To the point: Gemini Omni Flash enables video generation and editing through natural language conversation with consistent characters, realistic physics, and multiple editing rounds.
Google has introduced Gemini Omni Flash, a model that generates high-quality videos from arbitrary inputs and edits them through natural language commands. The system combines image, audio, video, and text inputs, leveraging Gemini’s knowledge of the real world.
Google has unveiled the first version of the Omni model family – Gemini Omni Flash. The system is currently being rolled out in the Gemini app, Google Flow, and YouTube Shorts. In the course of further development, additional output formats such as image and audio are also planned to be supported.
The model allows videos to be edited through natural language instructions, with changes building on each other while maintaining character consistency, physical correctness, and scene continuity. Users can modify specific elements or completely reshape entire scenes – for example, by adding objects, changing actions, or reinterpreting a recorded moment entirely. Edits are possible across multiple rounds without losing the original scene context.
A core feature is Gemini’s understanding of physics: the model processes forces such as gravity, inertia, and fluid dynamics with improved intuition. At the same time, it combines this physical knowledge with Gemini’s expertise in history, natural sciences, and cultural context to meaningfully connect scenic representations with storytelling.
Users can, for example, transform a video as a starting point: place a violinist in a different environment, make instruments invisible, or change camera perspectives – all through conversation. The system combines inputs from images, audio, video, and text into coherent outputs based on Gemini’s understanding of the world.
Source: ainews-dev.lumi-systems.io · Published 20 May 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.5.2.