Google has updated the Gemini Omni 1.2 Flash model

Google has introduced an updated version of its multimodal model, Gemini Omni 1.2 Flash. The main focus of this release is on improving video content processing and optimizing API usage costs.

The updated model now supports the generation of video clips up to 15 seconds long. A key technical improvement is the neural network's ability to maintain consistency of characters and objects between frames. Users can now edit videos using text prompts, allowing for scene modifications without the need to regenerate the entire clip.

The developers announced a 30% reduction in the model's usage cost compared to the previous version. This change is aimed at increasing the accessibility of real-time video creation and processing tools for a wide range of users and business applications.

For practitioners, this means the ability to work more flexibly with video content while reducing computational resource costs. Integrating edits via prompts simplifies the iterative video creation process, reducing the time spent refining details in complex scenes.