Limitations:
- Omni offers 10-second video streams at the moment, which will soon take longer.
- Loading audio and scene extension references is not yet supported in the Gemini API for this model.
- Video references up to 3 seconds long are accepted by the schema API but are not processed correctly by the model at this time.
- Character consistency when changing scenes or pan movements has some limitations but we are working on making this better.
Gemini Omni is available in public preview starting today in Google AI Studio and Gemini API. To see a full list of model capabilities and region-specific limitations check the developer's documentation.
Build with both models today
The real magic happens when you combine these models together. Use the Nano Banana 2 Lite as a model that produces a high-speed image, then pass that image as a reference to the Gemini Omni Flash for high-quality video. And, by using the Interactions API for this experience multiple times, you can save session history and context so users can stack up to three consecutive edits.
To help you get started we've created a few remixable demo apps that let you see how you can pair both the Nano Banana 2 Lite and Gemini Omni Flash in a single workflow.
