Gemini Omni explained: Google's AI model for video creation from any input
Google unveiled Gemini Omni, a new AI model capable of generating and editing cinematic videos from text, images, audio, or video inputs, at its Google I/O 2026 developer conference on May 19. The model incorporates realistic physics and supports conversational video editing while ensuring continuity across edits, with plans to integrate digital watermarking for transparency.
Google introduced Gemini Omni, a multimodal AI model designed for video creation and editing, during its Google I/O 2026 keynote on May 19. The model can process text, images, audio, and video inputs to produce cohesive cinematic clips, while also allowing users to edit scenes conversationally using natural language prompts. Gemini Omni maintains visual consistency across edits, including physics-based realism such as gravity and fluid dynamics, by leveraging Gemini’s broader reasoning capabilities. The tool enables users to modify videos by typing prompts, transform scenes or environments, alter actions, and refine edits incrementally without losing continuity. It supports reference-based generation, allowing users to upload drawings, photos, or footage to guide AI-generated scenes, styles, and transitions. Google also announced plans for users to create digital avatars using their own voice and appearance, with voice reference support launching initially. To promote transparency, all Gemini Omni-generated videos will include Google’s SynthID digital watermark, detectable via the Gemini app, Chrome, and Google Search. The company open-sourced SynthID text watermarking technology to encourage broader adoption among developers. Gemini Omni builds on Google’s Gemini 3.5 AI family, integrating scientific, historical, and cultural knowledge to enhance storytelling beyond pattern-matching generation.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.