Two days before Google I/O 2026, a previously unannounced video model surfaced inside the Gemini app. The name on the card — Gemini Omni — has set off a week of speculation about what Google is actually shipping on May 19. Here is what the leak shows, what is still rumor, and what it means for AI video editing workflows today.
What Actually Leaked
On May 11, a UI string appeared briefly inside Gemini’s video tab. It read: “Create with Gemini Omni: meet our new video model, remix your videos, edit directly in chat, try templates, and more.”
A handful of users captured short generations before the model card disappeared. Two clips circulated widely — a professor writing trigonometric proofs on a chalkboard with legible handwriting, and a tightly detailed dining scene at a seaside restaurant. Reporting noted that each prompt consumed a large share of the daily Gemini Pro allowance, suggesting the model is expensive to run.
The leak is real. The product is not announced.
The Interesting Word Is “Remix”
The leaked feature list reads less like a typical video generation launch than like a pivot. Generation is one bullet. The rest — remix, edit in chat, templates — is editing language.
Early observers reported that the in-chat editing felt unusually capable: removing watermarks, swapping objects inside clips, and rewriting scenes via plain-text instructions. Generation fidelity on the leaked clips appeared strong but not necessarily ahead of Seedance 2 or HappyHorse 1.0.
If that holds, Omni would follow the same playbook Google ran on the image side last year with Nano Banana — lead with editing strength rather than raw generation quality. Editing is the harder problem and the one that gets used every day.
What Is Still Rumor
The model card is the only confirmed surface. Everything else is informed speculation:
- The name itself. Omni may be the public brand for a Veo-powered pipeline, a separate Gemini-native model, or an umbrella that covers image and video in one system.
- Tier structure. Reporting speculates a Flash and Pro split, mirroring the rest of the Gemini family. Not confirmed.
- API availability. No public API, no pricing, no rate limits. A model behind a closed waitlist is not a production tool.
- Clip length, audio, conditioning. None of the dimensions that actually decide whether a model belongs in a pipeline have been disclosed.
This is the same evaluation discipline we wrote about last week. Until Google publishes specs, the leak is a hint about positioning, not a roadmap.
What This Means for Tellers Users
Tellers is multi-model by design. The agent already orchestrates AI video creation across Runway Gen 4.5, LTX Video, Kling, Hailuo, Seedance 2, Veo 3.1, and others, picking the right tool per shot rather than locking the platform to one provider.
Omni — if it ships with a stable API we’ll integrate it on the day of the launch. In the following days our team will extensively test the model and give all the best practices insights to Tellers agent.
A few practical notes if you are planning around I/O:
- The in-chat editing pattern Omni is teasing is already how the Tellers agent works today, across multiple models. The interface bet is not new; the question is whether Google’s single-model implementation closes the quality gap on specific tasks like object swap and scene rewrites.
- The leaked cost signal — heavy credit consumption per prompt — matters. The model with the best demo is not always the model relevant for your productions.
- The most useful integration target is rarely the headliner. Look for the unflashy capabilities: identity preservation across edits, audio sync, and time to first frame.
What is Gemini Omni?
Gemini Omni is an unannounced Google video model that briefly surfaced inside the Gemini app on May 11, 2026. Leaked UI text described it as a model that can generate video, remix existing clips, and edit directly in chat. Google has not made an official announcement.
When will Gemini Omni launch?
Google I/O 2026 runs May 19-20, 2026. The leak window and Google's previous I/O cadence make a launch announcement likely, but until Google confirms it, treat the timing as expectation rather than fact.
How is Omni different from Veo?
It is not officially confirmed. The leaked positioning — remix, edit in chat, templates — suggests Omni leans toward editing and conversational workflows, while Veo has so far been positioned as a generation model. Whether Omni replaces Veo, sits alongside it, or unifies image and video is unclear.
Will Gemini Omni be available on Tellers?
Yes as soon as we get access to the API.
Two Days to I/O
The keynote is on May 19. By the end of the week, we will know whether Omni was the actual reveal, whether Veo 4 ships separately, and whether either has the API surface that makes it a real production tool.
If you want a workflow that already combines top AI video editing models with full timeline control open Tellers and start with one project. The model selection updates as the landscape moves; your workflow does not need to.