Early look: Gemini Omni generates realistic AI video in new leak

Google I/O 2026 is next week, and we’re already seeing glimpses of upcoming features. Gemini users have spotted the new animation on iOS and Android apps, but that’s not all. A user now reports spotting references to a new “Gemini Omni” model, which could be Google’s next-generation AI video model.

Reddit user Zacatac_391 opened their Gemini app and got a pop-up suggesting they “create with Gemini Omni,” Google’s “new video model” (h/t 9to5Google).

The user then tried generating a video with the following prompt:

A professor writes out a mathematical proof for trigonometric identities on a traditional chalkboard, explaining the step he is currently on in the equation.

This prompt gave them this video output with Gemini Omni:

This video is not perfect, but still impressive on several counts. The Omni model not only got the reasoning correct but also generated a video that is very lifelike and reasonably accurate. There are some tells that it’s an AI video, like some of the writing actions not matching the chalk output and the vanishing chalk towards the end (which shows inconsistency), but overall, it’s still pretty good.

The user tried the classic “Will Smith eating spaghetti” test used to benchmark AI-generated video output, but apparently couldn’t get an output due to the model’s guardrails. The user then tried out a different prompt:

Can you create a scene with two men at a table seaside at an upscale restaurant on outdoor deck seating. They are at a circular table with a nice white table cloth, and all of the fancy accessories, all the spoons forks and knives, fancy napkins, centerpiece. One man is Distinguished: A mature African-American man in his 50s with a short beard and confident posture, wearing a tailored, sophisticated suit, the other is is friend, both approaching the table to eat a plate of spaghetti. In the beginning the men approach the table, exchange brief niceties, and begin to eat the spaghetti calmly In between bites sharing conversation.

Here’s the output from Gemini Omni for this prompt:

This video is a little less impressive. The spaghetti appears out of thin air on empty plates, and there’s not enough chewing happening for the bites taken.

For reference, Reddit user janekm3 tried the same prompt with ByteDance’s Seedance 2 and got the following result:

Seedance 2’s output appears more consistent, though the video jitters on my end.

Max Weinbach on X found metadata suggesting that Omni is an extension of Veo. We’ll have to wait for Google’s official announcement to learn more about Gemini Omni and how it improves upon Veo.