Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips
deepmind ·
Google DeepMind has released Gemini Omni 1.2 Flash, a new video generation model that supports native clips up to 15 seconds long, an increase from the previous 10-second limit. The update introduces multi-turn conversational editing, allowing users to refine videos through iterative prompts without regenerating entire sequences from scratch. Pricing has been reduced by 30%, with 720p video now costing $0.07 per second, making it more cost-effective for scalable production. Enhanced reference adherence ensures better consistency in character identity, product details, and brand colors across different camera angles and transitions. The model also features improved natural speech with clearer lip-syncing and distinct multi-speaker conditioning for more realistic dialogue scenes. Users can direct complex cinematic camera movements, such as continuous FPV flythroughs, with fewer jump cuts and steadier scaling. This release is generally available via the Gemini API, Google Cloud, and various Google applications like Flow, Google Vids, and Google Ads.
deepmind