The tool adds instruction-based editing, improved lip sync, and new accessibility options in Google applications
Regarding sound, Google announced advancements in dialogue generation, with smoother and more natural voices and better lip synchronization. It also incorporates greater control for cinematic productions, such as FPV shots and dramatic camera pullbacks.
The new model will be incorporated into several Google applications and services, with different access conditions:
Gemini App: allows creating 15-second videos, although it initially requires a subscription to Google AI Ultra.
Google Flow: will be available for free users, who will be able to direct scenes and perform multi-step conversational edits. Google has not confirmed whether they will have access to all features.
Google Vids: users will be able to generate up to three 15-second videos per month at no cost. The tool will also include reference images and enhanced narration.
Gemini API: developers will be able to access the model through the Gemini API.
Google Cloud: companies will be able to test it from the Google Cloud console.
Google introduced Gemini Omni 1.2 Flash, its new model for video generation with artificial intelligence. The tool incorporates the ability to create clips of up to 15 seconds, improve visual continuity between scenes, and modify results through conversational instructions.
The update also adds advancements in dialogue generation, lip synchronization, and control over the duration of videos. The model will be available on various Google platforms, although access conditions vary depending on the service and type of user.
One of the main improvements of Gemini Omni 1.2 Flash is the duration of the videos. While version 1.1 Flash allowed generating clips of up to 10 seconds, the new model extends the limit to 15 seconds, representing a 50% increase.
Additionally, users can choose the duration of each video between three and 15 seconds, in one-second intervals. This allows adapting the clips to the needs of each project.
Google also improved the model's ability to follow reference images and maintain consistency of characters, objects, and settings. The company highlighted that the tool reduces visual alterations between shots, movements, and camera transitions.
This feature is particularly useful for generating multiple scenes of the same story. According to Google, visual elements can maintain greater continuity from the first to the last frame, even when extending a clip or interpolating between keyframes.
Another new feature of Gemini Omni 1.2 Flash is video editing through successive instructions. Users can make changes step by step without having to regenerate the content from scratch or repeat the initial instructions.
The model also promises to preserve elements that are not intended to be modified during edits. This facilitates the creation of different versions of the same video for different audiences.
