Google announced the launch of the AI video generation model Gemini Omni 1.2 Flash on November 5, local time. The model supports native generation of videos up to 15 seconds long and introduces multi-turn conversational editing capabilities. Compared to the previous generation model, the maximum video length has increased from 10 seconds to 15 seconds, a 50% increase. Users can set the generation duration in 1-second increments, ranging from 3 seconds to 15 seconds.
Google stated that the new model enhances adherence to reference content, better maintaining consistency of characters, objects, props, scenes, and environments across multiple shots, reducing visual drift or form changes during motion and shot transitions. This capability also applies to interpolation between keyframes at the beginning and end, as well as extending videos based on existing clips.
In terms of editing, Gemini Omni 1.2 Flash allows users to incrementally modify existing results through multi-turn conversations without needing to resubmit complete requests or regenerate videos from scratch. During the modification process, the model strives to retain elements that users do not request to adjust, facilitating the creation of multiple versions for different audiences.
Gemini Omni 1.2 Flash Offers Multi-Turn Conversational Editing Features
Google also indicated that the new model improves voice generation effects, producing smoother, more natural dialogue that matches lip movements, while enhancing control over visual formats such as first-person flight shots and wide-angle shots. In terms of usage, Google AI Ultra subscribers can generate 15-second videos through the Gemini app; developers can access the model via the Gemini API, and enterprise users can utilize it in the Google Cloud console.
Free users can try scene control and multi-turn editing features in Flow, but Google has not clarified whether they will have access to all new features. Additionally, Google Vids will also offer a free 15-second video generation feature, but it remains unclear whether other improvements will be available to free users.

