Google's Gemini Omni Video Model Spotted Ahead of I/O

The Gemini Omni video model was spotted via a Reddit capture ahead of Google I/O. The new Gemini UI showed a "Create with Gemini Omni" card, along with video remix, in-chat direct editing, and template features displayed together. Whether this was an accidental exposure or a limited A/B test is not yet clear.

A usage limits tab was also added to settings. Some users reported that video generation quickly consumed credits, suggesting that Google is extending the usage-based billing system it has been testing across Gemini to video as well.

Early evaluations were mixed.

• Generation quality was said to lag one tier behind ByteDance's Seedance 2.
• On the other hand, editing features such as watermark removal, object replacement, and rewriting scenes via chat instructions were strong for a first release.

This pattern resembles Gemini's image model Nano Banana. Back then, generation scores were also average but it was strong on editing leaderboards, and it later rose to become a frontier image system. This video model also appears to follow a strategy that prioritizes modality integration, and the results currently circulating are likely from the Flash tier, with a Pro variant reportedly in preparation as well.

The timing also lines up. Google I/O will be held on May 19-20, and this early exposure is read as a pre-signal to gauge pre-keynote reactions and refine messaging.