Limitations:
- Omni gives 10-second video generations presently, with longer durations coming quickly.
- Importing audio references and scene extension just isn’t but supported within the Gemini API for this mannequin.
- Video references as much as 3 seconds in period are accepted by the API schema however will not be appropriately processed by the mannequin right now.
- Character consistency when altering scenes or panning actions has some limitations however we’re working to make this higher.
Gemini Omni is accessible in public preview beginning as we speak in Google AI Studio and the Gemini API. To see the complete listing of mannequin capabilities and regional particular limitations take a look at the developer docs.
Construct with each fashions as we speak
The actual magic occurs if you chain these fashions collectively. Use Nano Banana 2 Lite as a high-speed picture era mannequin, then go that picture as a reference to Gemini Omni Flash to animate it right into a high-quality video. Plus, through the use of the Interactions API for these multi-turn experiences, you’ll be able to keep session historical past and context so customers can stack as much as three sequential edits.
That will help you get began we created a number of demo apps you’ll be able to remix that allow you to expertise how one can pair each Nano Banana 2 Lite and Gemini Omni Flash into one workflow.
