Gemini Omni Flash Preview

Gemini Omni Flash Preview is Google's video generation and editing model, introduced at Google I/O and rolled out to developers through the Gemini API in June 2026. Rather than a pipeline of separate tools, it folds video generation into a single multimodal Gemini model, producing clips with audio the prompt can direct, down to the style of the background music. Its defining trait is conversational editing, where each turn builds on the last and the model preserves everything you did not ask to change, with reference images available to pin specific subjects. Generations run up to 10 seconds in the current preview, which suits iterative creative exploration, quick concept videos, and edit-heavy workflows more than long-form output.