Google’s New AI Tool Can Turn Almost Anything Into A Video

Google has announced Gemini Omni Flash at I/O 2026, and it’s basically the company’s next big step in AI-generated media. Unlike earlier Gemini tools that mostly focused on text or images, Omni Flash is centered around video creation and editing.

The bigger idea here is pretty simple. Google wants users to create and edit videos by talking naturally instead of learning complicated editing software.

Google’s New AI Tool Can Turn Almost Anything Into A Video

Here are the biggest things announced with Gemini Omni Flash.

You Can Edit Videos Through Simple Conversations

Probably the most interesting part of Omni Flash is how editing works.

Instead of opening a timeline and manually adjusting clips, users can simply type instructions like changing an object, modifying an environment, or adding effects to a scene.

Google showed demos where sculptures turned into bubbles, mirrors became liquid-like, and environments kept evolving across multiple edits.

What’s interesting is that the AI remembers earlier prompts, so users can continue refining the same video without starting over each time.

It Can Generate Videos From Multiple Inputs

Omni Flash isn’t limited to text prompts.

Google says users can combine text, images, video clips, and audio references together while generating videos. So instead of describing everything manually, you can upload references and let the AI build around them.

The company also demonstrated stylized clips built from existing photos, music references, and motion samples.

At launch, audio support is mostly limited to voice references, though Google says broader audio input support is coming later.

Google Says The AI Understands Physics Better

One thing Google repeatedly focused on during the announcement was realism.

According to the company, Gemini Omni Flash has a stronger understanding of movement, gravity, fluid motion, and object interactions compared to earlier video AI tools.

That may sound like a small detail, but it matters a lot for AI-generated videos because unrealistic motion is usually the first thing people notice.

Google showed examples involving chain-reaction tracks, moving objects, and cinematic scenes where the AI tries to maintain more natural motion and continuity.

Users Can Create AI Versions Of Themselves

Google is also adding AI avatar support.

Users will be able to create digital avatars using their own appearance and voice, allowing the system to generate videos that resemble them.

The company says it’s still testing broader speech editing and voice modification features before making them widely available, which probably explains why Google spent extra time talking about safety and responsible AI use during the presentation.