Google DeepMind is making Gemini Omni 1.2 Flash generally available for video generation. The model extends the maximum clip length to 15 seconds, lowers the price of 720p video, and lets users refine a clip through follow-up instructions.
Longer clips and a lower 720p price
Generated clips can now run from three to 15 seconds, compared with a previous maximum of 10 seconds. Google lists the cost for 720p video as $0.07 per second, a 30 percent reduction from its predecessor.
The model is rolling out through the Gemini API, Google AI Studio, Google Cloud Console, Flow, Google Vids, Google Ads, and the Gemini app for Google AI Ultra subscribers. Google says each generated video includes an imperceptible SynthID watermark that indicates its origin.
Edits can build on an existing clip
Google calls the step-by-step editing feature “Conversational Editing.” A user can ask for changes to elements such as the background, lighting, or aspect ratio during an ongoing conversation, rather than starting the whole video generation process again. The company says this editing works across the full 15-second clip.
Google describes a workflow that uses lower-resolution drafts to manage costs. A creator can begin at 360p or 720p, make targeted corrections with follow-up edits, and upscale the finished version to 1080p or 4K. This approach reserves the higher-resolution output for the version that is ready to use.
Reference inputs aim to keep details stable
The model accepts character images, product photos, motion clips of up to three seconds, and audio cues as inputs. Google says these references can help preserve faces, clothing, packaging typography, and brand colors when the camera angle changes.
One example in the announcement is a virtual try-on sequence. Using one model photo and product shots of three outfits, Gemini Omni 1.2 Flash creates a 15-second video with 360-degree turns. Google says the output maintains the model’s identity as well as each garment’s cut and drape.
Reference adherence matters when a video needs to show a particular person or product throughout changing scenes. If those details remain recognizable from shot to shot, users may be able to make variations without losing the identity of the subject. These are capabilities Google describes; the source does not provide independent performance measurements.
More natural motion, according to Google
Google also says the new model reduces visual drift and morphing during motion and camera transitions compared with its predecessor. For dialogue, the company describes more fluid, natural-sounding speech, clearer lip-sync, and improved handling of multiple speakers.
For camera work, Gemini Omni 1.2 Flash is designed to handle continuous movements such as FPV drone flights and dramatic zoom-outs, with fewer unwanted jump cuts, according to Google. Alongside longer clips and conversational editing, those changes target videos that need to maintain continuity across movement, speech, and scene details.