Mishal Rahman / Android Authority
TL;DR
- Reddit users had early access to the “Gemini Omni” model within the Gemini app, billed as Google’s new AI video generation tool.
- Early tests showed impressive logical reasoning in math-themed videos, but more complex dinner scenes revealed common AI glitches, such as objects appearing out of thin air.
- Metadata suggests that this model is an extension of Google’s “Veo,” and user reports indicate that this model imposes severe usage limits on Google AI Pro plans.
Google I/O 2026 is next week, but we’re already getting a glimpse of what’s to come. Gemini users noticed new animations in the iOS and Android apps, but that’s not all. Now, a user is reporting that they’ve found references to a new “Gemini Omni” model that could be Google’s next-generation AI video model.
When Reddit user Zacatac_391 opened the Gemini app, he saw a pop-up suggesting Google’s “new video model,” Create with Gemini Omni (h/t 9to5Google).
you don’t want to miss out on the best Android permissions?
The user then tried to generate a video using the following prompt:
The professor writes a mathematical proof of a trigonometric identity on a traditional blackboard and explains the ongoing steps of the equation.
This prompt resulted in the following video output on Gemini Omni.
This video isn’t perfect, but it’s still impressive in several ways. Not only was the inference correct, the Omni model produced a video that was very realistic and fairly accurate. There are some parts that appear to be AI videos, such as some of the writing actions not matching the chalk’s output, and the chalk disappearing towards the end (showing inconsistency), but overall it’s still pretty good.
The user tried the classic “Will Smith Eating Spaghetti” test, which is used to benchmark AI-generated video output, but apparently was unable to get any output due to guardrails in the model. The user then tried a different prompt.
Can you create a scene where two men are seated at a table by the sea or on an outdoor deck in a high-end restaurant? They are seated at a circular table with a nice white tablecloth, complete with all the fancy accessories, all the spoons, forks, knives, fancy napkins, and centerpieces. One man is an outstanding person. A mature African American man in his fifties with a short beard and a confident posture, wearing a well-tailored suit. The other person is a friend, and they both approach the table to eat a plate of spaghetti. First, the men approach the table, exchange brief greetings, and then settle down and begin eating their spaghetti. Have a conversation between bites.
Here is the output from Gemini Omni for this prompt:
This video is a little less impressive. The spaghetti appears out of thin air on an empty plate, and each bite doesn’t generate enough chewing.
For reference, Reddit user janekm3 tried the same prompt in ByteDance’s Seedance 2 and got the following result:
The Seedance 2’s output seems more stable, but the video jitters on my end.
X’s Max Weinbach discovered metadata that suggested Omni was an extension of Veo. To learn more about Gemini Omni and how it improves Veo, we’ll have to wait for Google’s official announcement.
Thank you for joining our community. Please read our comment policy before posting.
