accessible to Google AI Plus, Pro, and Ultra subscribers worldwide, as well as through Google Flow. It will also be integrated into the YouTube Create app this week. The initial version, Gemini Omni Flash, is available to paid subscribers and YouTube Shorts users at no charge. Developer and enterprise access through APIs is scheduled for the near future. Google plans to expand the Omni family to include image and audio generation in the future, leveraging its inherently multimodal architecture.
