Google DeepMind released Gemini Omni 1.2 Flash on November 5, an update to its native video generation and editing model. According to Google's announcement, the model produces clips of up to 15 seconds, up from 10 seconds with Omni 1.1, with duration adjustable in one-second steps from 3 to 15 seconds.
Google says the model holds characters, objects and scenes more consistently from the first frame to the last, follows reference inputs more closely, and produces camera moves with "significantly fewer unwanted jump cuts." It accepts character, product or style images, motion clips of up to three seconds, and audio cues as references. Speech output has improved lip-sync and multi-speaker conditioning.
The release also adds multi-turn editing. Users can swap a product or background, change lighting or weather, or switch aspect ratio without regenerating the whole clip. Google recommends drafting at 360p or 720p, iterating with edits, and then upscaling to 1080p or 4K.
API pricing drops 30% from Omni 1.1 to $0.07 per second for 720p video. The model is available in the Gemini API through Google AI Studio and is generally available on Google Cloud with Provisioned Throughput. It is also rolling out to Google Vids at no additional cost, to Flow, to Google Ads video campaigns, and to Google AI Ultra subscribers in the Gemini app. All outputs carry an imperceptible SynthID watermark.
Google did not publish quantitative benchmarks for the release and said it is still working on fast sports action, rapid script changes and complex physics. For developers, the practical change is cost and control: a cheaper per-second rate combined with targeted edits means fewer full regenerations when iterating on a shot.