Google DeepMind has released its Gemini Omni 1.2 Flash video-generation model with a 30 percent price reduction and extended clip lengths. The updated artificial intelligence model is now generally available across multiple platforms, including the Gemini API, Google AI Studio, Google Cloud Console, Flow, Google Vids, the Gemini app for Google AI Ultra subscribers, and Google Ads. Every generated video automatically includes an imperceptible SynthID watermark to verify its origin.
The model now produces video clips ranging from three to 15 seconds, an increase from the previous 10-second limit, with pricing dropping to $0.07 per second for 720p video. A key addition is conversational editing, which lets users modify backgrounds, lighting, or aspect ratios step by step across the entire 15-second duration without regenerating the whole clip. Additionally, Gemini Omni 1.2 Flash improves reference adherence by accepting character images, product photos, short motion clips, and audio cues to maintain consistent faces, clothing, packaging typography, and brand colors across different camera angles.
The update also delivers improved visual consistency with reduced morphing during camera transitions, alongside more fluid dialogue with clearer lip-sync and multi-speaker recognition. It supports complex camera movements like FPV drone flights and zoom-outs with fewer jump cuts.
This release matters because it lowers video generation costs while expanding creative control and consistency for developers and enterprises using Google's AI ecosystem.