[I/O 2026] Gemini Omni extends Gemini's multimodality to video generation

Google launches Gemini Omni Flash, a multimodal AI model generating realistic videos from text, audio, and images, now available for Google AI subscribers.

With Gemini Omni, Google is taking a major new step in multimodality: a family of models designed to create anything from any type of input (text, image, audio, video). The first member of this family, Gemini Omni Flash, is already available and initially focuses on video, but with a much broader ambition: to simulate the real world.

Omni Flash doesn't just generate videos from prompts. It understands the physical world and Gemini's cultural context to produce high-quality sequences rooted in reality. You can combine multiple sources (a photo + audio + text + an existing video) and get a coherent result.

Key Technical Highlights Emphasized by Google 100% Natural Control: everything is done through conversational language, instruction by instruction. You can refine a video by simply chatting (“make the character turn left,” “change the lighting to sunset,” “add a surprise reaction”). Consistency and Realistic Physics: characters, faces, clothing, and settings remain consistent from one scene to another. The model respects gravity, kinetic energy, fluid dynamics, and physical interactions (demos, for example, showed a ball in an ultra-realistic Rube Goldberg machine). World Knowledge: beyond photorealism, Omni draws on Gemini's vast knowledge (history, science, culture) to give meaning to images and actions.

Key Features Avatars: you can create a digital clone that replicates your appearance and voice. This feature is already operational but subject to strict identity verification to prevent abuse. Audio and Speech Modification: still in responsible testing phase at launch (Google prefers to proceed cautiously on this sensitive point).

Security and Traceability Each generated video automatically integrates Google's imperceptible SynthID watermark. You can verify it directly in the Gemini app, Gemini in Chrome, or via Google Search.

Immediate Availability Gemini Omni Flash is being rolled out today for Google AI Plus, Pro, and Ultra subscribers via: the Gemini app Google Flow (the video creation tool)

It will be available for free on YouTube Shorts and YouTube Create. API access for developers and businesses is planned for the coming weeks.