Google launches Gemini Omni Flash 1.1 model focused on video generation

Google has released Flash 1.1, the new version of its multimodal artificial intelligence family, Gemini Omni. The model, which can process text, image, audio, and video data together, aims to provide more consistent results, particularly in video generation and editing processes.

According to the company's announcement, Gemini Omni Flash 1.1 has been made available to developers via AI Studio. One of the most notable changes in the new version is that videos are extended not just by looking at the last frame, but through a 10-second retrospective analysis.

With this method, the model is intended to better track scene integrity, maintain camera movements more naturally, and reduce transition errors, especially in dynamic scenes. In previous approaches, extension processes based solely on the last frame could lead to continuity issues.

SPEED AND EDITING TOOLS STAND OUT

Gemini Omni Flash 1.1 can extend videos in 10-second segments up to a total of 40 seconds. Users can also determine the start and end points of scenes by referencing three-second short video clips.

Google states that it has increased system speed by up to 60 percent by reducing the resolution from 720p to 360p during the draft generation phase. However, the ability to export completed content in 1080p or 4K resolution allows for a balance between fast previewing and high-quality output.

The new scene editing tools are highlighted as features that can accelerate AI-supported workflows, particularly in advertising, social media, and digital content production. The use of reference videos offered by the model aims to help create the desired composition in a shorter amount of time.